---
tags:
- fp8
- vllm
---

# Qwen2-1.5B-Instruct-FP8

## Model Overview
Qwen2-1.5B-Instruct quantized to FP8 weights and activations using per-tensor quantization, ready for inference with vLLM >= 0.5.0.

## Usage and Creation
Produced using [AutoFP8 with calibration samples from ultrachat](https://github.com/neuralmagic/AutoFP8/blob/147fa4d9e1a90ef8a93f96fc7d9c33056ddc017a/example_dataset.py).

```python
from datasets import load_dataset
from transformers import AutoTokenizer

from auto_fp8 import AutoFP8ForCausalLM, BaseQuantizeConfig

pretrained_model_dir = "Qwen/Qwen2-1.5B-Instruct"
quantized_model_dir = "Qwen2-1.5B-Instruct-FP8"

tokenizer = AutoTokenizer.from_pretrained(pretrained_model_dir, use_fast=True, model_max_length=4096)
tokenizer.pad_token = tokenizer.eos_token

ds = load_dataset("mgoin/ultrachat_2k", split="train_sft").select(range(512))
examples = [tokenizer.apply_chat_template(batch["messages"], tokenize=False) for batch in ds]
examples = tokenizer(examples, padding=True, truncation=True, return_tensors="pt").to("cuda")

quantize_config = BaseQuantizeConfig(quant_method="fp8", activation_scheme="static")

model = AutoFP8ForCausalLM.from_pretrained(
    pretrained_model_dir, quantize_config=quantize_config
)
model.quantize(examples)
model.save_quantized(quantized_model_dir)
```

## Evaluation

### Open LLM Leaderboard evaluation scores
|                      | Qwen2-1.5B-Instruct       | Qwen2-1.5B-Instruct-FP8<br>(this model) |
| :------------------: | :----------------------: | :------------------------------------------------: |
| arc-c<br>25-shot     | 43.09                    | 41.81                                              |
| hellaswag<br>10-shot | 67.48                    | 67.18                                              |
| mmlu<br>5-shot       | 55.87                    | 55.60                                              |
| truthfulqa<br>0-shot | 43.34                    | 43.09                                              |
| winogrande<br>5-shot | 63.61                    | 63.38                                              |
| gsm8k<br>5-shot      | 57.70                     | 56.48                                             |
| **Average<br>Accuracy**  | **55.18**                    |              **54.59**                                      |
| **Recovery**             | **100%**                     |              **98.93%**                                     |