Fix EP post merge#43730
Merged
Merged
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
…lecting for mergeModuleList
3outeille
reviewed
Feb 4, 2026
| shape[dim] = end - start | ||
| return tuple(shape) | ||
|
|
||
| class AllReduce(TensorParallelLayer): |
Member
There was a problem hiding this comment.
we should rename it else it becomes hard to understand given we have
def all_reduce_forward(x, device_mesh):
"""All-reduce forward, identity backward. Use after rowwise layers."""
return _AllReduceForward.apply(x, device_mesh)
Member
There was a problem hiding this comment.
we will use moe_tp_experts instead (handle backwards)
3outeille
added a commit
that referenced
this pull request
Feb 5, 2026
3outeille
reviewed
Feb 5, 2026
3outeille
reviewed
Feb 5, 2026
3outeille
reviewed
Feb 5, 2026
3outeille
reviewed
Feb 5, 2026
Contributor
|
[For maintainers] Suggested jobs to run (before merge) run-slow: glm4_moe_lite, gpt_oss, minimax_m2, solar_open, fbgemm_fp8 |
Member
|
lgtm nice job ! |
3outeille
approved these changes
Feb 5, 2026
ArthurZucker
added a commit
that referenced
this pull request
Feb 5, 2026
* restore * add all reduce for ep * fix init and bias sharding * fix finalize weight init * add full stacktracing * fix * okay big improvement here * the only case shard index should be used is when we are acctually collecting for mergeModuleList * more fixes * fix EP forward gpt oss * revert some shit * when you are stupid sometimes you really need a brain :) :) :) :) * fix TP * Ok GPT oss is fixed now * try to fix perms * attempt to fix * am I a doomer and AI is not that bad? * fix * it "passes" but the output is shit * style my man * outputs are gonna be giberish but at least the forward pass "works" * dtyle * fix mixtral * okay shape fixes * tensor idx is only for groupped gemm / EP * fix gate_up shard * fix :) * revert some EP changes that are breaking other stuff * style * use moe_tp_experts * revert unrelated, last nits and style * good? * fix modlar --------- Co-authored-by: 3outeille <ferdinand.mom@epita.fr>
jiosephlee
pushed a commit
to jiosephlee/transformers_latest
that referenced
this pull request
Feb 11, 2026
* restore * add all reduce for ep * fix init and bias sharding * fix finalize weight init * add full stacktracing * fix * okay big improvement here * the only case shard index should be used is when we are acctually collecting for mergeModuleList * more fixes * fix EP forward gpt oss * revert some shit * when you are stupid sometimes you really need a brain :) :) :) :) * fix TP * Ok GPT oss is fixed now * try to fix perms * attempt to fix * am I a doomer and AI is not that bad? * fix * it "passes" but the output is shit * style my man * outputs are gonna be giberish but at least the forward pass "works" * dtyle * fix mixtral * okay shape fixes * tensor idx is only for groupped gemm / EP * fix gate_up shard * fix :) * revert some EP changes that are breaking other stuff * style * use moe_tp_experts * revert unrelated, last nits and style * good? * fix modlar --------- Co-authored-by: 3outeille <ferdinand.mom@epita.fr>
2 tasks
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
What does this PR do?
Fix GPT-OSS ???
EP sharding works, the forward is broken for eager (does not work) and output giberissh for grouped_mm.
Will work on a fix to have in the patch!
Fix any TP model
tensor_idx