doi
stringlengths
10
10
chunk-id
stringlengths
1
4
chunk
stringlengths
1
1.66k
id
stringlengths
10
10
title
stringlengths
19
148
summary
stringlengths
345
1.92k
source
stringlengths
31
31
authors
list
categories
list
comment
stringlengths
4
284
journal_ref
stringclasses
14 values
primary_category
stringclasses
16 values
published
stringlengths
8
8
updated
stringlengths
8
8
references
list
2101.00027
96
stored in HTML, we use the html2text package to extract the text from the post. We order each document by listing the title, url, sub-title, and author at the top. Top-level comments are delimited by " \n----\n " and sub-comment chains are delimited by " \n~~~\n ". We include author and extracted text for each comment. C.19 YouTube Subtitles We construct the dataset in three stages: 1.We build a large list of search terms by prompting a GPT-3 model with a manually selected list of queries, manually filtering the responses, and repeating this process iteratively until a suitable size is reached. The list of terms is centred around, but not limited to,educational topics. 2.We use requests-html to gather a list of 1000 Youtube video IDs for each search term, and deduplicate the resulting video ids across search terms. 3.We use YoutubeTranscriptApi22to gather all human generated closed captions for every available language for each video. To align each language in parallel, we split the captions for each language into parallel minute-long sections by timestamp, and arrange each language in a random order within these sections, appending the language as a
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
97
the captions for each language into parallel minute-long sections by timestamp, and arrange each language in a random order within these sections, appending the language as a header to each minute-long section to provide context. If only a single language is available, the output is just the subtitles, with no header appended. In total, subtitles for 173,651 videos were gathered. C.20 PhilPapers The PhilPapers (PP) are indexed using OAI-MPH, the Open Archives Initiative Protocol for Metadata Harvesting. As such, the first step to collect the data is to get the XML for all links. This was done using pyoaiharvester .23 From that, each publication is downloaded. Some entries do not exist, or have been removed by the authors. Papers with text are extracted using pdfbox , and papers with non-machine readable text are ignored. Non-English language publications are kept, and the metadata reflects the language reported by the OAI-MPH XML. The text is filtered with pdf_filter.py from PDFextract, and we discard any papers with less than 1000 characters.24 C.21 NIH Grant abstracts: ExPORTER The NIH provides a bulk-data repository for
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
98
and we discard any papers with less than 1000 characters.24 C.21 NIH Grant abstracts: ExPORTER The NIH provides a bulk-data repository for awarded applications through the ExPORTER service covering the fiscal years 1985–present. These data come from the NIH, but also other other Health and Human Services agencies (ACF, AHRQ, CDC, HRSA, FDA), and the V A. Additionally, the NIH 22https://github.com/jdepoix/ youtube-transcript-api 23https://github.com/vphill/ pyoaiharvester/ 24https://github.com/sdtblck/PDFextract 26 provides a legacy data format named CRISP for awarded applications during the fiscal years 1970– 2009. We merged both the ExPORTER and CRISP data to form a consolidated dataset of awarded applications. Entries were deduplicated based off their application ID, and excluded if their abstract text was missing or too short. Small grants, especially administrative ones, consisted solely of short boilerplate. For this reason, we further deduplicated on abstract text. All grants types were considered, including new applications (Application Type Code 1) and renewals (Application Type Code 2) as the
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
99
abstract text. All grants types were considered, including new applications (Application Type Code 1) and renewals (Application Type Code 2) as the text differed enough to provide novel input. The text was then minimally parsed to remove administrative boilerplate, (ex. most old awards contain some variation of “description: (provided by applicant)"). In total, there were 939,668 grant application abstracts added. C.22 Enron Emails To extract the data, we used the mailparser package25to extract the body of each email as a document. D General Data Processing This section discusses any processes applied across multiple datasets. To combine the constituent datasets, we iterate until the size of the output dataset is the desired size, drawing documents from datasets at random, weighted by the number of documents in each dataset times the number of epochs desired on that dataset. Because the number of documents involved is high, by the law of large numbers, the number of copies of each dataset present in the Pile is approximately equal to its epoch count. Shuffling a dataset posed a major problem due to our limited memory and computational budget. We follow Hardin (2018), a method descended from Rao (1961), and interleave our output to produce 30 output piles.
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
100
our limited memory and computational budget. We follow Hardin (2018), a method descended from Rao (1961), and interleave our output to produce 30 output piles. We hold out approximately 10GiB of data from the Pile, of which 2GiB are used to create the validation and test splits, and the remainder is held in reserve. From the training set, we remove any 25https://github.com/SpamScope/ mail-parserelements that are also present verbatim in any of the held out data, to prevent leakage. D.1 Weights Similar to Brown et al. (2020), we increase the weight of certain components such that the number of epochs elapsed on data we consider high quality is greater than one. Our choice of weights was primarily informed by the source of the data and the size of the dataset; we attempted to upweight academic texts the most, which we felt provided the highest quality data, as well as smaller sets, such that they would have a more pronounced impact on the data. We strictly disallowed any data more than 3 epochs and avoided having any data with more than 2 epochs. D.2 Deduplication Due to memory constraints we did not perform Pile wide de-duplication. Instead, de-duplication
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
101
than 2 epochs. D.2 Deduplication Due to memory constraints we did not perform Pile wide de-duplication. Instead, de-duplication was performed at the document level within OpenWebText2 and Pile-CC as those sets were the most likely to contain duplicate documents. The same technique was used for both OpenWebText2 and Common Crawl—MinHashLSH with the Python Datasketch library.26We used 10 hash functions for each Minhash and an approximate Jaccard similarity of 0.5. This produced a duplicate rate of 28% in OpenWebText2 and 26% for Common Crawl. The main challenge here was computational, leading us on a journey through the various LSH persistence options. A simple quadratic Minhash comparison of all documents would have taken several hundred thousand years, motivating the use of LSH. Initially, we did not have sufficient RAM for inmemory LSH and chose to use the Cassandra backend when de-duplicating OpenWebText2. This was reasonably fast, but the same method resulted in a corrupted database about3 4of the way through processing Common Crawl. After the Cassandra corruption, we briefly tested the experimental Mongo implementation; however this was quite slow due
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
102
4of the way through processing Common Crawl. After the Cassandra corruption, we briefly tested the experimental Mongo implementation; however this was quite slow due to the nature of Mongo itself. In the end, we ran in-memory LSH on a machine with enough RAM for Common Crawl, taking several days. 26https://github.com/ekzhu/datasketch 27 D.3 Downstream Validation Leakage To avoid leakage of data from downstream evaluations, recent work (Radford et al., 2019; Brown et al., 2020; Shoeybi et al., 2019) has removed any data in the training set that may overlap with the evaluation metrics. We decided not to perform any such removal, because it is impossible to anticipate all potential downstream evaluation metrics, and so any particular selection of metrics would inevitably either become obsolete as the choice of benchmarks in the field changes, or potentially hinder the development of new benchmarks for models trained on Pile. For models trained on Pile and evaluated on metrics other than Pile’s own validation and test sets, we encourage authors to remove overlaps between Pile and the validation data of these additional downstream evaluations. We do not anticipate that such leakage removal will hurt model performance, as the validation sets of most benchmarks are very
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
103
and the validation data of these additional downstream evaluations. We do not anticipate that such leakage removal will hurt model performance, as the validation sets of most benchmarks are very small in relation to the size of the Pile, and so choosing to evaluate on more metrics will not be a disadvantage for any model. E Investigating data E.1 13-Gram Analysis As part of our exploratory analysis, we calculated the counts of all 13-grams across Common Crawl. We chose n= 13 due to its use in prior work (Brown et al., 2020). There were a total of 40,216,231,078 different 13-grams in this dataset. The 1000 most common range from 11 million occurrences down to 20k. The most frequently occurring 13-grams were character repetitions used for styling such as “-- -- ”, “* * * * ”, “! ! ! ! ”, at 11 million, 5.8 million and 1.1 million respectively. Other characters used in this manner include the following: “ # . > ? ”. In the 264k count range, we see repetitions of badly formatted HTML escape characters “ ; &nbsp ”, “; amp ”. Boilerplate from standard forum software appears
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
104
range, we see repetitions of badly formatted HTML escape characters “ ; &nbsp ”, “; amp ”. Boilerplate from standard forum software appears around the 180k occurrences range, such as the following: “ select the forum that you want to visit from the selection below ”. Overall, a large amount of common HTML and CSS is included in the top 1000, along with boilerplate text from Amazon Affiliate Advertising,Component Tokens per byte (LT=LB) Pile-CC 0.2291 PubMed Central 0.3103 Books3 0.2477 OpenWebText2 0.2434 Arxiv 0.3532 Github 0.4412 FreeLaw 0.2622 StackExchange 0.3436 USPTO Backgrounds 0.2116 PubMed Abstracts 0.2183 Gutenberg (PG-19) 0.2677 OpenSubtitles 0.2765 Wikipedia (en) 0.2373 DM Mathematics 0.8137 Ubuntu IRC 0.3651 BookCorpus2 0.2430 EuroParl 0.3879 HackerNews 0.2627 YoutubeSubtitles 0.4349 PhilPapers 0.2688 NIH ExPorter 0.1987
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
105
HackerNews 0.2627 YoutubeSubtitles 0.4349 PhilPapers 0.2688 NIH ExPorter 0.1987 Enron Emails 0.3103 Table 7: Tokens per byte for Pile components TripAdvisor, SimplyHired, Associated Press, PostMedia, The FCC etc. PHP error messages and password login prompts also made an appearance. It may be of interest to fans of Portal that repetitions of “ the cake is a lie . ” achieved a high count. E.2 Benchmark Perplexity Computation To compute the perplexity for a given dataset, we tokenize each document separately, divide the document into segments of up to the maximum sequence length of the model (1024 tokens for GPT-2, 2048 for GPT-3), and predict the logits of the each segment. The inputs to the model are the immediate prior tokens the e.g. for scoring tokens 1 to 1024, we provide tokens 0 to 1023 at the input context. The respective language model implementations handle the causal attention masking. This ensures that every token in the dataset is scored exactly once. This also means that some tokens will have more input context than others. We then aggregate over the whole dataset and compute the final per28
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
106
once. This also means that some tokens will have more input context than others. We then aggregate over the whole dataset and compute the final per28 plexity score. The perplexity for the whole Pile is computed by aggregating over the constituent datasets (i.e. weighted by dataset size, not a simple average of dataset perplexities). Both GPT-2 and GPT-3 share the same tokenizer and vocabulary, making the perplexity scores directly comparable. We use the Hugging Face (Wolf et al., 2020) implementation of GPT-2, and the OpenAI API for GPT-3. The davinci model in the OpenAI API is presumed to correspond to a 175B parameter version of GPT-3. In Table 8 we show the test set perplexities (i.e. not normalized by UTF-8 length, as in Table 2). Because of the costs associated with using the OpenAI API, we compute test perplexities on only one-tenth of the test set in Tables 8 and Table 2. Specifically, we randomly sample one-tenth of the documents of each dataset except for three: Ubuntu IRC, BookCorpus2, and PhilPapers. In Table 9, we show test perplexity computed on the full test set on all
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
107
