Llama-3.1-8B-Instruct-NVFP4/README.md

---
tags:
- fp4
- vllm
language:
- en
- de
- fr
- it
- pt
- hi
- es
- th
pipeline_tag: text-generation
license: llama3.1
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
---

# Meta-Llama-3.1-8B-Instruct-NVFP4

## Model Overview
- **Model Architecture:** Meta-Llama-3.1
  - **Input:** Text
  - **Output:** Text
- **Model Optimizations:**
  - **Weight quantization:** FP4
  - **Activation quantization:** FP4
- **Intended Use Cases:** Intended for commercial and research use in multiple languages. Similarly to [Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct), this models is intended for assistant-like chat.
- **Out-of-scope:** Use in any manner that violates applicable laws or regulations (including trade compliance laws). Use in languages other than English.
- **Release Date:** 10/23/2025
- **Version:** 1.0
- **License(s):** [llama3.1](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B/blob/main/LICENSE)
- **Model Developers:** RedHatAI

This model is a quantized version of [Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct).
It was evaluated on a several tasks to assess the its quality in comparison to the unquatized model.

### Model Optimizations

This model was obtained by quantizing the weights and activations of [Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) to FP4 data type, ready for inference with vLLM>=0.9.1
This optimization reduces the number of bits per parameter from 16 to 4, reducing the disk size and GPU memory requirements by approximately 75%.

Only the weights and activations of the linear operators within transformers blocks are quantized using [LLM Compressor](https://github.com/vllm-project/llm-compressor).

## Deployment

### Use with vLLM

This model can be deployed efficiently using the [vLLM](https://docs.vllm.ai/en/latest/) backend, as shown in the example below.

```python
from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4"
number_gpus = 1

sampling_params = SamplingParams(temperature=0.6, top_p=0.9, max_tokens=256)

tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

prompts = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)

llm = LLM(model=model_id, tensor_parallel_size=number_gpus)

outputs = llm.generate(prompts, sampling_params)

generated_text = outputs[0].outputs[0].text
print(generated_text)
```

vLLM aslo supports OpenAI-compatible serving. See the [documentation](https://docs.vllm.ai/en/latest/) for more details.

## Creation

This model was created by applying [LLM Compressor with calibration samples from UltraChat](https://github.com/vllm-project/llm-compressor/blob/main/examples/quantization_w4a4_fp4/llama3_example.py), as presented in the code snipet below.

<details>

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

from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier
from llmcompressor.modifiers.smoothquant import SmoothQuantModifier
from llmcompressor.utils import dispatch_for_generation

MODEL_ID = "meta-llama/Meta-Llama-3-8B-Instruct"

# Load model.
model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

DATASET_ID = "HuggingFaceH4/ultrachat_200k"
DATASET_SPLIT = "train_sft"

# Select number of samples. 512 samples is a good place to start.
# Increasing the number of samples can improve accuracy.
NUM_CALIBRATION_SAMPLES = 512
MAX_SEQUENCE_LENGTH = 2048

# Load dataset and preprocess.
ds = load_dataset(DATASET_ID, split=f"{DATASET_SPLIT}[:{NUM_CALIBRATION_SAMPLES}]")
ds = ds.shuffle(seed=42)

def preprocess(example):
    return {
        "text": tokenizer.apply_chat_template(
            example["messages"],
            tokenize=False,
        )
    }

ds = ds.map(preprocess)

# Tokenize inputs.
def tokenize(sample):
    return tokenizer(
        sample["text"],
        padding=False,
        max_length=MAX_SEQUENCE_LENGTH,
        truncation=True,
        add_special_tokens=False,
    )

ds = ds.map(tokenize, remove_columns=ds.column_names)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp4 with per group 16 via ptq
#   * calibrate a global_scale for activations, which will be used to
#       quantize activations to fp4 on the fly
smoothing_strength = 0.5
recipe = [
    SmoothQuantModifier(smoothing_strength=smoothing_strength),
    QuantizationModifier(
        ignore=["re:.*lm_head.*"],
        config_groups={
            "group_0": {
                "targets": ["Linear"],
                "weights": {
                    "num_bits": 4,
                    "type": "float",
                    "strategy": "tensor_group",
                    "group_size": 16,
                    "symmetric": True,
                    "observer": "mse",
                },
                "input_activations": {
                    "num_bits": 4,
                    "type": "float",
                    "strategy": "tensor_group",
                    "group_size": 16,
                    "symmetric": True,
                    "dynamic": "local",
                    "observer": "minmax",
                },
            }
        },
    )
]

# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-NVFP4"

