llm

app.py

@@ -2,6 +2,8 @@ import gradio as gr
 from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer
 import pandas as pd
 from sklearn.metrics import accuracy_score, precision_recall_fscore_support
+import openai
+import requests
 
 # Define the available models and tasks
 TASKS = ["sentiment-analysis", "ner", "text-classification"]
@@ -9,26 +11,54 @@ MODELS = {
     "DistilBERT": "distilbert-base-uncased",
     "BERT": "bert-base-uncased",
     "RoBERTa": "roberta-base",
-    "LLaMA": "decapoda-research/llama-7b-hf"
-    # Add other models here
+    "LLaMA2_7B_chat": "meta-llama/Llama-2-7b-chat-hf",
+    "LLaMA2_70B": "meta-llama/Llama-2-70b-hf",
+    "ChatGLM3_6B": "THUDM/chatglm-6b",
+    "InternLM_7B": "internlm/internlm-7b",
+    "Falcon_7B": "tiiuae/falcon-7b",
+    # Add other Hugging Face models here
 }
 
+# Set up OpenAI API key
+openai.api_key = "YOUR_OPENAI_API_KEY"
+
+# Function to load pipeline for Hugging Face models
 def load_pipeline(task, model):
     model_name = MODELS[model]
     return pipeline(task, model=model_name)
 
+# Function to predict using Hugging Face models
 def predict(task, model, text):
-    selected_pipeline = load_pipeline(task, model)
-    results = selected_pipeline(text)
+    if model.startswith("ChatGPT"):
+        response = openai.Completion.create(
+            engine="text-davinci-003",
+            prompt=text,
+            max_tokens=50
+        )
+        results = [{"label": response.choices[0].text.strip()}]
+    else:
+        selected_pipeline = load_pipeline(task, model)
+        results = selected_pipeline(text)
     return results
 
+# Function to benchmark Hugging Face models
 def benchmark(task, model, file):
     data = pd.read_csv(file.name)
-    texts = data['text'].tolist()
-    true_labels = data['label'].tolist()
+    texts = data['query'].tolist()
+    true_labels = data['answer'].tolist()
     
-    selected_pipeline = load_pipeline(task, model)
-    predictions = [selected_pipeline(text)[0]['label'] for text in texts]
+    if model.startswith("ChatGPT"):
+        predictions = []
+        for text in texts:
+            response = openai.Completion.create(
+                engine="text-davinci-003",
+                prompt=text,
+                max_tokens=50
+            )
+            predictions.append(response.choices[0].text.strip())
+    else:
+        selected_pipeline = load_pipeline(task, model)
+        predictions = [selected_pipeline(text)[0]['label'] for text in texts]
     
     accuracy = accuracy_score(true_labels, predictions)
     precision, recall, f1, _ = precision_recall_fscore_support(true_labels, predictions, average='macro')
@@ -44,7 +74,7 @@ def benchmark(task, model, file):
 with gr.Blocks() as demo:
     with gr.Row():
         task_input = gr.Dropdown(TASKS, label="Task")
-        model_input = gr.Dropdown(list(MODELS.keys()), label="Model")
+        model_input = gr.Dropdown(list(MODELS.keys()) + ["ChatGPT"], label="Model")
     
     with gr.Tab("Predict"):
         with gr.Row():