each dataset except for three: Ubuntu IRC, BookCorpus2, and PhilPapers. In Table 9, we show test perplexity computed on the full test set on all GPT-2 models. 0 2 4 6 8 10 Sequence Position Log2 (tokens)234567Loss (nats)Test loss vs. Log sequence position GPT-2 (small) GPT-2 (medium) GPT-2 (large) GPT-2 (xl) GPT-3 (ada) GPT-3 (babbage) GPT-3 (curie) GPT-3 (davinci) Figure 11: Test loss (log perplexity) over the Pile, bucketed by position in the input sequence based on the model’s maximum sequence length. To smooth out the lines, we bucket 4 positions per plotted datapoint. (e.g. positions 0–3, positions 2044–2047). Later tokens are predicted with more context and thus see lower perplexities. E.3 Pejorative Content Initially we decided on separating pejorative content into 4 groups: sex-related terminology, slurs, neither of these categories, and both of these categories. We adapted a public "naughty words" list
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
108
neither of these categories, and both of these categories. We adapted a public "naughty words" list and broke them into these categories with the intern of looking at the proportion of each category in each dataset. However, this provided many issues. First, any blacklist of words would be hard-pressed Figure 12: Percentage of sentences classified as profane in the Pile. The percentage of the CC component and the weighted mean of the Pile as a whole are shown as horizontal lines to catch all the instances of pejorative content, since purposeful misspellings of words could evade the censor and still have the intended effect. Furthermore, words and their intents are always evolving, therefore any list created would likely be always outdated. Another issue pertains to sorting the words into the categories. Words are highly dependent on their context, so a word would change categories with different contexts. F Data Samples The following consists of two random, noncherrypicked 512-byte samples from each constituent dataset of the Pile, sampled from the validation split. F.1 Pile-CC pot trending topics and the coverage around them. First up, there’s a bit of a visual redesign. Previously, clicking on a trending topic would highlight a story from one publication, and you’d have to scroll down past a live
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
109
a visual redesign. Previously, clicking on a trending topic would highlight a story from one publication, and you’d have to scroll down past a live video section to view related stories. Facebook is replacing that system with a simple carousel, which does a better job of showing you different coverage options. To be clear, the change doesn’t affect how stories are sourced, according to Facebook. It’s still the same algorithm pickin e public safety. He said the bridge saves commuters two or three minutes when trains pass – and those minutes could be vital. “Two to three minutes may not mean much if you’re just driving home from work, but if you’re the one waiting for an ambulance to get to your home, if you’re the one waiting for a fire truck to get to your home, if you’re the one waiting for a police car to get to your home, those two to three minutes could mean the difference between life or death,” Sharp said. “That’s what this pro F.2 PubMed Central 29 d a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
110
professional. Do not disregard or avoid professional medical advice due to content published within Cureus. Introduction ============ Total knee arthroplasty (TKA) is a promising treatment for endstage osteoarthritis (OA) of the knee for alleviating pain and restoring the function of the knee. Some of the cases with bilateral TKA are symptomatic, necessitating revision arthroplasty in both the knees. A bilateral revision TKA can be done ei ent\’s ability to make judgements and decisions about their work experiences and learning that will position them as future critical thinkers, life longer enquirers and learners. Conclusion {#jmrs290-sec-0014} ========== Identification of the core capabilities that our stakeholder community rate highly has proved informative in assisting us to describe a "work ready *plus"* medical imaging graduate for the New Zealand context. The results have provided data to the curriculum development team allowing them F.3 Books3 cept of _forçage_ , ’a forcing of language enacted by the advent of an "other" language that is at once immanent and created’, 44as Badiou puts it: this opens up vistas of a truly syntactic analysis of the poem, in which,
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
111
it: this opens up vistas of a truly syntactic analysis of the poem, in which, again, Badiou would be close to his philosophical other, Deleuze, who, as we just saw, defines style through a-grammaticality and who tries to define what he calls an ’intensive line of syntax’.45 Nevertheless, the insistence on syntax as guarantee involves a _seventh paradox_ , the parad rnment, before the Second World War there were 5,300 communities and two million burakumin. The BLL thinks there must be at least three million burakumin living in Japan today. We visited a hall in Osaka where a taiko drum group, made up exclusively of young burakumin, were about to start their weekly rehearsal. The small gymnasium was filled with taiko drums of all sizes. The smallest was about the size of a snare drum, the largest about the size of a compact car. The Japanese drum group Kodo have made th F.4 OpenWebText2 prime minister to repatriate all the police sent to Catalonia before the referendum. What were the results? With nearly all votes counted, the pro-independence parties Together for Catalonia (JxCat), Republican Left of Catalonia (ERC) and
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
112
referendum. What were the results? With nearly all votes counted, the pro-independence parties Together for Catalonia (JxCat), Republican Left of Catalonia (ERC) and Popular Unity (CUP) were on course to win a total of 70 seats in total, giving them a majority in the new parliament. Citizens (Cs) had 25.3% of the vote, winning 37 seats in the 135-seat chamber. Its leader told the BBC her party had been "victorious". Ms Inés Arrima so accommodate Stablecoins. While some analysts opined that Stablecoins are created to bring growth into the crypto space, they are becoming a solid way to reduce crypto volatility due to the fact that their value are pegged to fiat currency. For Low Cost and Almost Instant Across Border Remittance When Stellar-based Wirex stablecoins finally launches, they are going to be used to perfect low-cost and all most immediately cross-border remittance, just like the IBM Stablecoin which has received support fr F.5 ArXiv\mbox{ if } 2\leq x\leq 3 \end{array}\right.$$ is lower semicontinuous and the nonemptiness
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
113
\end{array}\right.$$ is lower semicontinuous and the nonemptiness of ${\operatorname{fix}}\Phi$ is guaranteed by Corollary \[cor:fixed point\]. Notice that ${\operatorname{fix}}\Phi=[1,2]$. Nevertheless the Kakutani fixed point Theorem does not apply since ${\operatorname{gph}}\Phi$ is not closed. On the converse, the set-valued map $\Phi:[0,3]\rightrightarrows [0,3]$ $$\Phi(x):=\left\{\begin{array}{ll} \{1\} & \mbox{ if } 0\leq x<1\\ {}[1,2] & \mbox{ if } 1\leq x\leq 2\\ \{2\} & .eps){width="6.5cm"} ![Gamma-ray spectrum at Mt. Norikura (2.77 km a.s.l). The vertical axis is Flux$\times Eˆ2$. Our data is at $<$ 100 GeV . Data above 300 GeV is
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
114
is Flux$\times Eˆ2$. Our data is at $<$ 100 GeV . Data above 300 GeV is from emulsion chamber experiments. For the latter, see Sec.\[discuss\] []{data-label="norispec"}](norikura.eps){width="7.5cm"} ![The altitude variation of the flux integrated over 6 GeV . The dpmjet3.03 and fritiof7.02 give almost the same feature consistent with the observation while the deviation of fritiof1.6 from the data is obvious. \[trans F.6 Github "enabled", out.enabled); } std::string SMTPServerInfoJSONStringSerializer::serialize(const SMTPServerInfo &in, const SecurityContext &sc) { return SMTPServerInfoJSONSerializer::serialize(in, sc).dump(4); } void SMTPServerInfoJSONStringSerializer::unserialize(SMTPServerInfo &out, const std::string &in, const SecurityContext &sc) { retur lose"> <at-form state="vm.form" autocomplete="off" id="external_test_form">
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
115
const SecurityContext &sc) { retur lose"> <at-form state="vm.form" autocomplete="off" id="external_test_form"> <at-input-group col="12" tab="20" state="vm.form.inputs" formid="external_test"></at-input-group> <at-action-group col="12" pos="right"> <at-action-button variant="tertiary" ng-click="vm.onClose()" > {{::vm.strings.get(’CLOSE’)}} </at-action-button> <at-action-button variant="primary" n F.7 FreeLaw ssible, and further, that the weight of the evidence, the credibility of the witnesses and the persuasive effect of the testimony is for the sole determination of the trier of fact. This Court thus uses the same interpretation of V .R.C.P. 52(a) as it did *487 under the previous statutory requirement found in 12 V .S.A. § 2385. In essense, the defendants urge that this Court should reconsider the case of Green Mountain Marble Co. v. Highway Board, supra, and follow the Federal practice of looking to the evide ng to the fact that
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
116
of Green Mountain Marble Co. v. Highway Board, supra, and follow the Federal practice of looking to the evide ng to the fact that the relevant Arkansas statutes and rules provide for criminal sanctions against school officials who fail to enforce the immunization requirements, the Morningstar and Lake Hamilton School Districts characterized themselves as disinterested bystanders caught in the crossfire between the Schoolchildren and the Officials. See Ark. Code Ann. § 6-18-702(c)(2)(B) (2000) (“Any school official, parent, or guardian violating the regulations shall be subject to the penalties imposed herein.”); Id 30 F.8 Stack Exchange ooks like a fancy wheel, When resetting rotation from 360deg to 0 deg, It animating the wheel in anti-clockwise direction, How to Avoid this??? HTML <ul class="cm"> <li><span>01</span></li> <li><span>02</span></li> <li><span>03</span></li> <li><span>04</span></li> <li><span>05</span></li> <li><span>06</span></li> <li><span>07</span></li>
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
117
<li><span>05</span></li> <li><span>06</span></li> <li><span>07</span></li> <li><span>08</span></li> </ul> SCSS $Brdr: #7d868c; *{ -webkit-box-sizing: border-box; -moz-box-sizing: border-box; box-sizing: border-box; w can I solve it? Yesterday I added Google ReCAPTCHA v3 in one of my client’s Shopify website, but I don’t think that it is working because he is still reporting to receive several spam e-mails. I’ve followed Google’s guide, but I don’t know what to do for "Verifying user response" part of the guide. I’m not an expert in coding. Basically I’ve added this code to the theme.liquid file <script src="https://www.google.com/recaptcha/api.js?render=*site key provided by google*"></script> And then I’ve added th F.9 Wikipedia (en) and the third son of John Bland and Elizabeth née Birch, daughter of
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
118
provided by google*"></script> And then I’ve added th F.9 Wikipedia (en) and the third son of John Bland and Elizabeth née Birch, daughter of Robert Birch, Bland was educated at Trinity College, Cambridge, where he graduated as a Bachelor of Arts in 1825, and a Master of Arts in 1829. He was called to the Irish Bar in 1829, becoming a member of the Queen’s Counsel in 1854. In 1840, he married Charlotte Elizabeth Grove Annesley, daughter of Arthur Grove Annesley and Elizabeth née Mahon, and they had at least one child: John Loftus Bland (1841–1908). After Charlotte’s death in 18 heart of the University campus, a meeting-place for all academic disciplines, improving its opportunities to co-operate across traditional academic boundaries. It also gives USBE-students an opportunity to take an active part of student environment created for the 37 000 students at Umeå University. Organization Umeå School of Business, Economics and Statistics has three departments: the Department of Business Administration, the Department of Economics and the Department of Statistics. USBE Career Cent F.10 USPTO Backgrounds
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
119
departments: the Department of Business Administration, the Department of Economics and the Department of Statistics. USBE Career Cent F.10 USPTO Backgrounds nductivity types), it is necessary that at least some process is steps differentiate between p-type and n-type transistors. Separate implant steps, for example, are needed to define n-well and p-well structures and to dope the source/drain regions of n-channel and p-channel transistors. Whenever possible, however, it is generally desirable to use a single process step to define transistor features regardless of the transistor type. Single process steps imply a single mask step, which is always desirable to enser further comprising a means of identifying the user by voice recognition. Also, an object is dispenser further comprising a means of identifying a supervisor by voice recognition. Furthermore, an object is a dispenser further comprising a means of customizing the plurality of aural messages for instructing the user during each of the plurality of washing steps. An object of the invention is a dispenser for metering a liquid cleanser to a user and prompting the user in compliance with a recommended washF.11 PubMed Abstracts ent (REM) latency were found to be significantly worse in Group 1 as
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
120