# Apply quantization.
oneshot(
    model=model,
    dataset=ds,
    recipe=recipe,
    max_seq_length=MAX_SEQUENCE_LENGTH,
    num_calibration_samples=NUM_CALIBRATION_SAMPLES,
    output_dir=SAVE_DIR,
)

print("\n\n")
print("========== SAMPLE GENERATION ==============")
dispatch_for_generation(model)
input_ids = tokenizer("Hello my name is", return_tensors="pt").input_ids.to("cuda")
output = model.generate(input_ids, max_new_tokens=100)
print(tokenizer.decode(output[0]))
print("==========================================\n\n")

model.save_pretrained(SAVE_DIR, save_compressed=True)
tokenizer.save_pretrained(SAVE_DIR)

```
</details>

## Evaluation

This model was evaluated on the well-known OpenLLM v1, OpenLLM v2 and HumanEval_64 benchmarks. All evaluations were conducted using [lm-evaluation-harness](https://github.com/neuralmagic/lm-evaluation-harness).

### Accuracy
<table>
<thead>
  <tr>
    <th>Category</th>
    <th>Metric</th>
    <th>Meta-Llama-3.1-8B-Instruct</th>
    <th>Llama-3.1-8B-Instruct-NVFP4 (this model)</th>
    <th>Recovery</th>
  </tr>
</thead>
<tbody>
<tr>
  <td rowspan="8"><b>OpenLLM V1</b></td>
  <td>arc_challenge_llama</td><td>83.35</td><td>82.32</td><td>98.76</td></tr>
<tr>
  <td>gsm8k_llama</td><td>78.17</td><td>79.30</td><td>101.45</td></tr>
<tr><td>hellaswag</td><td>78.43</td><td>78.01</td><td>99.46</td></tr>
<tr><td>mmlu_llama</td><td>69.37</td><td>65.95</td><td>95.07</td></tr>
<tr><td>mmlu_cot_llama</td><td>72.86</td><td>68.60</td><td>94.15</td></tr>
<tr><td>truthfulqa_mc2</td><td>55.09</td><td>52.95</td><td>96.12</td></tr>
<tr><td>winogrande</td><td>75.77</td><td>74.03</td><td>97.70</td></tr>
<tr><td><b>Average</b></td><td><b>73.29</b></td><td><b>71.59</b></td><td><b>97.68</b></td></tr>
</tbody></table>

<table>
  <thead>
    <tr>
      <th>Category</th>
      <th>Metric</th>
      <th>Meta-Llama-3.1-8B-Instruct</th>
      <th>RedHatAI/Llama-3.1-8B-Instruct-NVFP4 (this model)</th>
      <th>Recovery (%)</th>
    </tr>
  </thead>
  <tbody>
  <tr>
    <td rowspan="7"><b>OpenLLM V2</b></td>
    <td>MMLU-Pro (5-shot)</td>
    <td>37.69</td>
    <td>34.43</td>
    <td>91.35</td>
  </tr>
  <tr>
    <td>IFEval (0-shot)</td>
    <td>80.94</td>
    <td>79.98</td>
    <td>98.81</td>
  </tr>
  <tr>
    <td>BBH (3-shot)</td>
    <td>50.76</td>
    <td>48.62</td>
    <td>95.78</td>
  </tr>
  <tr>
    <td>Math-|v|-5 (4-shot)</td>
    <td>22.05</td>
    <td>14.65</td>
    <td>66.44</td>
  </tr>
  <tr>
    <td>GPQA (0-shot)</td>
    <td>28.44</td>
    <td>27.94</td>
    <td>98.24</td>
  </tr>
  <tr>
    <td>MuSR (0-shot)</td>
    <td>38.10</td>
    <td>37.83</td>
    <td>99.29</td>
  </tr>
  <tr>
    <td><b>Average</b></td>
    <td><b>43.00</b></td>
    <td><b>40.58</b></td>
    <td><b>94.37</b></td>
  </tr>
  <tr>
    <td><b>Coding</b></td>
    <td>HumanEval_64 pass@2</td>
    <td>71.90</td>
    <td>71.44</td>
    <td>99.36</td>
  </tr>
 </tbody>
</table>



### Reproduction

The results were obtained using the following commands:

<details>

#### MMLU_LLAMA
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks mmlu_llama \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --batch_size auto
```

#### MMLU_COT_LLAMA
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks mmlu_cot_llama \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --batch_size auto
```

#### ARC-Challenge
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks arc_challenge_llama \
  --apply_chat_template \
  --batch_size auto
```

#### GSM-8K
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks gsm8k_llama \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --batch_size auto
```

#### Hellaswag
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks hellaswag \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --batch_size auto
```

#### Winogrande
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks winogrande \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --batch_size auto
```

#### TruthfulQA
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,add_bos_token=True,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True \
  --tasks truthfulqa \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --batch_size auto
```

#### OpenLLM v2
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True\
  --apply_chat_template \
  --fewshot_as_multiturn \
  --tasks leaderboard \
  --batch_size auto
```

#### HumanEval and HumanEval_64
```
lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Meta-Llama-3.1-8B-Instruct-NVFP4",dtype=auto,max_model_len=4096,tensor_parallel_size=1,enable_chunked_prefill=True,enforce_eager=True\
  --apply_chat_template \
  --fewshot_as_multiturn \
  --tasks humaneval_64_instruct \
  --batch_size auto
```

</details>