CoolFace
Modelpublic

RedHatAI/Qwen3-32B-FP8-dynamic

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
15likes1.5kdownloads
Model Card

Qwen3-32B-FP8-dynamic

Model Overview

  • —Model Architecture: Qwen3ForCausalLM
  • —Input: Text
  • —Output: Text
  • —Model Optimizations:
  • —Activation quantization: FP8
  • —Weight quantization: FP8
  • —Intended Use Cases:
  • —Reasoning.
  • —Function calling.
  • —Subject matter experts via fine-tuning.
  • —Multilingual instruction following.
  • —Translation.
  • —Out-of-scope: Use in any manner that violates applicable laws or regulations (including trade compliance laws).
  • —Release Date: 05/02/2025
  • —Version: 1.0
  • —Model Developers: RedHat (Neural Magic)

Model Optimizations

This model was obtained by quantizing activations and weights of Qwen3-32B to FP8 data type. This optimization reduces the number of bits used to represent weights and activations from 16 to 8, reducing GPU memory requirements (by approximately 50%) and increasing matrix-multiply compute throughput (by approximately 2x). Weight quantization also reduces disk size requirements by approximately 50%.

Only weights and activations of the linear operators within transformers blocks are quantized. Weights are quantized with a symmetric static per-channel scheme, whereas activations are quantized with a symmetric dynamic per-token scheme. The llm-compressor library is used for quantization.

Deployment

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

python
from vllm import LLM, SamplingParams
from transformers import AutoTokenizer

model_id = "RedHatAI/Qwen3-32B-FP8-dynamic"
number_gpus = 1
sampling_params = SamplingParams(temperature=0.6, top_p=0.95, top_k=20, min_p=0, max_tokens=256)

messages = [
    {"role": "user", "content": prompt}
]

tokenizer = AutoTokenizer.from_pretrained(model_id)

messages = [{"role": "user", "content": "Give me a short introduction to large language model."}]

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 for more details.

Creation

<details> <summary>Creation details</summary> This model was created with llm-compressor by running the code snippet below.

python
  from llmcompressor.modifiers.quantization import QuantizationModifier
  from llmcompressor.transformers import oneshot
  from transformers import AutoModelForCausalLM, AutoTokenizer
  
  # Load model
  model_stub = "Qwen/Qwen3-32B"
  model_name = model_stub.split("/")[-1]

  model = AutoModelForCausalLM.from_pretrained(model_stub)

  tokenizer = AutoTokenizer.from_pretrained(model_stub)

  # Configure the quantization algorithm and scheme
  recipe = QuantizationModifier(
      ignore=["lm_head"],
      targets="Linear",
      scheme="FP8_dynamic",
  )

  # Apply quantization
  oneshot(
      model=model,
      recipe=recipe,
  )
  
  # Save to disk in compressed-tensors format
  save_path = model_name + "-FP8-dynamic"
  model.save_pretrained(save_path)
  tokenizer.save_pretrained(save_path)
  print(f"Model and tokenizer saved to: {save_path}")

</details>

Evaluation

The model was evaluated on the OpenLLM leaderboard tasks (versions 1 and 2), using lm-evaluation-harness, and on reasoning tasks using lighteval. vLLM was used for all evaluations.

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

lm-evaluation-harness

  lm_eval \
    --model vllm \
    --model_args pretrained="RedHatAI/Qwen3-32B-FP8-dynamic",dtype=auto,gpu_memory_utilization=0.5,max_model_len=8192,enable_chunk_prefill=True,tensor_parallel_size=2 \
    --tasks openllm \
    --apply_chat_template\
    --fewshot_as_multiturn \
    --batch_size auto
  lm_eval \
    --model vllm \
    --model_args pretrained="RedHatAI/Qwen3-32B-FP8-dynamic",dtype=auto,gpu_memory_utilization=0.5,max_model_len=8192,enable_chunk_prefill=True,tensor_parallel_size=2 \
    --tasks mgsm \
    --apply_chat_template\
    --batch_size auto
  lm_eval \
    --model vllm \
    --model_args pretrained="RedHatAI/Qwen3-32B-FP8-dynamic",dtype=auto,gpu_memory_utilization=0.5,max_model_len=16384,enable_chunk_prefill=True,tensor_parallel_size=2 \
    --tasks leaderboard \
    --apply_chat_template\
    --fewshot_as_multiturn \
    --batch_size auto

