[Serve] add allocator in Storage as the upstream change #1996
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This PR introduces the IPC memory and customized all-reduce kernel dispatches for tensor parallelism. We add a new compiler flag `--allreduce-strategy`, which supports `"ring"`, `"one-shot"` and `"two-shot"`. The flag defaults to `"ring"`, which means this PR makes no difference if people do not manually change the all-reduce strategy. As of now the IPC-memory-backed customized all-reduce kernels are only available on CUDA. To enable all-reduce strategies other than "ring", here are some example compile commands: ```python python -m mlc_llm compile model/mlc-chat-config.json --device cuda --opt "allreduce-strategy=one-shot" -o model/lib.so python -m mlc_llm compile model/mlc-chat-config.json --device cuda --opt "allreduce-strategy=two-shot" -o model/lib.so ``` Please be aware that, you probably also need to specify other compiler flags, for example, like `--opt "cublas_gemm=1;allreduce-strategy=one-shot"`.
Contributor
Author
|
failed to reopen it, will create a new one |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The changes in apache/tvm#16750 modified the signature of the Storage, this pull request updates the caller code in mlc-llm to accommodate the new Storage class signature. Ran into build error w/o the change.
cc: @MasterJH5574 @vinx13 @tqchen