Llama-3.1-8B-Instruct-FP8-block

Llama-3.1-8B-Instruct-FP8-block

Model Overview

• Model Architecture: LlamaForCausalLM
  • Input: Text
  • Output: Text
• Model Optimizations:
  • Weight quantization: FP8
  • Activation quantization: FP8
• Release Date:
• Version: 1.0
• Model Developers:: Red Hat

Quantized version of meta-llama/Llama-3.1-8B-Instruct.

Model Optimizations

This model was obtained by quantizing the weights and activations of meta-llama/Llama-3.1-8B-Instruct to FP8 data type. 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 of the language model are quantized.

Deployment

Use with vLLM

1. Initialize vLLM server:

vllm serve RedHatAI/Llama-3.1-8B-Instruct-FP8-block --tensor_parallel_size 1

1. Send requests to the server:

from openai import OpenAI

# Modify OpenAI's API key and API base to use vLLM's API server.
openai_api_key = "EMPTY"
openai_api_base = "http://<your-server-host>:8000/v1"

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

model = "RedHatAI/Llama-3.1-8B-Instruct-FP8-block"


messages = [
    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
]

outputs = client.chat.completions.create(
    model=model,
    messages=messages,
)

generated_text = outputs.choices[0].message.content
print(generated_text)

Creation

This model was quantized using the llm-compressor library as shown below.

<details> <summary>Creation details</summary>

from transformers import AutoProcessor, LlamaForCausalLM

from llmcompressor import oneshot
from llmcompressor.modeling import replace_modules_for_calibration
from llmcompressor.modifiers.quantization import QuantizationModifier

MODEL_ID = "meta-llama/Llama-3.1-8B-Instruct"

# Load model.
model = LlamaForCausalLM.from_pretrained(MODEL_ID, dtype="auto")
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = replace_modules_for_calibration(model)

# Configure the quantization algorithm and scheme.
# In this case, we:
#   * quantize the weights to fp8 with per-block quantization
#   * quantize the activations to fp8 with dynamic token activations
recipe = QuantizationModifier(
    targets="Linear",
    scheme="FP8_BLOCK",
    ignore=["lm_head"],
)

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

# Save to disk in compressed-tensors format.
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-block"
model.save_pretrained(SAVE_DIR)
processor.save_pretrained(SAVE_DIR)

</details>

Evaluation

The model was evaluated on the OpenLLMv1 leaderboard task, using lm-evaluation-harness, on reasoning tasks using lighteval. vLLM was used for all evaluations.

<details> <summary>Evaluation details</summary>

Openllm V1

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Llama-3.1-8B-Instruct-FP8-block",dtype=auto,add_bos_token=True,max_model_len=16384,tensor_parallel_size=1,gpu_memory_utilization=0.9,enable_chunked_prefill=True,trust_remote_code=True \
  --tasks openllm \
  --write_out \
  --batch_size auto \
  --show_config

Openllm V2

lm_eval \
  --model vllm \
  --model_args pretrained="RedHatAI/Llama-3.1-8B-Instruct-FP8-block",dtype=auto,add_bos_token=False,max_model_len=16384,tensor_parallel_size=1,gpu_memory_utilization=0.7,disable_log_stats=True,enable_chunked_prefill=True,trust_remote_code=True \
  --tasks leaderboard \
  --apply_chat_template \
  --fewshot_as_multiturn \
  --write_out \
  --batch_size auto \
  --show_config

Coding Benchmarks

evalplus.evaluate --model "RedHatAI/Llama-3.1-8B-Instruct-FP8-block" \
                  --dataset "humaneval" \
                  --backend vllm \
                  --tp 1 \
                  --greedy
evalplus.evaluate --model "RedHatAI/Llama-3.1-8B-Instruct-FP8-block" \
                --dataset "mbpp" \
                --backend vllm \
                --tp 1 \
                --greedy

</details>

Accuracy

<table> <thead> <tr> <th>Category</th> <th>Metric</th> <th>meta-llama/Llama-3.1-8B-Instruct</th> <th>RedHatAI/Llama-3.1-8B-Instruct-FP8-block</th> <th>Recovery (%)</th> </tr> </thead> <tbody> <!-- OpenLLM Leaderboard V1 --> <tr> <td rowspan="7"><b>OpenLLM V1</b></td> <td>ARC-Challenge (Acc-Norm, 25-shot)</td> <td>60.92</td> <td>60.92</td> <td>100.00</td> </tr> <tr> <td>GSM8K (Strict-Match, 5-shot)</td> <td>71.11</td> <td>70.66</td> <td>99.36</td> </tr> <tr> <td>HellaSwag (Acc-Norm, 10-shot)</td> <td>80.75</td> <td>80.48</td> <td>99.67</td> </tr> <tr> <td>MMLU (Acc, 5-shot)</td> <td>68.20</td> <td>67.96</td> <td>99.64</td> </tr> <tr> <td>TruthfulQA (MC2, 0-shot)</td> <td>54.54</td> <td>54.18</td> <td>99.34</td> </tr> <tr> <td>Winogrande (Acc, 5-shot)</td> <td>78.45</td> <td>78.14</td> <td>99.60</td> </tr> <tr> <td><b>Average Score</b></td> <td><b>69.00</b></td> <td><b>68.72</b></td> <td><b>99.59</b></td> </tr> <!-- OpenLLM Leaderboard V2 --> <tr> <td rowspan="7"><b>OpenLLM V2</b></td> <td>IFEval (Inst Level Strict Acc, 0-shot)</td> <td>81.89</td> <td>81.41</td> <td>99.41</td> </tr> <tr> <td>BBH (Acc-Norm, 3-shot)</td> <td>50.70</td> <td>50.96</td> <td>100.51</td> </tr> <tr> <td>Math-Hard (Exact-Match, 4-shot)</td> <td>20.24</td> <td>20.77</td> <td>102.61</td> </tr> <tr> <td>GPQA (Acc-Norm, 0-shot)</td> <td>29.53</td> <td>29.95</td> <td>101.42</td> </tr> <tr> <td>MUSR (Acc-Norm, 0-shot)</td> <td>38.89</td> <td>38.62</td> <td>99.32</td> </tr> <tr> <td>MMLU-Pro (Acc, 5-shot)</td> <td>37.71</td> <td>37.48</td> <td>99.38</td> </tr> <tr> <td><b>Average Score</b></td> <td><b>43.16</b></td> <td><b>43.20</b></td> <td><b>100.09</b></td> </tr> <tr> <td rowspan="4" ><strong>Coding</strong> </td> <td>HumanEval pass@1 </td> <td>68.90 </td> <td>68.90 </td> <td>100.00 </td> </tr> <tr> <td>HumanEval+ pass@1 </td> <td>62.20 </td> <td>61.00 </td> <td>98.07 </td> </tr> <tr> <td>MBPP pass@1 </td> <td>67.70 </td> <td>71.40 </td> <td>105.47 </td> </tr> <tr> <td>MBPP+ pass@1 </td> <td>55.60 </td> <td>57.90 </td> <td>104.14 </td> </tr> </tbody> </table>