Open
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #3240 +/- ##
=======================================
Coverage 28.33% 28.33%
=======================================
Files 33 33
Lines 4094 4094
=======================================
Hits 1160 1160
Misses 2934 2934 ☔ View full report in Codecov by Sentry. |
ssaliceTT
requested review from
AleksKnezevic,
acicovicTT,
ajakovljevicTT,
jameszianxuTT,
kmabeeTT,
mrakitaTT,
ndrakulicTT,
sdjukicTT,
sgligorijevicTT and
vzeljkovicTT
as code owners
ssaliceTT
force-pushed
the
ssalice/kimi-attn-mlp
branch
2 times, most recently
from
fe77298 to
2e3b912
Compare
…ce. No idea of sharding works tho.
… single device and TP. Need to test MoE. Also, cleaned up attention test a bit.
… Added seq_len = 32 hardcoding for testing MoE for now. Also, kimi k2 MoE needs to be set to .eval() manually due to some training thing in the model.
…CollectivePermuteOp not implemented.
… L1 cache error. Xfailed deepseek MoE due to ttnn.sort op not supporting float32. Xfailed kimi k2 MoE for dynamo error.
…h shard for Kimi MLP. Fails due to collect_permute op. Not sure why. Attention tests work fine with Aleks's shard specs
…cs. It all works now.
ssaliceTT
force-pushed
the
ssalice/kimi-attn-mlp
branch
from
2e3b912 to
fcde0ee
Compare
|
|
||
| # Override for single layer testing | ||
| config.num_hidden_layers = 1 | ||
| config.use_cache = False |
Contributor
There was a problem hiding this comment.
don't we want to use cache?
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Ticket
tt-xla #2954
Problem description
As a part of ongoing effort to run Kimi K2, attention and MLP tests needed to be added.
What's changed
Test running the attention module of Kimi K2 have been added to run on nightly.
Tests running the MLP of Kimi K2 have been added to run on nightly.
MLP tests are parametrized on MLP type, as it has regular MLP as well as MoE.
MoE is currently xfailed: fails due to a Dynamo compilation error.
Checklist