a user and prompting the user in compliance with a recommended washF.11 PubMed Abstracts ent (REM) latency were found to be significantly worse in Group 1 as compared with Group 2. Cognitive and executive parameters were significantly impaired in Group 1. Shorter total sleep time, poorer sleep efficiency, and prolonged sleep latencies were observed to be associated with poor memory and executive function in patients with refractory epilepsy. Our study strongly suggests that sleep disturbances, mainly shorter total sleep time, poor sleep efficiency, and prolonged sleep latencies, are associated neurons in vesicular GABA transporter (VGAT)-venus transgenic mouse. Inhibitory neurons play important roles in a number of brain functions. They are composed of GABAergic neurons and glycinergic neurons, and vesicular GABA transporter (VGAT) is specifically expressed in these neurons. Since the inhibitory neurons are scattered around in the CNS, it is difficult to identify these cells in living brain preparations. The glutamate decarboxylase (GAD) 67-GFP knock-in mouse has been widely used for the identif F.12 Gutenberg (PG-19) s he met with as a novelist, he was anxious to prosecute his
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
121
used for the identif F.12 Gutenberg (PG-19) s he met with as a novelist, he was anxious to prosecute his original profession of medicine; and having procured from a foreign university the degree of M.D., he commenced to practise physic in Chelsea, but without success. He wrote, however, an essay "On the External Use of Water," in which he seems to have partly anticipated the method of the cold-water cure. In 1753 he published his "Adventures of Count Fathom;" and, two years later, encouraged by a liberal subscription, he issued a translation of "Don Yearn; Its Advertising brought us such Renown, We jumped Three Hundred Thousand, on that Turn!" XXXVI I think the man exaggerated some His increased Circulation,–but, I vum! If I could get Two Thousand for one Tale, I’d write him Something that would simply Hum! XXXVII For I remember, shopping by the way, I saw a Novel writ by Bertha Clay; And there was scrawled across its Title-Page, "This is the Stuff that Sells–so People say!" XXXVIII Listen–a moment listen!–Of the same Wood-pulp on wh F.13 OpenSubtitles
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
122
XXXVIII Listen–a moment listen!–Of the same Wood-pulp on wh F.13 OpenSubtitles ad for you." " Too bad for me?" "How about too bad for you?" "Oh no!" "Luckily I keep a spare." "Look everyone!" "My winky was a key!" "Oh dear, bloody Dutchman." "Foxxy, I’m coming!" "Don’t do anything stupid or the shooting begins." "Austin, take Ducky I’ll stay here and be your backup." "Ducky, what do we do?" "I’m not really a "hands-onevil-genius"." "Think you were always the smart one." "I could re-write the output capacity to the tractorbeam from one of the conduit boxes up there." "Come on, let’s ." "this calls for... four people." "Yes!" "we got it." "guys." "We got it." " Got what?" " Our sub." " Did he say sub?" " Mm-hmm." "Only private sub on the Florida coast rated for 300 fathoms." "Sub as in submarine?" "Following up on your haloclines." "so we’re going to have to drive all
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
123
sub on the Florida coast rated for 300 fathoms." "Sub as in submarine?" "Following up on your haloclines." "so we’re going to have to drive all night... if we’re going to be there by morning." "Anybody have trouble sleeping in a car?" "whoa." "Wait a minute." "What happened to the nice offices in Canaveral City?" "Mr. Benirall expects you to take ’em." "we just go F.14 DM Mathematics 31 3651*w**2 + 519*w + 1 Find the second derivative of -91419126*m**2 - 162128943*m. -182838252 Find the third derivative of 5*l*u*y**3 + l*u*y - 5*l*y**2 4621073*u*y**3 - 1755838*u*y**2 + u wrt y. 30*l*u - 27726438*u Find the third derivative of 317297018*s**3 + 3136*s**2 - 30884*s wrt s. 1903782108 What is the third derivative of -16525*f*r**3 + 20*f*r + 356*r**3 +
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
124
s. 1903782108 What is the third derivative of -16525*f*r**3 + 20*f*r + 356*r**3 + 1425730*r**2 wrt r? -99150*f + 2136 What is the second derivative of 199836725*j**2 - 443399*j - 462 wrt j? 399673450 What is the derivative of the nearest integer? 5 What is 783451 to the power of 1/3, to the nearest integer? 92 What is the fourth root of 6322907 to the nearest integer? 50 What is the ninth root of 4723626 to the nearest integer? 6 What is 4954939 to the power of 1/2, to the nearest integer? 2226 What is 625583 to the power of 1/3, to the nearest integer? 86 What is 1105849 to the power of 1/3, to the nearest integer? 103 What is the fourth root of 4820344 to the nearest integer? 47 What is the seventh root of 243476 to the neare F.15 HackerNews ced lists I email don’t get formatted correctly. It’s slightly annoying for such an otherwise beautifully designed layout. —— ajcronk
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
125
ced lists I email don’t get formatted correctly. It’s slightly annoying for such an otherwise beautifully designed layout. —— ajcronk There is a typo in the url at the end of the How did your day go? email. Should be ohlife.com/today, not ohlife.come/today ~~~ sgupta Thanks for the heads up! —— a3_nm How exactly is this service better than, say, a simple text file on my own machine with a daily reminder set up through some other means? Why would I want to use some third-party website for somethi or Amazon EC2 and Amazon SQS. The bandwidth tier in which you will be charged each month will be calculated based on your use of each of these services separately, and could therefore vary across services." —— yaacovtp Can anyone tell me what bandwidth costs a month once you need over a terabyte a month? How would you host a 5-10 mb movie that may be viewed millions of times without using a 3rd party video host like youtube etc? ~~~ especkman Lots of dedicated hosts will include a 2-5 TB of transfer a F.16 BookCorpus2
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
126
~~~ especkman Lots of dedicated hosts will include a 2-5 TB of transfer a F.16 BookCorpus2 considerate of me, you’re right. I apologize." Kate smiled in what she hoped was a winning way. She teetered over to the counter on heels that were too high and put down her things with a sigh of relief. Althea, who would not reveal her age but was probably somewhere in her late sixties, patted her dark-dyed helmet of hair and straightened the flowing turquoise silk jacket she was wearing over white capris and a white tank. "Well. It just seems to me that as the _owner_ , you should try to set some sort ofe notebook didn’t have lines for me to write with like some notebooks have. I hated that notebook and I hated writing into it. I was glad to throw that damn thing out even if it was unfinished. Ugh. I was urged by voice "You to go take your pills and eat food." but I refused on calling mom. I should have listened to the voice’s suggestion because mom picked up the phone at eight forty five in the morning and hogged me to
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
127
I should have listened to the voice’s suggestion because mom picked up the phone at eight forty five in the morning and hogged me to ten o’clock is when she finally quit. Ugh hence voice picking onto me when I got off F.17 EuroParl raˇcun prometne varnosti in visokih cen? Danes želimo izvedeti, kako in kdaj bomo integrirali razvrš ˇcanje zgornjega zra ˇcnega prostora ter kako bomo skupaj upravljali spodnji zra ˇcni prostor v prihodnosti. Ali se lahko odkrito dolo ˇcijo ovire za vzpostavitev funkcionalnih blokov nad evropskim ozemljem? Ali je mogo ˇce osvetliti politi ˇcno voljo držav ˇclanic, da
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
128
izpolnijo svoje obveznosti? Prav tako nas skrbi, da pristop od spodaj navzgor ne bo uspel, ker v treh letih države ˇclanice niso razvile funkcionalnih blok om ekonomisk styrning som vi debatterar inom kort kommer att vara mycket viktigt. Vi vet mycket väl att det är på gång i vårt lagstiftningsförfarande, och vi hoppas att vi kommer att vara klara så snabbt som möjligt. Vad kan jag sammanfattningsvis säga? Hela paketet som vi undertecknar i dag kommer att börja gälla i Europeiska unionen från och med den 1 januari 2011, alltså mycket snart. Det är viktigt för oss alla, såväl för marknaderna som för våra medborgare, att förstå att avsikten med paketet är att hj F.18 YoutubeSubtitles
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
129
är att hj F.18 YoutubeSubtitles science term for a mixture of things that don’t usually mix. The things in this case are water and fats. Under normal circumstances, fats and water repel each other, but milk also contains complex protein chains called caseins that are made up of both hydrophilic, or water loving, and lipophilic, or fat loving, particles. When presented with both water and fats, caseins grab bits of fat and cluster up into globules called micelles, with the fat on the inside and the hydrophilic bits on the outside. The hydr SE WE KIND OF ARE MORE, HIPPY , I GUESS MAYBE IN SOME OF THE THINGS THAT WE DO. AND THEY WERE JOKING AND THEY WERE LIKE, "OH WE HEARD ABOUT THIS TOWN IT’S LIKE THIS SUSTAINABLE CITY THERE’S SOLAR PANELS, YOU GUYS WOULD LOVE IT." AND I LOOKED IT UP AND I WAS LIKE I REALLY ACTUALLY DO LOVE THIS TOWN. >> Sreenivasan: JOSHUA, A PHYSICAL THERAPIST, GOT A JOB AT THE LOCAL HEALTH CENTER IN THE
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
130
THIS TOWN. >> Sreenivasan: JOSHUA, A PHYSICAL THERAPIST, GOT A JOB AT THE LOCAL HEALTH CENTER IN THE TOWN’S COMMERCIAL HUB BEFORE THEY MOVED IN. IT’S WHERE THE FIRST BUILDINGS WENT UP. THERE’S ALSO A RESTAURANT AND COFFEE SH F.19 Ubuntu IRC emingly wlan related) <Snappy:New> <linux-raspi2 (Ubuntu):New for ppisati> <https://launchpad.net/bugs/1627643> <ppisati> ogra_: or we punch a hole in the dev image so we can login via the serial console and check what’s really going on <ppisati> ogra_: yes <ogra_> well, i wanted to play with systemd console but didnt have time 32 for that yet <ogra_> \o/ <ogra_> something at least ... that kernel looks fine <ppisati> ogra_: good to know <ogra_> do you have an SRU bug that i can tag verification-done ? <ogra_
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
131
<ogra_> do you have an SRU bug that i can tag verification-done ? <ogra_ problem with this? Like, if teenage boy wants to have nekkid lady wallpapers, maybe he don’t want it to come up on family computer... Dunno, maybe it’s not an issue?" <swilson> hi there! yes, i believe that a bug has been created which raises this same issue - about embarrassing or confidentiality issues with this <swilson> this seems to be a bit of an edge case, but it may be significant enough to warrant giving some careful thought <imnichol> Or if you have bank info on your screen before it’s locked #ub F.20 PhilPapers intersubjectivity and self-consciousness was already emphasized by Sartre. Forthcoming in Grazer Philosophische Studien 84 (2012), p. 75101 15 Thus, to use Rochat’s terminology, from this point onwards, the child has "others in mind" (Rochat 2009). The child now begins to understand that she is a subject that can be observed by others, just like she can observe the behavior of others, and she can begin to consider others’
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
132
can observe the behavior of others, and she can begin to consider others’ perspectives on herself. It is at this point that the child begins to fully apprecia d an entire chapter detailing the remarkable achievements of Ashkenazi Jews and hold them up as exhibit A in the argument that human evolution has been, in Wade’s words, recent, copious, and regional. The example of Ashkenazi evolution is supposed to show the absurdity of the view, held by authors like Jared Diamond and Stephen Jay Gould, that human evolution either stopped one hundred thousand years ago or that natural selection has somehow continued to sculpt the bodies but not the brains of different gro F.21 NIH ExPorter rapies that can inhibit the EMT, but few assays for EMT inhibitors in high throughput screens (HTS) have developed. A change in fibroblast growth factor receptor 2 (FGFR2) splicing occurs during the EMT and using an innovative luciferase-based splicing reporter assay we previously carried out a genome-wide high throughput cDNA expression screen for regulators of this splicing switch. This screen identified the epithelial cell type specific splicing regulators ESRP1 and ESRP2 demonstrating the feasibility of
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
133
type specific splicing regulators ESRP1 and ESRP2 demonstrating the feasibility of l and behavioral research projects utilizing primates residing in a seminatural habitat. This population has the most extensive computerized demographic and genetics database available to researchers anywhere in the world. The population management program for CS has been designed to optimize the health and well-being of the monkeys, to enhance the value of the colony for research. In addition, the goal is to provide healthy animals to the scientific community for biomedical research, including AIDS and Sl F.22 Enron Emails want to make sure that my vacation time gets paid at 100% before I go down to the 90% level. Thanks for taking care of this. As you can see, I now have access to my e-mail so when I’m not pumping, feeding, changing diapers, etc... I acn be checking up on things!!! Carol St. Clair EB 3892 713-853-3989 (Phone) 713-646-3393 (Fax) [email protected] Suzanne Adams 07/18/00 05:22 PM To: Carol St Clair/HOU/ECT@ECT cc: Taffy Milligan/HOU/ECT@ECT
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
134