lighteval

lightevalmodelarguments.yaml

yaml
  model_parameters:
    model_name: RedHatAI/Qwen3-32B-FP8-dynamic
    dtype: auto
    gpu_memory_utilization: 0.9
    tensor_parallel_size: 2
    max_model_length: 40960
    generation_parameters:
      temperature: 0.6
      top_k: 20
      min_p: 0.0
      top_p: 0.95
      max_new_tokens: 32768
  lighteval vllm \
    --model_args lighteval_model_arguments.yaml \
    --tasks lighteval|aime24|0|0 \
    --use_chat_template = true
  lighteval vllm \
    --model_args lighteval_model_arguments.yaml \
    --tasks lighteval|aime25|0|0 \
    --use_chat_template = true
  lighteval vllm \
    --model_args lighteval_model_arguments.yaml \
    --tasks lighteval|math_500|0|0 \
    --use_chat_template = true
  lighteval vllm \
    --model_args lighteval_model_arguments.yaml \
    --tasks lighteval|gpqa:diamond|0|0 \
    --use_chat_template = true
  lighteval vllm \
    --model_args lighteval_model_arguments.yaml \
    --tasks extended|lcb:codegeneration \
    --use_chat_template = true

</details>

Accuracy

<table> <tr> <th>Category </th> <th>Benchmark </th> <th>Qwen3-32B </th> <th>Qwen3-32B-FP8-dynamic<br>(this model) </th> <th>Recovery </th> </tr> <tr> <td rowspan="7" ><strong>OpenLLM v1</strong> </td> <td>MMLU (5-shot) </td> <td>80.96 </td> <td>80.89 </td> <td>99.9% </td> </tr> <tr> <td>ARC Challenge (25-shot) </td> <td>69.03 </td> <td>68.00 </td> <td>98.5% </td> </tr> <tr> <td>GSM-8K (5-shot, strict-match) </td> <td>87.64 </td> <td>88.32 </td> <td>100.8% </td> </tr> <tr> <td>Hellaswag (10-shot) </td> <td>71.10 </td> <td>71.44 </td> <td>100.5% </td> </tr> <tr> <td>Winogrande (5-shot) </td> <td>69.77 </td> <td>69.85 </td> <td>100.1% </td> </tr> <tr> <td>TruthfulQA (0-shot, mc2) </td> <td>58.63 </td> <td>59.13 </td> <td>100.9% </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>72.86</strong> </td> <td><strong>72.94</strong> </td> <td><strong>100.1%</strong> </td> </tr> <tr> <td rowspan="7" ><strong>OpenLLM v2</strong> </td> <td>MMLU-Pro (5-shot) </td> <td>54.24 </td> <td>54.78 </td> <td>101.0% </td> </tr> <tr> <td>IFEval (0-shot) </td> <td>86.23 </td> <td>86.23 </td> <td>100.0% </td> </tr> <tr> <td>BBH (3-shot) </td> <td>44.29 </td> <td>43.70 </td> <td>98.7% </td> </tr> <tr> <td>Math-lvl-5 (4-shot) </td> <td>54.61 </td> <td>57.26 </td> <td>104.9% </td> </tr> <tr> <td>GPQA (0-shot) </td> <td>5.53 </td> <td>5.46 </td> <td>--- </td> </tr> <tr> <td>MuSR (0-shot) </td> <td>7.85 </td> <td>8.81 </td> <td>--- </td> </tr> <tr> <td><strong>Average</strong> </td> <td><strong>42.13</strong> </td> <td><strong>42.71</strong> </td> <td><strong>101.4%</strong> </td> </tr> <tr> <td><strong>Multilingual</strong> </td> <td>MGSM (0-shot) </td> <td>32.57 </td> <td> </td> <td> </td> </tr> <tr> <td rowspan="6" ><strong>Reasoning<br>(generation)</strong> </td> <td>AIME 2024 </td> <td>79.37 </td> <td>79.37 </td> <td>100.0% </td> </tr> <tr> <td>AIME 2025 </td> <td>71.77 </td> <td>70.42 </td> <td>98.1% </td> </tr> <tr> <td>GPQA diamond </td> <td>66.67 </td> <td>68.69 </td> <td>103.0% </td> </tr> <tr> <td>Math-lvl-5 </td> <td>96.20 </td> <td>96.40 </td> <td>100.2% </td> </tr> <tr> <td>LiveCodeBench </td> <td>62.45 </td> <td>63.32 </td> <td>101.4% </td> </tr> </table>