[Bugfix] Adjust mllama to regional compilation #15112
Merged
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
|
👋 Hi! Thank you for contributing to the vLLM project. 💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging. To run CI, PR reviewers can either: Add 🚀 |
Signed-off-by: Jan Kaniecki <[email protected]>
Signed-off-by: Jan Kaniecki <[email protected]>
jkaniecki
force-pushed
the
mllama_regional_compilation
branch
from
2f5b299 to
49cd230
Compare
heheda12345
approved these changes
Mar 19, 2025
DarkLight1337
approved these changes
Mar 19, 2025
github-actions
bot
added
the
ready
gmarinho2
pushed a commit
to gmarinho2/vllm
that referenced
this pull request
Apr 1, 2025
Signed-off-by: Jan Kaniecki <[email protected]>
jkaniecki
added a commit
to HabanaAI/vllm-fork
that referenced
this pull request
Apr 2, 2025
Signed-off-by: Jan Kaniecki <[email protected]>
afierka-intel
pushed a commit
to HabanaAI/vllm-fork
that referenced
this pull request
Apr 3, 2025
This PR involves cherry-pick of vllm-project#15112 from the upstream and a fix for cos_sin preparation in emb layers to match regional compilation. --------- Signed-off-by: Jan Kaniecki <[email protected]>
lulmer
pushed a commit
to lulmer/vllm
that referenced
this pull request
Apr 7, 2025
Signed-off-by: Jan Kaniecki <[email protected]> Signed-off-by: Louis Ulmer <[email protected]>
imangohari1
pushed a commit
to imangohari1/vllm-fork
that referenced
this pull request
Apr 8, 2025
This PR involves cherry-pick of vllm-project#15112 from the upstream and a fix for cos_sin preparation in emb layers to match regional compilation. --------- Signed-off-by: Jan Kaniecki <[email protected]>
tvoas
pushed a commit
to tvoas/vllm-fork
that referenced
this pull request
Apr 9, 2025
This PR involves cherry-pick of vllm-project#15112 from the upstream and a fix for cos_sin preparation in emb layers to match regional compilation. --------- Signed-off-by: Jan Kaniecki <[email protected]>
nishith-fujitsu
pushed a commit
to nishith-fujitsu/vllm
that referenced
this pull request
Apr 9, 2025
Signed-off-by: Jan Kaniecki <[email protected]>
shreyankg
pushed a commit
to shreyankg/vllm
that referenced
this pull request
May 3, 2025
Signed-off-by: Jan Kaniecki <[email protected]>
RichardoMrMu
pushed a commit
to RichardoMrMu/vllm
that referenced
this pull request
May 12, 2025
Signed-off-by: Jan Kaniecki <[email protected]> Signed-off-by: Mu Huai <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
When trying to perform regional compilation with t.compile (compiling layers separately instead of calling t.compile on whole model) on mllama model with Gaudi devices, such an error occures:
ValueError: Unknown decoder layer type <class 'torch._dynamo.eval_frame.OptimizedModule'>
Regional compilation for Gaudi devices has been added with #13213
Cause for this issue is checking layer classes by their names inside mllama code e.g.:
if isinstance(decoder_layer, MllamaCrossAttentionDecoderLayer):
Torch.compile wraps module after compilation with torch._dynamo.eval_frame.OptimizedModule name, that's wht we see mismatch in isinstance function. To resolve that we can distinguish layers basing on self.cross_attention_layers ids - and so do proposed changes. We don't also need raise ValueError in layer instance checking as there is no option for decoder layers to be of different types than desired ones.