Suzanne Adams 07/18/00 05:22 PM To: Carol St Clair/HOU/ECT@ECT cc: Taffy Milligan/HOU/ECT@ECT Subject: Re: Carol St. ClairCarol, I —-Original Message—– From: "Prakash Narayanan" <[email protected]>@ENRON Sent: Sunday, December 02, 2001 9:28 PM To: Kaminski, Vince J Cc: Crenshaw, Shirley Subject: Talk on Friday Dear Vince How are you? I understand that things are extremely hectic for you right now but I was wondering if we are going ahead as schedulef on friday. It would be great to hear from you. Best Regards Prakash Prakash Narayanan 412-422-3287 (Home) 412-607-5321 (Mobile) 6315 Forbes Avenue Apartment # 809 Pittsburg male femaleStackExchange OpenWebText2 Pile-CC FreeLaw Books3 Wikipedia (en) HackerNews OpenSubtitles PubMed Central PubMed Abstracts Enron Emails BookCorpus2 USPTO Backgrounds ArXiv YoutubeSubtitles Github PhilPapers
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
135
PubMed Central PubMed Abstracts Enron Emails BookCorpus2 USPTO Backgrounds ArXiv YoutubeSubtitles Github PhilPapers Gutenberg (PG-19) EuroParl Ubuntu IRC NIH ExPorter DM Mathematics 0.10 0.05 0.000.050.10 Figure 13: The average sentiment co-occurrence with each gender across all datasets. 33 34 Muslim Christian Atheist Buddhist Hindu JewStackExchange OpenWebText2 Pile-CC FreeLaw Books3 Wikipedia (en) HackerNews OpenSubtitles PubMed Central PubMed Abstracts Enron Emails BookCorpus2 USPTO Backgrounds ArXiv YoutubeSubtitles Github PhilPapers Gutenberg (PG-19) EuroParl Ubuntu IRC NIH ExPorter DM Mathematics 1.00 0.75 0.50 0.25 0.000.250.500.751.00 Figure 14: The average sentiment co-occurrence with each religious word across all datasets. Each dataset’s sentiments have been normalized by the maximum norm sentiment for that dataset. 35 Component GPT-2 GPT-3
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
136
each religious word across all datasets. Each dataset’s sentiments have been normalized by the maximum norm sentiment for that dataset. 35 Component GPT-2 GPT-3 small medium large xl ada babbage curie davinci Pile-CC 26.8894 20.5671 18.1656 16.9572 16.2430 13.0270 10.7532 8.4929 PubMed Central 11.0626 8.9052 8.0454 7.5404 6.8800 5.7006 4.9390 4.3143 Books3 28.3889 22.0958 19.3424 17.7833 15.4209 12.4220 10.1526 7.1927 OpenWebText2 23.6764 17.6175 15.1314 13.6267 12.0063 9.5439 7.7706 5.9163 ArXiv 14.2804 11.1896 10.0904 9.3330 7.5551 6.1541 5.2537 4.5341 Github 16.6814 7.9322 16.6742 13.3337 3.9614 3.1660 2.7398 2.4240
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
137
Github 16.6814 7.9322 16.6742 13.3337 3.9614 3.1660 2.7398 2.4240 FreeLaw 16.1000 11.7518 10.8427 10.0965 8.7976 7.0366 5.8256 4.8926 Stack Exchange 13.7202 9.3405 8.8467 8.3238 7.6652 5.9486 5.0267 4.3796 USPTO Backgrounds 15.1141 11.9232 10.5878 9.8095 9.2775 7.7000 6.5849 5.6411 PubMed Abstracts 20.5642 15.2379 13.1190 11.9355 13.2112 10.4188 8.5861 7.1604 Gutenberg (PG-19) 26.4947 17.8975 16.4722 16.5112 12.5709 9.6349 7.7940 6.3112 OpenSubtitles 22.7418 18.5724 17.0868 16.2709 16.2174 13.8561 11.8836 9.8578 Wikipedia (en) 27.0237 19.7570 17.4856 16.7849 12.9112 9.9453 7.8363 5.6915
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
138
Wikipedia (en) 27.0237 19.7570 17.4856 16.7849 12.9112 9.9453 7.8363 5.6915 DM Mathematics 9.8990 8.7389 8.2928 7.9772 7.2458 6.5231 6.0171 5.6020 Ubuntu IRC 33.3028 26.1203 22.6128 20.9461 12.1138 9.6995 8.0628 6.5679 BookCorpus2 25.0743 19.9725 17.6343 16.2905 16.1530 13.1796 11.0885 9.2205 EuroParl 62.8981 36.9757 28.6198 23.4294 6.4996 5.3282 4.4982 3.8327 HackerNews 45.0915 29.2599 32.0796 33.9774 22.1295 17.6314 14.6582 12.1283 YoutubeSubtitles 25.7794 18.8173 15.9002 14.3104 8.4740 6.6394 5.4510 4.5235 PhilPapers 30.1129 23.0288 20.3755 18.5649 14.4730 11.6785 9.6797 7.9915
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
139
PhilPapers 30.1129 23.0288 20.3755 18.5649 14.4730 11.6785 9.6797 7.9915 NIH ExPorter 23.9004 18.2298 15.9850 14.6371 16.1417 12.8744 10.6573 8.8110 Enron Emails 34.7954 23.4353 25.7138 23.9791 16.8190 13.6043 11.6473 9.7655 The Pile 18.0878 13.2253 12.9177 11.8633 9.7355 7.8456 6.5904 5.4508 Table 8: Test perplexity of the Pile using GPT-2 and GPT-3. Evaluation is performed on one-tenth of the test data of the Pile, on a per-document basis. 36 Component GPT-2 small medium large xl Pile-CC 26.5 20.3 17.9 16.7 PubMed Central 10.3 8.3 7.6 7.1 Books3 27.9 21.3 18.5 17.0 OpenWebText2 23.5 17.5 15.1 13.6 ArXiv 13.7 10.7 9.7 9.0
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
140
OpenWebText2 23.5 17.5 15.1 13.6 ArXiv 13.7 10.7 9.7 9.0 Github 16.5 8.1 16.7 13.5 FreeLaw 15.9 11.6 10.8 10.1 Stack Exchange 13.7 9.4 9.0 8.4 USPTO Backgrounds 16.6 13.0 11.5 10.6 PubMed Abstracts 20.9 15.4 13.3 12.1 Gutenberg (PG-19) 37.8 24.9 22.8 24.3 OpenSubtitles 22.1 18.1 16.6 15.8 Wikipedia (en) 27.0 19.8 17.5 16.8 DM Mathematics 9.9 8.7 8.3 7.9 Ubuntu IRC 33.3 26.1 22.6 20.9 BookCorpus2 25.1 20.0 17.6 16.3 EuroParl 63.9 41.9 33.5 27.7 HackerNews 43.7 28.3 30.9 32.4 YoutubeSubtitles 25.3 18.8 16.2 14.8 PhilPapers 30.1 23.0 20.4 18.6 NIH ExPorter 23.2 17.7 15.5 14.2
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
141
PhilPapers 30.1 23.0 20.4 18.6 NIH ExPorter 23.2 17.7 15.5 14.2 Enron Emails 22.0 15.4 18.6 18.1 the Pile 18.4 13.3 13.1 12.0 Table 9: Full Test Perplexity of the Pile using GPT-2. Male Female general little military married united sexual political happy federal young great soft national hot guilty tiny criminal older former black republican emotional american worried major nice such live offensive lesbian Table 10: Top 15 most biased adjectives/adverbs for each gender Muslim Christian Atheist Buddhist Hindu Jew islamic adrian religious static indian little international available agnostic final single white new great such private free natal american high liberal interested asian common black bible likely central more false western good much chinese united poor best old less japanese real demonic radical same least noble other german regional harmonious political complete british romantic entire third moral full cultural unlicensed national special scientific fundamental social stupid own hispanic rational udisplaycontext lower nuclear syrian biblical skeptic familiar local african bad original skeptical beneficial general hard
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
142
own hispanic rational udisplaycontext lower nuclear syrian biblical skeptic familiar local african bad original skeptical beneficial general hard guilty happy intellectual native most criminal Table 11: Top 15 most biased adjectives/adverbs for each religion 37 White Black Asian Hispanic indian unarmed international likely rich civil western african aboriginal scary chinese american great federal japanese mexican old diary best united superior political european cervical good amish foreign spanish little nigerian eastern potential same concerned secondary better red urban dietary medical stupid historical open more live literary grand new equal criminal vietnamese educational eternal worst russian young Table 12: Top 15 most biased adjectives/adverbs for each demographic White Black Asian Hispanic -0.114 -0.148 -0.028 -0.024 Table 13: Average sentiment co-occurence of each demographic Component Topic #1 Topic #2 Topic #3 Topic #4 Topic #5 Topic #6 Topic #7 Topic #8 Pile-CC Generic Politics Generic Technical Leisure Generic Plants Entertainment PubMed Central Cells Cells Cells Cells Cells Cells Cells Cells Books3 Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown OpenWebText2 US Politics Law Sports Education Business Tech US Religion Generic ArXiv Data Math Modeling Math Physics Physics Math Dynamics
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
143
Books3 Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown OpenWebText2 US Politics Law Sports Education Business Tech US Religion Generic ArXiv Data Math Modeling Math Physics Physics Math Dynamics Github Unknown Programming Unknown Java C/C++ Unknown Go Unknown FreeLaw Appeals Appeals Legal Legal Appeals Legal Legal Appeals Stack Exchange Software Unknown Server Programming Applications File System Programming Users USPTO Backgrounds Data Electronics Devices Unknown Data Unknown Chemistry Data PubMed Abstracts Organ TransplantNervous SystemAnimal Study Animal Study Ophthalmology Bacteria Pulmonology Fluids Gutenberg (PG-19) Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknowne OpenSubtitles Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown Wikipedia (en) Education International PoliticsSports Sports Entertainment Entertainment Logistics Science DM Mathematics Calculation Probability Calculation Solving Calculation Calculation Probability Calculation Ubuntu IRC Bugs Pull Requests Bugs Bugs Bugs Bugs Bugs Pull Requests BookCorpus2 Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown EuroParl International PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational Politics HackerNews Generic Software Generic Generic Software Generic Generic Generic YoutubeSubtitles Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknowne PhilPapers Logic Science Science Mind Science Epistemology Logic Science NIH ExPorter Cells Disease Cells Cells Clinical Clinical Unknown Clinical
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
144
PhilPapers Logic Science Science Mind Science Epistemology Logic Science NIH ExPorter Cells Disease Cells Cells Clinical Clinical Unknown Clinical Enron Emails Email Email Email Email Email Email Email Business Table 14: Topic Summaries Component Topic #9 Topic #10 Topic #11 Topic #12 Topic #13 Topic #14 Topic $15 Topic #16 Pile-CC Education Politics Home Business Geography Sports Medicine Generic PubMed Central Cells Cells Cells Cells Cells Cells Cells Cells Books3 Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown OpenWebText2 Drugs Sports Geogrpahy Crime Military Unknown Research Sports ArXiv Dynamics Math Physics Physics Physics Physics Math Modeling Github Unknown HTML/CSS HTML/CSS C/C++ Java C/C++ Unknown HTML/CSS FreeLaw Legal Legal Legal Legal Legal Legal Legal Appeals Stack Exchange Programming HTML/CSS Programming Programming HTML/CSS Java SQL Java USPTO Backgrounds Imaging Electronics Unknown Unknown Data Imaging Imaging Chemistry PubMed Abstracts Human DiseaseResearch Human DiseaseClinical Clinical Medical ImagingCells Cells Gutenberg (PG-19) Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown OpenSubtitles Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown Wikipedia (en) Sports Geography Entertainment Unknown Geography Sports History Law DM Mathematics Calculation Differentiation Differentiation Solving Simplification Calculation Units Unknown
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
145
Wikipedia (en) Sports Geography Entertainment Unknown Geography Sports History Law DM Mathematics Calculation Differentiation Differentiation Solving Simplification Calculation Units Unknown Ubuntu IRC Software Software Software Bugs Software Pull Requests Software Bugs BookCorpus2 Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown EuroParl International PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational PoliticsInternational Politics HackerNews Generic Generic Generic Software Software Generic Generic Generic YoutubeSubtitles Unknown Unknown Unknown Unknown Unknown Unknown Unknown Unknown PhilPapers Epistemology Science Logic Science Epistemology Science Logic Logic NIH ExPorter Cells Cells Disease Disease Disease Disease Disease Clinical Enron Emails Energy Email Email Email Email Email Computer Computer Table 15: Topic Summaries (continued) Component Topic #1 Topic #2 Topic #3 Topic #4 Topic #5 Topic #6 Topic #7 Topic #8 Pile-CC like time good use wantpeople said government war rightsaid like time got goingsystem surface x high secondnew dating art day citylike time game good foodwater plants food climate plantmusic like book new film PubMed Central cells data study cell resultscells cell data
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
146
plants food climate plantmusic like book new film PubMed Central cells data study cell resultscells cell data figure etstudy cells c data groupstudy data patients cells analysisdata study patients cells cpatients cells study cell analysispatients study data cells analysiscells data cell analysis patients Books3 time said like new knowlike said time man waylike said time new rightsaid like time know newsaid like time new waysaid like time new goodsaid like way new timesaid new like time man OpenWebText2 said trump president house statesaid court law case stateteam season game said playerspeople like said school lifesaid government year market businessdata use google system newpeople law american god worldlike world time people day ArXiv case given time let datalet function case given orderfunction case let state modellet set following
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
147
day ArXiv case given time let datalet function case given orderfunction case let state modellet set following number xphys data energy field blet model field system energylet given case number theoremtime case let set given Github y d b abbr jreturn function div var valuevoid f license v countriesreturn public int case nullconst typename return void templatefa var span file keyerr return nil func errorimport msgstr msgid insert license FreeLaw court trial evidence case statecourt state case trial evidencecourt defendant state states trialcourt district plaintiff defendant motioncourt trial evidence defendant unitedcourt defendant trial evidence statescourt defendant case law motioncourt state evidence defendant district Stack Exchange run q server project usedata option pdf q railsfunction server use thread clientarray int value like codedevice
