gpt-oss model enablement #1754
Open
gpt-oss model enablement #1754
+1,453
−10
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
meta-cla
bot
added
the
CLA Signed
wwwjn
commented
Sep 24, 2025
torchtitan/models/attention.py
Outdated
| block_mask = FlexAttention.block_masks[self.mask_key] | ||
| return FlexAttention.flex_attn(q, k, v, block_mask=block_mask, scale=scale) | ||
|
|
||
| def _forward_with_sink( |
There was a problem hiding this comment.
LGTM.
I'm curious how expensive it is to always return lse. If it is actually no cost, we can merge the FlexAttention call to the original forward.
cc., @drisspg
wwwjn
force-pushed
the
gpt-oss
branch
2 times, most recently
from
48b2a11 to
07c0ff4
Compare
|
Need to rebase onto #1776 |
wwwjn
commented
Sep 30, 2025
tianyu-l
reviewed
Oct 5, 2025
|
|
||
|
|
||
| # TODO(jianiw): This need to be merged with expert_parallel | ||
| def expert_parallel(func: Callable) -> Callable: |
There was a problem hiding this comment.
sorry I'll merge my refactor, and then please rebase
There was a problem hiding this comment.
Are you referring to #1569 ?
| if use_sliding_attention: | ||
| self.attn = build_attention( | ||
| use_flex_attn=True, | ||
| attn_mask_type="sliding_window", |
There was a problem hiding this comment.
I think sliding_window should be orthogonal to causal vs. block-causal.
Namely, with document masking, the sliding window should only attend within single documents.
This should be much easier after #1776
…ks but reduces mfu for 20b
|
Summary of current status: There are some prerequisite PRs:
Once these PRs are landed, I will refactor:
|
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Outdated suggestions cannot be applied.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.
Keep developing on top of #1559. Thanks @KhoomeiK for initial contribution!
Initialized by the same seed checkpoint, set seed=0 and deterministic = True.
Run 1: dp_shard = 2
Run 2: dp_shard = 2, TP degree = 2 (NGPU=4)
Run 3: dp_shard = 2, TP degree =2, EP degree = 2 (NGPU=4)
Run 4: dp_shard = 2, TP degree = 2, EP degree = 2, ETP degree = 2 (NGPU=4)
Run 5: dp_shard=2, EP degree = 2 (NGPU=2)
