Instructions to use AICP-Labs/qwen3-32b-dflash-en-zh with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AICP-Labs/qwen3-32b-dflash-en-zh with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="AICP-Labs/qwen3-32b-dflash-en-zh")

# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("AICP-Labs/qwen3-32b-dflash-en-zh")
model = AutoModel.from_pretrained("AICP-Labs/qwen3-32b-dflash-en-zh")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use AICP-Labs/qwen3-32b-dflash-en-zh with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AICP-Labs/qwen3-32b-dflash-en-zh"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AICP-Labs/qwen3-32b-dflash-en-zh",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/AICP-Labs/qwen3-32b-dflash-en-zh

SGLang

How to use AICP-Labs/qwen3-32b-dflash-en-zh with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "AICP-Labs/qwen3-32b-dflash-en-zh" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AICP-Labs/qwen3-32b-dflash-en-zh",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "AICP-Labs/qwen3-32b-dflash-en-zh" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "AICP-Labs/qwen3-32b-dflash-en-zh",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use AICP-Labs/qwen3-32b-dflash-en-zh with Docker Model Runner:
```
docker model run hf.co/AICP-Labs/qwen3-32b-dflash-en-zh
```

Qwen3-32B DFlash Draft Model (EagleChat 400K Mix)

A DFlash draft model trained from Qwen3-32B using a EagleChat subset (English 200K + Chinese 200K) to accelerate speculative decoding.

Model Summary

This repository provides a DFlash draft model for Qwen3-32B. The draft model is intended to be used together with the target model in SpecForge, improving throughput (output tokens/sec) under standard speculative verification.

Base / Target model: Qwen/Qwen3-32B
Draft model type: DFlash (speculative decoding draft)
Training data: EagleChat subset (English 200K + Chinese 200K; total ~400K)
Training hardware: H100
Primary use case: accelerate inference with DFlash / SpecForge

Training Details

Data

Dataset: EagleChat subset
Composition:
- English: ~200,000 samples
- Chinese: 200,000 samples
Total: ~400,000 samples

Procedure

Epochs: 6
Sequence length: 4096
Precision: bf16
Codebase: SpecForge (DFlash training)

Evaluation

Benchmark settings

Target model: /models/Qwen3-32B
Draft model: sx-aicp/qwen3-32b-dflash-en-zh (or local path)
Max new tokens: 2048
Attention backend: fa3
Tensor parallel (tp_size): 4
device_sm: 90 (H100)
drop_first_batch: true
Concurrencies: 1 / 4 / 32 (varies by suite)

Speed Bench Results

Environment: H100 (SM90), tp=4, attention=fa3, max_new_tokens=2048, drop_first_batch=true.

Unified Summary

Benchmark	Conc=1	Conc=4	Conc=32
Math500	109.20 → 392.63 3.595× / L=5.564	409.44 → 1351.51 3.301× / L=5.582	2554.68 → 4554.81 1.783× / L=5.588
HumanEval	108.93 → 331.66 3.045× / L=4.769	407.34 → 1129.16 2.772× / L=4.756	2482.40 → 3632.36 1.463× / L=4.757
MT-Bench	109.19 → 233.75 2.141× / L=3.791	409.97 → 804.64 1.963× / L=3.852	2470.75 → 2767.16 1.120× / L=3.917

Format: baseline tok/s → DFlash tok/s; Speedup× / L(acceptance length).

How to Evaluate (z-lab / dflash)

python benchmark_sglang.py \
  --tp-size 4 \
  --target-model /models/Qwen3-32B \
  --draft-model /path/to/draft_model \
  --concurrencies 1,4,32 \
  --dataset-name math500 \
  --attention-backends fa3 \
  --output-md sglang_results.md

Downloads last month: 17

Safetensors

Model size

0.5B params

Tensor type

BF16

Model tree for AICP-Labs/qwen3-32b-dflash-en-zh

Base model

Qwen/Qwen3-32B

Finetuned

(505)

this model

Collection including AICP-Labs/qwen3-32b-dflash-en-zh

DFLASH

Collection

1 item • Updated Mar 2 • 1