Text Generation
Transformers
Safetensors
qwen3
feature-extraction
dflash
speculative-decoding
speculative-decoding-draft
block-diffusion
draft-model
diffusion-language-model
efficiency
qwen
qwen3.5
sglang
custom_code
text-generation-inference
Instructions to use lmsys/Qwen3.5-397B-A17B-DFlash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use lmsys/Qwen3.5-397B-A17B-DFlash with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="lmsys/Qwen3.5-397B-A17B-DFlash", trust_remote_code=True)# Load model directly from transformers import AutoTokenizer, AutoModel tokenizer = AutoTokenizer.from_pretrained("lmsys/Qwen3.5-397B-A17B-DFlash", trust_remote_code=True) model = AutoModel.from_pretrained("lmsys/Qwen3.5-397B-A17B-DFlash", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use lmsys/Qwen3.5-397B-A17B-DFlash with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "lmsys/Qwen3.5-397B-A17B-DFlash" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "lmsys/Qwen3.5-397B-A17B-DFlash", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/lmsys/Qwen3.5-397B-A17B-DFlash
- SGLang
How to use lmsys/Qwen3.5-397B-A17B-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 "lmsys/Qwen3.5-397B-A17B-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": "lmsys/Qwen3.5-397B-A17B-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 "lmsys/Qwen3.5-397B-A17B-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": "lmsys/Qwen3.5-397B-A17B-DFlash", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use lmsys/Qwen3.5-397B-A17B-DFlash with Docker Model Runner:
docker model run hf.co/lmsys/Qwen3.5-397B-A17B-DFlash
| from __future__ import annotations | |
| import importlib.util | |
| import os | |
| import subprocess | |
| import sys | |
| from dataclasses import dataclass | |
| from pathlib import Path | |
| PATCH_DIR = Path(os.environ.get("MODAL_PATCH_DIR", "/root/patches")) | |
| class PatchSpec: | |
| name: str | |
| module: str | |
| patch_file: str | |
| strip: int | |
| includes: tuple[str, ...] = () | |
| PATCHES = ( | |
| PatchSpec( | |
| name="flashinfer-pr-3312", | |
| module="flashinfer", | |
| patch_file="flashinfer-pr-3312.patch", | |
| strip=1, | |
| ), | |
| ) | |
| def _package_parent(module_name: str) -> Path: | |
| spec = importlib.util.find_spec(module_name) | |
| if spec is None or spec.submodule_search_locations is None: | |
| raise RuntimeError(f"Could not find installed package {module_name!r}.") | |
| locations = list(spec.submodule_search_locations) | |
| if not locations: | |
| raise RuntimeError(f"Installed package {module_name!r} has no package path.") | |
| return Path(locations[0]).resolve().parent | |
| def _git_apply_command(spec: PatchSpec, patch_path: Path) -> list[str]: | |
| cmd = ["git", "apply", f"-p{spec.strip}"] | |
| for include in spec.includes: | |
| cmd.append(f"--include={include}") | |
| cmd.append(str(patch_path)) | |
| return cmd | |
| def _check(cmd: list[str], *, cwd: Path) -> subprocess.CompletedProcess[str]: | |
| return subprocess.run( | |
| cmd, | |
| cwd=cwd, | |
| text=True, | |
| stdout=subprocess.PIPE, | |
| stderr=subprocess.STDOUT, | |
| ) | |
| def _apply_patch(spec: PatchSpec) -> None: | |
| patch_path = PATCH_DIR / spec.patch_file | |
| if not patch_path.exists(): | |
| raise RuntimeError(f"Missing patch file: {patch_path}") | |
| cwd = _package_parent(spec.module) | |
| base_cmd = _git_apply_command(spec, patch_path) | |
| reverse_cmd = [*base_cmd[:2], "--reverse", "--check", *base_cmd[2:]] | |
| check_cmd = [*base_cmd[:2], "--check", *base_cmd[2:]] | |
| reverse = _check(reverse_cmd, cwd=cwd) | |
| if reverse.returncode == 0: | |
| print(f"[patch] {spec.name} already applied under {cwd}") | |
| return | |
| check = _check(check_cmd, cwd=cwd) | |
| if check.returncode != 0: | |
| print(check.stdout, file=sys.stderr) | |
| raise RuntimeError(f"Patch {spec.name} does not apply under {cwd}.") | |
| print(f"[patch] applying {spec.name} under {cwd}") | |
| subprocess.run(base_cmd, cwd=cwd, check=True) | |
| def main() -> None: | |
| for patch in PATCHES: | |
| _apply_patch(patch) | |
| if __name__ == "__main__": | |
| main() | |