# SpQR

Algoritma kuantisasi [SpQR](https://hf.co/papers/2306.03078) melibatkan struktur kuantisasi 3-bit grup bi-level bertile 16x16 dengan outlier yang jarang.

    

> [!TIP]
> Untuk mengkuantisasi model dengan SpQR, lihat repositori [Vahe1994/SpQR](https://github.com/Vahe1994/SpQR).

Muat model yang dikuantisasi SpQR dengan [from_pretrained()](/docs/transformers/v5.23.0/id/main_classes/model#transformers.PreTrainedModel.from_pretrained).

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

quantized_model = AutoModelForCausalLM.from_pretrained(
    "elvircrn/Llama-2-7b-SPQR-3Bit-16x16-red_pajama-hf",
    dtype=torch.half,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("elvircrn/Llama-2-7b-SPQR-3Bit-16x16-red_pajama-hf")
```

