Instructions to use z-lab/gpt-oss-20b-DFlash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use z-lab/gpt-oss-20b-DFlash with Transformers:

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

pipe = pipeline("text-generation", model="z-lab/gpt-oss-20b-DFlash", trust_remote_code=True)

# Load model directly
from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("z-lab/gpt-oss-20b-DFlash", trust_remote_code=True)
model = AutoModel.from_pretrained("z-lab/gpt-oss-20b-DFlash", trust_remote_code=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use z-lab/gpt-oss-20b-DFlash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "z-lab/gpt-oss-20b-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/gpt-oss-20b-DFlash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/z-lab/gpt-oss-20b-DFlash

SGLang

How to use z-lab/gpt-oss-20b-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/gpt-oss-20b-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/gpt-oss-20b-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/gpt-oss-20b-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/gpt-oss-20b-DFlash",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use z-lab/gpt-oss-20b-DFlash with Docker Model Runner:
```
docker model run hf.co/z-lab/gpt-oss-20b-DFlash
```

Question about DFlash design choices: 8 layers and block size 8

by K-Compression - opened 2 days ago

Discussion

K-Compression

2 days ago

I noticed that most DFlash models use 5 layers and a block size of 16, while the GPT-OSS DFlash model uses 8 layers and a block size of 8. Was there a model-specific reason for this choice? In particular, did the relatively smaller hidden size or intermediate size of GPT-OSS make larger block sizes less effective, leading to the use of more DFlash layers and smaller blocks instead? I'd also be curious to know whether you evaluated alternative configurations and what trade-offs (e.g., quality, acceptance rate, or efficiency) influenced the final design

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