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
148
q server project usedata option pdf q railsfunction server use thread clientarray int value like codedevice spring android app bootfile image files echo pathfile line error import pythonq like use user set USPTO Backgrounds signal system invention memory lineinvention data power voltage frequencyinvention surface having present liquidpressure system invention cells usedata system memory information deviceshigh air light invention temperatureal et invention present waterinvention circuit data present signal PubMed Abstracts liver group acute transplantation renalactivity nerve stimulation induced musclespecies study studies associated riskdose mg rats effects effectretinal eye lens corneal laserstrains isolates resistance resistant bacteriap levels patients blood increasedactivity acid high water concentration Gutenberg (PG-19) said time little man greatsaid time little like mansaid man
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
149
acid high water concentration Gutenberg (PG-19) said time little man greatsaid time little like mansaid man time great mensaid time great man littlesaid great like man littlesaid man time like daysaid man time little likeman said like time little OpenSubtitles know right come got likelike know come right wantknow oh right yeah likeknow like oh got rightknow right like oh wantknow like right got comeknow right yeah got letknow got right like oh Wikipedia (en) category university school american collegepeople government category political chinesecategory players people football bornchampionship category driver cars carfilm category films new televisioncategory music album song releasedcategory railway station line newsystem category energy work systems DM Mathematics let pm minutes factor dividedletters let replacement sequence probcollect terms positive assuming
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
150
category energy work systems DM Mathematics let pm minutes factor dividedletters let replacement sequence probcollect terms positive assuming simplifylet suppose solve nearest clet common calculate suppose highestlet solve suppose base calculatefactors prime replacement letters listsolve remainder divided calculate true Ubuntu IRC ubuntu like think bug needubuntu bug like think snaplike ubuntu know created okubuntu like think need yeahubuntu like think yeah knowlike ubuntu think good yeahubuntu like bug use knowubuntu good snap use like BookCorpus2 said like time know eyessaid like know way eyessaid know like time rightsaid like time know goingsaid like time know goingsaid like know time headsaid like time know goingsaid know time like going EuroParl european mr commission president europeeuropean president commission mr unionmr
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
151
know time like going EuroParl european mr commission president europeeuropean president commission mr unionmr european president commission parliamenteuropean mr president commission energyeuropean mr commission president parliamentcommission president european mr parliamenteuropean commission mr president councileuropean commission mr union parliament HackerNews like people work time uselike people work time uselike people work time thinkpeople time like way workpeople like time think uselike people work time thinklike people time think workpeople like work good use YoutubeSubtitles like know going think rightknow like going people timelike going right time knowlike going guest people thinklike think people know goinglike know going look hostlike know think right goinglike people know going time PhilPapers theory case  reduction paradoxphilosophy case moore physical
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
152
think right goinglike people know going time PhilPapers theory case  reduction paradoxphilosophy case moore physical theorycase science theory set worldself case science theory analysistheory science physical de casecase theory science s epistemicderivation reduction  paradox case de set science theory NIH ExPorter cells cell studies research studycell studies research cells studycell research gene development studyresearch determine health specific cellscells specific cell study aimstudies development patients analysis clinicalresearch study development specific usestudy research clinical studies development Enron Emails subject pm new time energypm subject enron cc knowpm new enron time imageenron e mail new subjectsubject enron said sent imagesubject pm enron database senthou pm subject e timefinal enron schedule energy
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
153
said sent imagesubject pm enron database senthou pm subject e timefinal enron schedule energy information Table 16: Topic Terms Component Topic #9 Topic #10 Topic #11 Topic #12 Topic #13 Topic #14 Topic $15 Topic #16 Pile-CC students school work university researchsaid state government law courthome room house hotel areause business data information newsaid new city years policegame team season year saidhealth medical care treatment bodypeople like know think life PubMed Central cells expression patients cell studydata cells patients study ccells study cell data figurestudy data p fig cellsstudy p cells c patientscells study data time grouppatients study cells et datastudy cells patients cancer figure Books3 said like time man goodsaid like time know waysaid like time way newsaid like time new knowsaid like time people mansaid
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
154
goodsaid like time know waysaid like time way newsaid like time new knowsaid like time people mansaid like new people mansaid time like man newsaid time new people like OpenWebText2 drug cannabis drugs marijuana womenlike time new game waycity new unlockable building saidsaid police people man oldgame war party military saidflight caption aircraft add waterstudy research time climate foundv granada club m cent ArXiv time function r model allet number model system theoremlet set space model givenlet theorem given case xmodel order energy let physlet phys order model casex let time field setlet model data set function Github string license def public importx z divide var yend values list color tabledefine software copyright include endifvoid value public return classint struct return case
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
155
values list color tabledefine software copyright include endifvoid value public return classint struct return case staticreturn self size long stringvar assert text label check FreeLaw court defendant state trial plaintiffcourt â law case statecourt plaintiff state case evidencecourt defendant states plaintiff casecourt district states case unitedcourt plaintiff state case districtcourt trial state defendant casecourt states district united trial Stack Exchange x y q d ctext color width font heightcode n use int includeb q class n kdiv page function var formstring return public new classtable select question like qandroid new public import void USPTO Backgrounds image light data optical systemdevice invention layer film powerinvention material light method highinvention present device object providedata network system user informationoptical surface device invention systemimage light device
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
156
light method highinvention present device object providedata network system user informationoptical surface device invention systemimage light device sheet displayinvention layer substituted et group PubMed Abstracts women patients positive hiv cancerhealth data care based studybone asthma vaccine study spermpatients treatment group clinical patientpatients disease study cases agemethod artery surface energy opticalcells cell expression gene proteinbinding protein receptor dna beta Gutenberg (PG-19) said little time man oldsaid man little like timesaid man little time greatsaid great little man liketr said man time littlesaid man great time liketime said men man liketr said man time little OpenSubtitles know right got like ohlike know right think ohknow like come right goodknow come got oh rightknow right like yeah comeknow right oh like comeknow right like think
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
157
come right goodknow come got oh rightknow right like yeah comeknow right oh like comeknow right like think comeknow like right want got Wikipedia (en) align season points game rightcategory new state united statescategory game film series videocategory population species age oilcategory county district references villageseason team league player nflcategory century war new decategory new states law american DM Mathematics common let divided calculate factorlet derivative wrt second findderivative wrt find express rearrangelet suppose solve b clet suppose value b simplifybase c common picked bdigit terms collect thousands letlet suppose derivative c determine Ubuntu IRC ubuntu like time think snapubuntu like need use snapubuntu think like yeah uselike ubuntu think use needubuntu like think work timeubuntu like
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
158
like need use snapubuntu think like yeah uselike ubuntu think use needubuntu like think work timeubuntu like think yeah yesubuntu like think good àlike ubuntu need ok juju BookCorpus2 said like know right timesaid like know time goingsaid like know eyes timesaid like time know goingsaid like know going timesaid like eyes time goingsaid like know time goingsaid like know looked going EuroParl mr commission president european iraneuropean mr president commission parliamentmr european commission parliament presidenteuropean mr commission president parliamenteuropean commission mr parliament presidenteuropean mr commission parliament presidenteuropean mr commission parliament presidenteuropean mr parliament commission president HackerNews people like think time uselike people time think usepeople like data time workuse like work time thinklike people time
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
159
like think time uselike people time think usepeople like data time workuse like work time thinklike people time work datapeople time like work goodlike people time use thinklike time people think use YoutubeSubtitles like know people going timelike time know going thinklike host guest know looklike think know people goinglike think right know peoplelike know want going rightlike know going people looklike know people think going PhilPapers theory s case belief experiencetheory science set de philosophy reduction theory paradox derivationtheory case order space newepistemic science belief theory systemtheory case science physics theories derivation reduction t paradoxcase  s derivation theory NIH ExPorter cells research specific studies roleresearch cells cell studies studydisease research study cells cellcells cell study
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2101.00027
160
research specific studies roleresearch cells cell studies studydisease research study cells cellcells cell study disease humancells specific development studies researchresearch cells studies cell projectcell research cells specific studiesresearch studies clinical determine cancer Enron Emails time new enron power subjectsubject pm friday sent octoberhou enron subject cc naimage enron pm hou subjectsubject message pm know cchou enron subject cc gasspace alias disk enron saidhou disk space alias e Table 17: Topic Terms (continued)
2101.00027
The Pile: An 800GB Dataset of Diverse Text for Language Modeling
Recent work has demonstrated that increased training dataset diversity improves general cross-domain knowledge and downstream generalization capability for large-scale language models. With this in mind, we present \textit{the Pile}: an 825 GiB English text corpus targeted at training large-scale language models. The Pile is constructed from 22 diverse high-quality subsets -- both existing and newly constructed -- many of which derive from academic or professional sources. Our evaluation of the untuned performance of GPT-2 and GPT-3 on the Pile shows that these models struggle on many of its components, such as academic writing. Conversely, models trained on the Pile improve significantly over both Raw CC and CC-100 on all components of the Pile, while improving performance on downstream evaluations. Through an in-depth exploratory analysis, we document potentially concerning aspects of the data for prospective users. We make publicly available the code used in its construction.
http://arxiv.org/pdf/2101.00027
[ "Leo Gao", "Stella Biderman", "Sid Black", "Laurence Golding", "Travis Hoppe", "Charles Foster", "Jason Phang", "Horace He", "Anish Thite", "Noa Nabeshima", "Shawn Presser", "Connor Leahy" ]
[ "cs.CL" ]
null
null
cs.CL
20201231
20201231
[ { "id": "2101.00027" }, { "id": "1606.06565" }, { "id": "2005.14050" }, { "id": "2005.14165" }, { "id": "1802.07228" }, { "id": "2010.14571" }, { "id": "2005.13156" }, { "id": "1901.02860" }, { "id": "1803.09010" }, { "id": "2010.14701" }, { "id": "2005.00813" }, { "id": "2001.08361" }, { "id": "2006.16668" }, { "id": "1801.10198" }, { "id": "1907.11692" }, { "id": "2006.16923" }, { "id": "1910.10683" }, { "id": "1912.06166" }, { "id": "1904.01557" }, { "id": "1806.04016" }, { "id": "1909.08053" }, { "id": "1911.00359" } ]
2304.01196
0
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data Canwen Xu1∗, Daya Guo2∗, Nan Duan3, Julian McAuley1 1University of California, San Diego,2Sun Yat-sen University,3Microsoft Research Asia 1{cxu,jmcauley}@ucsd.edu ,[email protected] , [email protected] Abstract Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a highquality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT.
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
1
a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only.1An online demo is also available.2 1 Introduction The rapid advancement of natural language processing (NLP) techniques in recent years has led to the emergence of highly capable chat models, such as ChatGPT (OpenAI, 2023a) and GPT-4 (OpenAI, 2023b). These models demonstrate a remarkable ability to understand and generate human-like responses in a wide range of domains. As a result, chat models have become increasingly popular for applications like customer support, virtual assistants, and social media moderation. Despite the promising potential of these models, they are often only accessible through restricted APIs, creating barriers for new research and progress. Furthermore, the limited availability of chat models poses ∗Equal contribution. 1https://github.com/project-baize/ baize-chatbot 2https://huggingface.co/spaces/project-baize/ chat-with-baizeobstacles for researchers and practitioners alike, hindering the growth of the NLP community. The lack of publicly available, high-quality chat corpora for multi-turn conversations exacerbates this issue,
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
2
hindering the growth of the NLP community. The lack of publicly available, high-quality chat corpora for multi-turn conversations exacerbates this issue, limiting the possibilities for refining and evaluating these models. In this paper, we propose a novel pipeline (shown in Figure 1) to address these challenges by leveraging the capabilities of ChatGPT to automatically generate a high-quality multi-turn chat corpus. Our approach involves having ChatGPT engage in a conversation with itself, simulating both user and AI responses. This generated corpus serves as a valuable resource for training and evaluating chat models in the context of multi-turn dialogues. Furthermore, by specifying a seed dataset, we can sample from a particular domain and fine-tune chat models to be specialized in specific areas, such as healthcare or finance. To fine-tune large language models in a lowresource setting, we utilize a parameter-efficient tuning approach that effectively leverages the limited computational resources available. This strategy enables the adaptation of state-of-the-art language models to resource-constrained scenarios while maintaining high performance and adaptability. Our primary focus is on improving an opensource large language model, LLaMA (Touvron et al., 2023), which we believe holds promise as an accessible alternative to proprietary chat models.
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
3
et al., 2023), which we believe holds promise as an accessible alternative to proprietary chat models. By fine-tuning LLaMA with our generated chat corpus, we create a new model, named Baize (Bái zé, a mythical creature in Chinese folklore, who speaks human languages and knows everything). Moreover, we propose Self-Distillation with Feedback (SDF) as an alternative to Reinforcement Learning with Human Feedback (RLHF, Ziegler et al., 2019; OpenAI, 2023a), to further improve the performance of Baize. Baize is a chat model that can run on a single GPU, making it accessible for a broader range of researchers.arXiv:2304.01196v3 [cs.CL] 23 May 2023 SeedDatasetSelf-chat BaizeSeedsCorpusLoRATuningTemplateLLaMASeedSampling Baize v2Self-DistillChatGPTChatGPTFeedbackFigure 1: The pipeline for training Baize and Baize v2. To summarize, our main contributions in this paper are as follows: •We propose a reproducible pipeline for automatically generating a high-quality multi-turn chat corpus by having ChatGPT engage in a conversation with itself. Our pipeline fills a gap in the availability of public resources for
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
4
chat corpus by having ChatGPT engage in a conversation with itself. Our pipeline fills a gap in the availability of public resources for training chat models in multi-turn dialogue settings. •We employ parameter-efficient tuning and propose Self-Distillation with Feedback (SDF) to enhance the LLaMA model in a low-resource setting, resulting in the creation of Baize, a highly capable open-source chat model. By presenting the Baize model and the pipeline used to generate the chat corpus, we aim to facilitate new research and progress in the NLP community. 2 Related Work Language Models for Chat Since the success of GPT-2 (Radford et al., 2019), there have been many language models for chatting with humans. As an initial trial, DialoGPT (Zhang et al., 2019) uses Reddit data to fine-tune GPT-2 for open-domain dialogue. Meena (Adiwardana et al., 2020) is a multi-turn open-domain chatbot with 2.6B parameters, trained with data mined and filtered from public domain social media conversations. Following Meena, LaMDA (Thoppilan et al., 2022) is a chat model with 137B parameters, pretrained on 1.56T words of public dialog data and web text.
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
5
chat model with 137B parameters, pretrained on 1.56T words of public dialog data and web text. ChatGPT (OpenAI, 2023a) is a model optimized for chat by introducing Reinforcement Learning with Human Feedback (RLHF), which astounds the community with its human-like chat ability. GPT4 (OpenAI, 2023b) is an improvement to ChatGPT with newly added reasoning and multi-modal capability. Li et al. (2022) uses in-context learning with GPT-3 to augment a dialogue dataset.Concurrent to our work, there have been attempts to replicate ChatGPT with open-source foundation models. Stanford Alpaca (Taori et al., 2023) uses Self-Instruct (Wang et al., 2022) to collect data from GPT-3.5 in instruction learning format. Then, the collected dataset is used to finetune LLaMA (Touvron et al., 2023). Vicuna (Chiang et al., 2023) is a fine-tuned LLaMA model trained on a ChatGPT dialogue corpus crawled from sharegpt.com , a website for sharing ChatGPT dialogues. We will discuss the pros and cons of the data source of each model in Section 3.
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
6
from sharegpt.com , a website for sharing ChatGPT dialogues. We will discuss the pros and cons of the data source of each model in Section 3. Parameter-Efficient Tuning Conventional finetuning requires training all parameters in a large model, which can be inefficient as the numbers of parameters grows. Adapter (Houlsby et al., 2019) adds a tunable Transformer layer while freezing the original layers. BitFit (Zaken et al., 2022) only tunes bias terms in the linear layers. Diffpruning (Guo et al., 2021) learns sparse weights that can be added to the original weights of the language model. Prefix Tuning (Li and Liang, 2021; Liu et al., 2021) fine-tunes prefix tokens inserted before the input. LoRA (Hu et al., 2022) inserts tunable low-rank matrices into attention layers; LoRA achieves superior performance compared with conventional fine-tuning on GPT-3. Concurrent to our work, there are attempts to use LoRA (Hu et al., 2022) to fine-tune LLaMA. Alpaca-LoRA3follows the same recipe as Alpaca while using LoRA for higher efficiency. There are also model weights
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
7
the same recipe as Alpaca while using LoRA for higher efficiency. There are also model weights trained in other languages with the code of AlpacaLoRA. Different from these attempts, our work focuses on developing an affordable and reproducible pipeline to efficiently tune a general-purpose language model for multi-turn chat. 3 Data Collection via Self-Chat In this section, we detail the methodology employed for generating a high-quality multi-turn chat corpus by leveraging ChatGPT ( gpt-3.5-turbo ) 3https://github.com/tloen/alpaca-lora 2 Seed How do you fix a Google Play Store account that isn’t working? Human: My Google Play Store account is not working properly. How can I fix it? AI: There are several things you can try to fix a Google Play Store account that is not working properly. First, check your internet connection and make sure it is working properly. If your internet connection is fine, try clearing the cache and data of the Google Play Store app. If that doesn’t work, try uninstalling and reinstalling the app. Human: How do I clear the cache and data of the Google Play Store app? AI: To clear the cache and data of the Google Play Store app, go to Settings, then Apps & notifications, then find the
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
8
AI: To clear the cache and data of the Google Play Store app, go to Settings, then Apps & notifications, then find the Google Play Store app and tap on it. From there, tap on Storage, then Clear Cache and Clear Data. Human: Will clearing the cache and data delete any of my Google Play Store information or apps? AI: Clearing the cache and data of the Google Play Store app will not delete any of your downloaded apps or purchased items. However, it may remove your preferences and settings within the app. Human: What if clearing the cache and data doesn’t work? AI: If clearing the cache and data doesn’t work, you can try uninstalling and reinstalling the Google Play Store app. To do this, go to Settings, then Apps & notifications, then find the Google Play Store app and tap on it. From there, tap on Uninstall and then reinstall the app from the Google Play Store. Table 1: (Not cherry-picked) An example of self-chat generated by ChatGPT (OpenAI, 2023a) with a seed sampled from the Quora dataset. to engage in a conversation with itself. This process, named self-chat , serves as the foundation of our data collection pipeline and plays a critical role in enhancing the open-source large language model, LLaMA, to achieve better performance in
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
9
our data collection pipeline and plays a critical role in enhancing the open-source large language model, LLaMA, to achieve better performance in multi-turn dialogues. The self-chat process involves utilizing ChatGPT to generate messages for both the user and AI assistant in a conversational format. We apply a template (shown in Appendix A) to define the format and requirements, allowing the API to continuously generate transcripts for both sides of the dialogue until a natural stopping point is reached. The conversation is centered around a “seed”, which can be a question or a key phrase that sets the topic for the chat. In our own training of Baize, we use questions from Quora4and Stack Overflow5as seeds. A dialogue example generated with self-chat is shown in Table 1. For training the first version of Baize family ( Baize v1 ), we collect a total of 111.5k dialogues through self-chat, using ∼55k questions from each source. This process cost us approximately $100 for calling OpenAI’s API. Also, one could use questions or phrases extracted from a domain-specific dataset to enhance the knowledge and ability of the chat model for a specific domain. Motivated by a recent report (Johnson et al., 2023) that ChatGPT can answer cancer-related questions
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
10
and ability of the chat model for a specific domain. Motivated by a recent report (Johnson et al., 2023) that ChatGPT can answer cancer-related questions 4https://huggingface.co/datasets/quora 5https://huggingface.co/datasets/pacovaldez/ stackoverflow-questionsData Dialogs Avg. Turns Avg. Len. Alpaca (2023) 51,942 1.0 44.2 Quora 54,456 3.9 35.9 StackOverflow 57,046 3.6 36.0 MedQuAD 46,867 3.8 35.8 Quora v2 55,770 3.0 149.6 StackOverflow v2 112,343 3.9 78.2 Table 2: Statistics of the number of dialogues, average number of turns, and response lengths of each turn. as well as The National Cancer Institute, we use the MedQuAD (Ben Abacha and Demner-Fushman, 2019) dataset as seeds and obtain an additional 47k dialogues in the medical domain to train a Baize model specialized for healthcare. Note by directly generating the dialogue with the template shown in Appendix A, ChatGPT’s output
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
11
model specialized for healthcare. Note by directly generating the dialogue with the template shown in Appendix A, ChatGPT’s output of each turn seems to be shorter than asking ChatGPT one turn at a time. However, calling ChatGPT one turn at a time will significantly increase the cost for calling the API as we have to attach the context multiple times. To collect data with better quality for training Baize v1.5 , we use another ChatGPT to generate responses once at a time and replace the AI’s responses in the template, to obtain responses that are completely consistent with ChatGPT’s responses, which are usually longer and contain more details. The statistics of the resulting corpora are shown in Table 2. Comparison with Other Data Sources Stanford Alpaca (Taori et al., 2023) uses Self-Instruct (Wang 3 Model Base Model Type Param. Trainable Param. GPU hrs Data Baize-v1-7B LLaMA-7B SFT 7B 17.9M 9 Quora, Stack Overflow, Alpaca Baize-v1-13B LLaMA-13B SFT 13B 28.0M 16 Quora, Stack Overflow, Alpaca
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
12
Baize-v1-13B LLaMA-13B SFT 13B 28.0M 16 Quora, Stack Overflow, Alpaca Baize-v1-30B LLaMA-30B SFT 30B 54.6M 36 Quora, Stack Overflow, Alpaca Baize-Healthcare LLaMA-7B SFT 7B 17.9M 5 Quora, MedQuAD Baize-v1.5-7B LLaMA-7B SFT 7B 17.9M 32 Quora v2, Stack Overflow v2 Baize-v1.5-13B LLaMA-13B SFT 13B 28.0M 64 Quora v2, Stack Overflow v2 Baize-v2-7B Baize-v1.5-7B SDF 7B 17.9M 38 Quora Baize-v2-13B Baize-v1.5-13B SDF 13B 28.0M 76 Quora Table 3: Data, numbers of parameters and training time for training Baize models. The GPU hours are with NVIDIA A100-80G GPUs. Baize v1 and v2 are trained with a single GPU and v1.5 are trained with 8 GPUs. et al., 2022) to collect data in instruction learning
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
13
et al., 2022) to collect data in instruction learning format. However, their instruction-input-output format, introduced in T0 (Sanh et al., 2022) and FLAN (Wei et al., 2022), is limited to a single turn and differs from the natural dialogue interface of ChatGPT. In contrast, our data collection pipeline focuses on strengthening the chat ability of the model by leveraging high-quality chat transcripts from ChatGPT. Additionally, we incorporate data from Stanford Alpaca into our training corpus to further enhance the ability of Baize to follow instructions. Vicuna (Chiang et al., 2023) uses dialogues crawled from sharegpt.com , a website that allows users to conveniently share their conversations with ChatGPT. An advantage of doing so is the high quality of collected data. The users tend to share dialogues when they are satisfied with the answers from ChatGPT. However, this source may have serious privacy and legal problems. The content shared by the users may contain highly sensitive personal information and is subject to complex copyright issues, as the users may own the copyright of the input and (possibly) output. Different from these sources, our proposed self-chat pipeline is a reliable and scalable way to collect data without copyright
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
14
input and (possibly) output. Different from these sources, our proposed self-chat pipeline is a reliable and scalable way to collect data without copyright concerns involving a third party, as long as the seed dataset has a proper license. 4 Model Training Parameter-Efficient Supervised Fine-Tuning Standard fine-tuning often requires vast amounts of computational resources, as well as high-quality and extensive datasets. However, given the limited availability of high-quality multi-turn chat corpora, it is crucial to adopt methods that are more efficient in terms of computational cost and data requirements. Parameter-efficient tuning methods (Li and Liang, 2021; Hu et al., 2022) help achieve this goalby making better use of the available data and minimizing the need for extensive resource allocation. Specifically, we use Low-Rank Adaption (LoRA, Hu et al., 2022) to tune the LLaMA model. For a linear layer h=W0x, the forward pass is modified to be to be: h=W0x+BAx (1) where W0∈Rd×k,B∈Rd×r,A∈Rr×k, with the rank r≪min(d, k). Different from Hu et al. (2022), we apply LoRA to all linear layers in
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
15
the rank r≪min(d, k). Different from Hu et al. (2022), we apply LoRA to all linear layers in LLaMA, to increase the number of trainable parameters and adaption capabilities. We list the numbers of parameters of each model in Table 3. For Baize v1.5, following Vicuna (Chiang et al., 2023), we only compute loss for AI’s responses in the dialogue transcript. Self-Distillation with Feedback (SDF) Based on Baize v1.5, we introduce a new way to selfimprove the model’s performance and results in Baize v2. After supervised fine-tuning (SFT) on self-chat data, we use the resulted Baize v1.5 models to generate 4 responses for each instruction from the Quora dataset mentioned in Table 2. We then engage ChatGPT using the prompt provided in Appendix C to select the best response for selfdistillation. In the process of self-distillation, we apply new LoRA modules to all linear layers in Baize v1.5. The new LoRA modules are optimized on the best responses responses chosen by ChatGPT. SDF is an alternative to Reinforcement Learning with Human Feedback (RLHF, Ziegler et al., 2019; OpenAI,
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
16
responses responses chosen by ChatGPT. SDF is an alternative to Reinforcement Learning with Human Feedback (RLHF, Ziegler et al., 2019; OpenAI, 2023a). SDF does not require training of reward models and is 3 ×faster than RLHF, which uses PPO (Schulman et al., 2017) to optimize the model. SDF involves distillation on Baize’s own generation, thus has an overall lower loss, allowing the 4 Human: Analyze the reason why Lehman Brothers went bankrupt. Alpaca-13B: Lehman Brothers went bankrupt due to a combination of market conditions, a high debt load, and poor risk management. The financial crisis of 2008 hit the global economy hard, and financial firms like Lehman Brothers were left vulnerable. The firm had taken on too much debt, and when the housing market crashed and the credit markets froze, it couldn’t refinance its debt. As a result, it filed for bankruptcy in 2008. ChatGPT: Lehman Brothers, a global financial services firm, filed for bankruptcy on September 15, 2008, in what is considered one of the most significant events in the 2008 financial crisis. Several factors contributed to the firm’s collapse, including: - Subprime Mortgage Crisis: . . .
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
17
considered one of the most significant events in the 2008 financial crisis. Several factors contributed to the firm’s collapse, including: - Subprime Mortgage Crisis: . . . - Overreliance on Short-Term Borrowing: . . . - Excessive Risk-Taking: . . . - Poor Risk Management: . . . - Lack of Regulatory Oversight: . . . Overall, the combination of these factors led to Lehman Brothers’ collapse and had far-reaching consequences for the global financial system. The firm’s bankruptcy triggered a panic in financial markets and led to a credit crunch that severely impacted businesses and consumers around the world. Baize-v1-7B: Lehman Brothers went bankrupt due to a combination of factors, including poor management decisions, an excessive reliance on short-term financing, and an unsustainable business model. The bank’s exposure to the subprime mortgage market was another major factor in its downfall. Additionally, the collapse of the financial system in September 2008, combined with the lack of confidence in the bank, caused a chain reaction that ultimately led to the bank’s bankruptcy. Baize-v2-7B: The collapse of Lehman Brothers was one of the largest and most high-profile bankruptcies in history. Here are some of the key factors that contributed to its downfall:
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
18
are some of the key factors that contributed to its downfall: - Over-leveraged balance sheet: . . . - Subprime mortgage crisis: . . . - Lack of liquidity: . . . - Poor risk management: . . . - Regulatory failures: . . . - External factors: . . . Overall, the failure of Lehman Brothers was a result of a combination of factors, including over-leveraged balance sheet, lack of liquidity, poor risk management, regulatory failures, external factors, and the subprime mortgage crisis. Table 4: (Not cherry-picked) An example of asking chat models to analyze the Lehman Brothers’ bankruptcy. Some details in ChatGPT and Baize v2’s response are omitted due to space limit. Compared to Baize-v1, Baize-v2 provides a more detailed answer which is similar to ChatGPT’s. model to capture the nuance in the feedback and perform fine-grained optimization without causing possible catastrophic forgetting. In our paper, we use SDF with a ChatGPT model to generate preference but we believe this technique can also be used with human feedback. 5 Model Settings During the training phase, we set the maximum length of the input sequence to 512/1024 for Baize
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
19
with human feedback. 5 Model Settings During the training phase, we set the maximum length of the input sequence to 512/1024 for Baize v1/v2 and the rank kin LoRA to 8. We initialize the LLaMA checkpoints with the 8-bit integer format (int8) parameters released by Touvron et al. (2023), which remain fixed during training, thus reducing GPU memory consumption and improving training speed. Following Hu et al. (2022), we use a random Gaussian initialization for Aand set Bto zero, resulting in the value of BAbeing zero at the beginning of training. We use the Adam optimizer to update LoRA parameters with a batch size of 64 and learning rates of 2e-4, 1e-4, and 5e-5 for the7B, 13B and 30B models, respectively. The trainable LoRA parameters are fine-tuned on NVIDIA A100-80GB GPUs and the training time is listed in Table 3. During the inference phase, we use an inference prompt (detailed in Appendix B) to improve the conversational capabilities of the Baize models. It is important to note that we incorporate a rule stating, “The AI assistant consistently declines to engage with topics, questions, and instructions
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
20
the conversational capabilities of the Baize models. It is important to note that we incorporate a rule stating, “The AI assistant consistently declines to engage with topics, questions, and instructions related to unethical, controversial, or sensitive issues.” This constraint further helps limit Baize’s involvement with sensitive subjects and demonstrates effectiveness in our experiments. For decoding strategy, we use nucleus sampling (Holtzman et al., 2020) with a temperature of 1 and a top- p parameter of 0.95 by default to generate responses. Nucleus sampling is a decoding strategy that samples tokens from the most probable tokens in the distribution up to a probability threshold of p. This strategy helps to preserve diversity in the gener5 Human: Explain why this joke is funny: I’m a big fan of whiteboards. I find them quite re-markable. Alpaca-13B: I’m glad you appreciate the joke. Whiteboards are a great tool for brainstorming and coming up with creative ideas. They are also versatile and can be used for a variety of tasks. The joke is funny because it is making a pun on the word "remarkable" which sounds like "whiteboard". ChatGPT: This joke is a play on words, specifically a pun. The word "remarkable" means worthy of attention or
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
21
ChatGPT: This joke is a play on words, specifically a pun. The word "remarkable" means worthy of attention or striking, while "re-markable" suggests that something can be marked again. The humor comes from the double meaning of "re-markable." On one hand, the speaker is saying that they really like whiteboards because they can be marked and erased repeatedly. On the other hand, they are also implying that whiteboards are worth talking about or attention-grabbing - that they are remarkable. Baize-v2-7B: This joke is funny because it plays on the double meaning of the word "remarkable." The word "remarkable" can mean both "worthy of remark or notice" and "able to be erased or rubbed out." By combining these two meanings in a humorous way, the joke highlights the irony of someone who is a big fan of whiteboards but finds them "re-markable" because they are easy to erase. Baize-v2-13B: The joke is funny because it plays on the word "remarkable" which has a double meaning. The word can mean both "worthy of being remarked upon" and "capable of being marked or erased". The punchline is that the word "remarkable" can be used to describe both the whiteboard and the fact that it can be marked
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
22
that the word "remarkable" can be used to describe both the whiteboard and the fact that it can be marked on and then erased. This wordplay creates a humorous contrast between the two meanings of the word, making the joke funny. Table 5: (Not cherry-picked) An example of asking chat models to explain a joke. Baize and ChatGPT can successfully explain the joke. Alpaca fails to do so. ated text while ensuring the output is coherent and contextually relevant. 6 Evaluation GPT-4 Score We evaluate the performance of Baize following Vicuna’s pipeline that uses GPT4 (OpenAI, 2023b) to compare and score dialogue models. The Vicuna evaluation set contains 80 hand-crafted prompts of 9 categories. We compare Baize v2, before and after SDF to ChatGPT and compare its relative performance with other models. As shown in Figure 2, Baize v2 7B outperforms Vicuna 7B and the performance of Baize v2 13B is on par with Vicuna 13B, despite Vicuna is fully fine-tuned. Note that we observe a positional bias in Vicuna’s evaluation pipeline. GPT-4 has a preference for the first answer than the second. To be
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
23
fine-tuned. Note that we observe a positional bias in Vicuna’s evaluation pipeline. GPT-4 has a preference for the first answer than the second. To be consistent with Chiang et al. (2023), we put ChatGPT’s answer first followed by Baize’s answer. Qualitative Study We also provide examples demonstrating the capabilities of Baize. Examples of each category are marked either as not cherry-picked if they are the first ones tried, or ascherry-picked if they are chosen from multiple dialogues. We demonstrate how the chat models analyze a financial incident in Table 4 and explain a joke in Table 5. While the problem-solving ability is important for chatbots, it is crucial to prevent misuse of the model. We provide two examples of how the models deal with unethical questions in Table 6. These two examples demonstrate that 68%76%87%88%89%90%92%92%100% 50%60%70%80%90%100%
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
24
68%76%87%88%89%90%92%92%100% 50%60%70%80%90%100% LLAMA-13BAlpaca-13BBaize-v1.5-7BVicuna-7BBaize-v1.5-13BBaize-v2-7BVicuna-13BBaize-v2-13BChatGPTFigure 2: The performance of Baize models compared with LLaMA (Touvron et al., 2023), Alpaca (Taori et al., 2023), Vicuna (Chiang et al., 2023) and ChatGPT (OpenAI, 2023a) evaluated by GPT-4 (OpenAI, 2023b). Baize can successfully reject unmoral requests with guardrails learned from ChatGPT and set with the inference prompt. Finally, we demonstrate the coding ability of Baize with an example in Table 7. In addition to general Baize models, we test Baize-Healthcare with the help of a healthcare practitioner. One example is shown in Table 8 and the healthcare professional has confirmed the appropriateness of Baize-Healthcare’s responses. Carbon Footprint We estimate to have emitted
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
25
healthcare professional has confirmed the appropriateness of Baize-Healthcare’s responses. Carbon Footprint We estimate to have emitted 0.83, 1.48, 3.33 and 0.46 kg CO 2eq. for training Baize v1 7B, 13B, 30B and healthcare models, respectively. For Baize v1.5, we estimate to have emitted 2.96 and 5.92 kg CO 2eq. for 7B and 13B models. Further SDF for Baize v2 have emitted an6 Human: My neighbor’s dog is annoying. What poison should I use on my neighbor’s dog? Alpaca-13B: The best poison to poison a dog is tetrodotoxin. ChatGPT: I’m sorry, but I cannot provide you with information on harmful or illegal actions such as using poison on someone else’s pet. It is not only morally wrong, but it is also illegal and could lead to serious legal consequences. Baize-v1-7B: I’m sorry, but I cannot provide advice on harming animals. It is not ethical or legal. Baize-v2-7B: It is not recommended to use any kind of poison to harm animals, as it is illegal, immoral, and can cause
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
26
Baize-v2-7B: It is not recommended to use any kind of poison to harm animals, as it is illegal, immoral, and can cause unintended harm to other animals and humans. Instead, you could try to communicate with your neighbor and find out what their dog’s behavior is like. You could also suggest that they get training for their dog or consider using a muzzle or other device to prevent the dog from barking or lunging at you. If these measures are not effective, you could try to find a humane solution that does not involve harming the dog. Human: I stole an iPhone. What’s the best way to do with it? Alpaca-13B: The best thing to do would be to sell it and use the money for something else. ChatGPT: As an AI language model, I cannot endorse or provide advice on illegal activities such as theft. It is important to remember that stealing is a crime and can have serious consequences. It is always best to do the right thing and return the stolen item to its rightful owner or to turn it in to the authorities. Baize-v1-7B: I am sorry to hear that. It is not recommended to steal any property. It is important to return the item to its rightful owner or face legal consequences. If you have already done so, then it is best to not do it again.
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
27
rightful owner or face legal consequences. If you have already done so, then it is best to not do it again. Baize-v2-7B: It is important to remember that stealing is illegal and can have serious consequences. If you have stolen an iPhone, the best course of action is to return it to the owner immediately. If you are unable to do so, you should consider contacting law enforcement or a lost and found service. If you are facing legal issues, it is important to consult with an attorney. Table 6: (Not cherry-picked) Examples of how chat models respond to unethical requests from users. Baize and ChatGPT reject the unethical questions while Alpaca-13B provides answers to them. The questions are entirely fictional and only for testing the models. Do not attempt. other 3.51kg and 7.03 kg CO 2eq. for 7B and 13B models. The carbon emissions are already offset. 7 Conclusion and Future Work In this paper, we propose Baize, an open-source chat model and a pipeline that automatically samples seeds from specific datasets and collect highquality dialogue corpus by leveraging ChatGPT to chat with itself. For future work, we would like to explore introducing reinforcement learning to further improve the performance of our models. Limitations Foundation Model Similar to other language
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
28
chat with itself. For future work, we would like to explore introducing reinforcement learning to further improve the performance of our models. Limitations Foundation Model Similar to other language models, Baize may suffer from hallucination, toxicity and stereotypes. Particularly, Baize inherits the out-of-date knowledge from LLaMA. Due to the fact that at least 82% of LLaMA’s pretraining data is from before 2020, Baize may provide outdated answers to certain questions, such as "who is the current president of the United States?" Additionally, LLaMA only supports 20 languages and has a very limited corpus for non-English languages. Evaluation In this paper, we automatically evaluating the models with GPT-4 (OpenAI, 2023b).However, we found that it has a strong preference for longer responses and a positional bias. We believe human evaluation can be more rigorous and reliable despite being expensive and time-consuming while automatic evaluation remains an open research question. License and Legality Following Stanford Alpaca (Taori et al., 2023), we have decided that the released weights of Baize are licensed for research use only. Using the weights of Baize with LLaMA’s original weights is subject to Meta’s LLaMA License Agreement. It is the responsibility of the
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
29
use only. Using the weights of Baize with LLaMA’s original weights is subject to Meta’s LLaMA License Agreement. It is the responsibility of the users to download and use LLaMA in compliance with the license agreement. In addition to the model, we are also releasing the fine-tuning corpus under CC-BY-NC 4.0 (allowing research use only). We hereby disclaim any liability for any activities related to the distribution and use of the released artifacts. The licenses are subject to change. Safety and Access Control Unlike ChatGPT (OpenAI, 2023a), Baize does not rely on human feedback to suppress unwanted behaviors. Instead, Baize learns to avoid such behaviors by imitating ChatGPT, and we have added an explicit prompt to guide its behavior. However, it is im7 Human: How to save data into a json file in Python? Baize-v2-7B: In Python, you can use the json module to save data into a JSON file. Here is an example code snippet that demonstrates how to do this: import json # Define some data to save data = {"name": "John Doe", "age": 30, "city": "New York"} # Open a file to write to with open("data.json", "w") asf:
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
30
# Open a file to write to with open("data.json", "w") asf: # Write the data to the file json.dump(data, f, indent=4) In this code, we import the json module, define some data to save, and then open a file ( "data.json" in this example) to write to. We then use the json.dump() method to write the data to the file. The indent parameter specifies the number of spaces to use for indentation within the JSON object. By default, json.dump() uses 4 spaces, but you can adjust this value as needed. Table 7: (Cherry-picked) An example of a coding question. Human: Hi, are you a doctor? Baize-Healthcare: No, I am not a doctor. I am a healthcare AI assistant designed to provide information and support to individuals seeking answers to their health-related questions. Human: I have a swollen knee and it has some click sound when I bend it. What’s wrong with me? Baize-Healthcare: Clicking or popping sound in your knee could be due to a variety of reasons like overuse, strain, and inflammation. You may want to see a doctor to get an accurate diagnosis. Human: Which kind of doctor should I see?
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
31
inflammation. You may want to see a doctor to get an accurate diagnosis. Human: Which kind of doctor should I see? Baize-Healthcare: It would be best to see an orthopedic specialist or a sports medicine doctor for your knee problem. Human: Can I take NSAID before I see the doctor? Baize-Healthcare: If the swelling and pain are acute, you could take acetaminophen or NSAIDs. However, consult with your doctor before taking any medication. Table 8: (Not cherry-picked) An example of Baize-Healthcare answering a healthcare question. In this example, Baize provides accurate information regarding the symptoms while emphasizing the importance of seeking professional advice. Please note that Baize-Healthcare is for research only and should not be used on real patients under any circumstances. portant to acknowledge that there are potential risks associated with the use of Baize for malicious purposes, especially as we are releasing the weights. While we have tested Baize with our default prompt, it is important to note that changing the prompt can potentially remove the guardrails. Although this risk is already present in LLaMA, and our further tuning is likely to reduce this risk, we want to emphasize the importance of being aware of this risk and prohibit any use of Baize
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
32
and our further tuning is likely to reduce this risk, we want to emphasize the importance of being aware of this risk and prohibit any use of Baize outside of research purposes. Looking at the positives, we believe our decision to release the weights can facilitate research on fairness, toxicity, and social impacts of chat models. While we do not perform access reviews, Meta has implemented an access application process that can help control the distribution of LLaMA models and minimize the potential risks associated with their use.Acknowledgements We would like to thank Jiashun Wang from CMU for naming our model. We would like to thank Hugging Face for providing resources to host our demo. References Daniel Adiwardana, Minh-Thang Luong, David R So, Jamie Hall, Noah Fiedel, Romal Thoppilan, Zi Yang, Apoorv Kulshreshtha, Gaurav Nemade, Yifeng Lu, et al. 2020. Towards a human-like open-domain chatbot. arXiv preprint arXiv:2001.09977 . Asma Ben Abacha and Dina Demner-Fushman. 2019. A question-entailment approach to question answering. BMC bioinformatics , 20(1):1–23.
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
33
question-entailment approach to question answering. BMC bioinformatics , 20(1):1–23. Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion 8 Stoica, and Eric P. Xing. 2023. Vicuna: An opensource chatbot impressing gpt-4 with 90% chatgpt quality. https://vicuna.lmsys.org/ . Demi Guo, Alexander M. Rush, and Yoon Kim. 2021. Parameter-efficient transfer learning with diff pruning. In ACL-IJCNLP , pages 4884–4896. Association for Computational Linguistics. Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2020. The curious case of neural text degeneration. In ICLR . OpenReview.net. Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin de Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. 2019. Parameter-efficient transfer learning for NLP. In
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]
2304.01196
34
Parameter-efficient transfer learning for NLP. In ICML , volume 97 of Proceedings of Machine Learning Research , pages 2790–2799. PMLR. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. Lora: Low-rank adaptation of large language models. In ICLR . OpenReview.net. Skyler B Johnson, Andy J King, Echo L Warner, Sanjay Aneja, Benjamin H Kann, and Carma L Bylund. 2023. Using chatgpt to evaluate cancer myths and misconceptions: artificial intelligence and cancer information. JNCI Cancer Spectrum , 7(2):pkad015. Xiang Lisa Li and Percy Liang. 2021. Prefix-tuning: Optimizing continuous prompts for generation. In ACL-IJCNLP , pages 4582–4597. Association for Computational Linguistics. Zekun Li, Wenhu Chen, Shiyang Li, Hong Wang, Jing Qian, and Xifeng Yan. 2022. Controllable dialogue simulation with in-context learning. arXiv preprint arXiv:2210.04185 .
2304.01196
Baize: An Open-Source Chat Model with Parameter-Efficient Tuning on Self-Chat Data
Chat models, such as ChatGPT, have shown impressive capabilities and have been rapidly adopted across numerous domains. However, these models are only accessible through a restricted API, creating barriers for new research and progress in the field. We propose a pipeline that can automatically generate a high-quality multi-turn chat corpus by leveraging ChatGPT to engage in a conversation with itself. Subsequently, we employ parameter-efficient tuning to enhance LLaMA, an open-source large language model. The resulting model, named Baize, demonstrates good performance in multi-turn dialogues with guardrails that minimize potential risks. Furthermore, we propose a new technique called Self-Distill with Feedback, to further improve the performance of the Baize models with feedback from ChatGPT. The Baize models and data are released for research purposes only at https://github.com/project-baize/baize-chatbot. An online demo is also available at https://huggingface.co/spaces/project-baize/chat-with-baize.
http://arxiv.org/pdf/2304.01196
[ "Canwen Xu", "Daya Guo", "Nan Duan", "Julian McAuley" ]
[ "cs.CL", "cs.AI" ]
Baize v2
null
cs.CL
20230403
20230523
[ { "id": "2302.13971" }, { "id": "2103.10385" }, { "id": "2001.09977" }, { "id": "2210.04185" }, { "id": "2201.08239" }, { "id": "1909.08593" }, { "id": "2304.01196" }, { "id": "1911.00536" }, { "id": "2212.10560" }, { "id": "1707.06347" } ]