Graph Mode Guide

Graph Mode Guide#

Note

This feature is currently experimental. In future versions, there may be behavioral changes around configuration, coverage, performance improvement.

This guide provides instructions for using Ascend Graph Mode with vLLM Ascend. Please note that graph mode is only available on V1 Engine.

Getting Started#

In v0.11.0, vLLM Ascend will run models in graph mode by default to keep the same behavior with vLLM. If you hit any issues, please feel free to open an issue on GitHub and fallback to the eager mode temporarily by set enforce_eager=True when initializing the model.

There are two kinds for graph mode supported by vLLM Ascend:

  • ACLGraph: This is the default graph mode supported by vLLM Ascend.

  • TorchAirGraph: This is the GE graph mode. In v0.11.0, only DeepSeek series models are supported.

Note

ACLGraph improves the performance of all the models by default, TorchAirGraph will be dropped in next version, and v0.11.0 will be the last stable version that support TorchAirGraph.

Using ACLGraph#

ACLGraph is enabled by default. Take Qwen series models as an example, just set to use V1 Engine is enough.

Offline example:

import os

from vllm import LLM

model = LLM(model="Qwen/Qwen2-7B-Instruct")
outputs = model.generate("Hello, how are you?")

Online example:

vllm serve Qwen/Qwen2-7B-Instruct

Using TorchAirGraph#

If you want to run DeepSeek series models with the graph mode, you should use TorchAirGraph. In this case, additional configuration is required.

Offline example:

import os
from vllm import LLM

# TorchAirGraph is only work without chunked-prefill now
model = LLM(model="deepseek-ai/DeepSeek-R1-0528", additional_config={"torchair_graph_config": {"enabled": True},"ascend_scheduler_config": {"enabled": True,}})
outputs = model.generate("Hello, how are you?")

Online example:

vllm serve Qwen/Qwen2-7B-Instruct --additional-config='{"torchair_graph_config": {"enabled": true},"ascend_scheduler_config": {"enabled": true,}}'

You can find more details about additional configuration here.

Fallback to the Eager Mode#

If both ACLGraph and TorchAirGraph fail to run, you should fallback to the eager mode.

Offline example:

import os
from vllm import LLM

model = LLM(model="someother_model_weight", enforce_eager=True)
outputs = model.generate("Hello, how are you?")

Online example:

vllm serve Qwen/Qwen2-7B-Instruct --enforce-eager
Contents