Mistral-Small-24B-Instruct-2501-FP8-Dynamic

Model Overview

• Model Architecture: Mistral-Small-24B-Instruct-2501
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Release Date: 3/1/2025
• Version: 1.0
• Model Developers: Neural Magic

Quantized version of Mistral-Small-24B-Instruct-2501. It achieves a flexible-extract filter score of 0.9030 on the evaluated on GSM8k task, where as the unquantized model achieves a flexible-extract filter score of 0.9060.

Model Optimizations

This model was obtained by quantizing the weights and activations to FP8 data type, ready for inference with vLLM >= 0.5.2. This optimization reduces the number of bits per parameter from 16 to 8, reducing the disk size and GPU memory requirements by approximately 50%. Only the weights and activations of the linear operators within transformers blocks are quantized.

Deployment

Use with vLLM

This model can be deployed efficiently using the vLLM backend, as shown in the example below.

from transformers import AutoTokenizer
from vllm import LLM, SamplingParams

max_model_len, tp_size = 4096, 1
model_name = "nm-testing/Mistral-Small-24B-Instruct-2501-FP8-Dynamic"
tokenizer = AutoTokenizer.from_pretrained(model_name)
llm = LLM(model=model_name, tensor_parallel_size=tp_size, max_model_len=max_model_len, trust_remote_code=True)
sampling_params = SamplingParams(temperature=0.3, max_tokens=256, stop_token_ids=[tokenizer.eos_token_id])

messages_list = [
    [{"role": "user", "content": "Who are you? Please respond in pirate speak!"}],
]

prompt_token_ids = [tokenizer.apply_chat_template(messages, add_generation_prompt=True) for messages in messages_list]

outputs = llm.generate(prompt_token_ids=prompt_token_ids, sampling_params=sampling_params)

generated_text = [output.outputs[0].text for output in outputs]
print(generated_text)

vLLM also supports OpenAI-compatible serving. See the documentation for more details.

Creation

This model was created with llm-compressor by running the code snippet below.

import argparse
from transformers import AutoModelForCausalLM, AutoTokenizer
from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.transformers import oneshot
import os

def main():
    parser = argparse.ArgumentParser(description='Quantize a transformer model to FP8')
    parser.add_argument('--model_id', type=str, required=True,
                        help='The model ID from HuggingFace (e.g., "mistralai/Mistral-Small-24B-Instruct-2501")')
    parser.add_argument('--save_path', type=str, default='.',
                        help='Custom path to save the quantized model. If not provided, will use model_name-FP8-dynamic')
    args = parser.parse_args()

    # Load model
    model = AutoModelForCausalLM.from_pretrained(
        args.model_id, device_map="auto", torch_dtype="auto", trust_remote_code=True,
    )
    tokenizer = AutoTokenizer.from_pretrained(args.model_id)

    # Configure the quantization algorithm and scheme
    recipe = QuantizationModifier(
        targets="Linear", scheme="FP8_DYNAMIC", ignore=["lm_head"]
    )

    # Apply quantization
    oneshot(model=model, recipe=recipe)

    save_path = os.path.join(args.save_path, args.model_id.split("/")[1] + "-FP8-dynamic")
    os.makedirs(save_path, exist_ok=True)

    # Save to disk in compressed-tensors format
    model.save_pretrained(save_path)
    tokenizer.save_pretrained(save_path)
    print(f"Model and tokenizer saved to: {save_path}")

if __name__ == "__main__":
    main()

Evaluation

The optimized model was evaluated on GSM8k task with the flexible-extract filter score of 0.9030 ± 0.0082, and strict-match filter score of 0.8976 ± 0.0083, where as the unquantized model with the flexible-extract filter score of 0.9060 ± .0080, and strict-match filter score of 0.8992 ± 0.0083.

Evaluations were carried out using the following commands.

For the quantized model:

lm_eval \
  --model vllm \
  --model_args pretrained="nm-testing/Mistral-Small-24B-Instruct-2501-FP8-Dynamic",add_bos_token=True \
  --tasks gsm8k \
  --batch_size auto

For the unquantized model

lm_eval \
  --model vllm \
  --model_args pretrained="mistralai/Mistral-Small-24B-Instruct-2501",add_bos_token=True \
  --tasks gsm8k \
  --batch_size auto

Accuracy

GSM8k evaluation scores for the optimized model

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
gsm8k	3	flexible-extract	5	exact_match	↑	0.9030	±	0.0082
		strict-match	5	exact_match	↑	0.8976	±	0.0083

GSM8k evaluation scores for the unquantized model

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
gsm8k	3	flexible-extract	5	exact_match	↑	0.9060	±	0.0080
		strict-match	5	exact_match	↑	0.8992	±	0.0083