-
Notifications
You must be signed in to change notification settings - Fork 468
[Doc] Add Disaggregation Prefill for dpsk-v32 doc #3302
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
Signed-off-by: wangli <[email protected]>
👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:
If CI fails, you can run linting and testing checks locally according Contributing and Testing. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Code Review
This pull request adds a new documentation page for setting up prefill-decode disaggregation with the DeepSeek v3.2 model. The new tutorial provides step-by-step instructions for environment verification, rank table generation, and service deployment. I've identified a critical syntax error in the prefiller deployment script and a likely configuration omission in the decoder deployment script that would prevent users from successfully following the tutorial. The suggested changes correct these issues.
--trust-remote-code \ | ||
--no-enable-prefix-caching \ | ||
--gpu-memory-utilization 0.92 \ | ||
--additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The vllm serve
command is missing a line continuation character \
at the end of this line. This will cause a shell syntax error and prevent the command from running correctly.
--additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}' | |
--additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}' \ |
--enable-expert-parallel \ | ||
--trust-remote-code \ | ||
--no-enable-prefix-caching \ | ||
--gpu-memory-utilization 0.9 \ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The decoder node configuration is missing the --additional-config
to enable torchair graph mode, which is enabled on the prefiller node. For consistency and to ensure correct distributed operation, the decoder should also have graph mode enabled.
--gpu-memory-utilization 0.9 \ | |
--gpu-memory-utilization 0.9 \ | |
--additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}' \ |
Signed-off-by: wangli <[email protected]>
What this PR does / why we need it?
Add Disaggregation Prefill for dpsk-v32 doc
Does this PR introduce any user-facing change?
How was this patch tested?