Instructions to use z-lab/Qwen3.5-4B-DFlash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use z-lab/Qwen3.5-4B-DFlash with Transformers:

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

pipe = pipeline("text-generation", model="z-lab/Qwen3.5-4B-DFlash", trust_remote_code=True)

# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("z-lab/Qwen3.5-4B-DFlash", trust_remote_code=True)
model = AutoModel.from_pretrained("z-lab/Qwen3.5-4B-DFlash", trust_remote_code=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use z-lab/Qwen3.5-4B-DFlash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "z-lab/Qwen3.5-4B-DFlash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "z-lab/Qwen3.5-4B-DFlash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/z-lab/Qwen3.5-4B-DFlash

SGLang

How to use z-lab/Qwen3.5-4B-DFlash 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 "z-lab/Qwen3.5-4B-DFlash" \
    --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": "z-lab/Qwen3.5-4B-DFlash",
		"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 "z-lab/Qwen3.5-4B-DFlash" \
        --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": "z-lab/Qwen3.5-4B-DFlash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use z-lab/Qwen3.5-4B-DFlash with Docker Model Runner:
```
docker model run hf.co/z-lab/Qwen3.5-4B-DFlash
```

Mismatch between full attention layer mentioned in Dflash config for target context

by RaghavvGoel - opened Apr 22

Discussion

RaghavvGoel

Apr 22

Hi,

I was trying to run Dflash draft model with Qwen3.5-4B target model. However, I found that Dflash is using target hidden-states from linear attention layers. I was curious if this was the intended behavior?
have the authors ablated with using target hidden-states from full attention layers only?

Thanks!

jianchen0311

Z Lab org Apr 22

We didn’t distinguish linear attention layers or softmax attention layers, we just select the hidden states uniformly from target model layers. We did try to get target hidden all from softmax attention layers in Qwen3-Coder-Next-DFlash model, but the difference seems small.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment