molt implementation #124
Open
molt implementation #124
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
Things to do: 1. a better rank distribution config setting strategy for pivoting model size easily; 2. distributed training logic 3. aliginging the output of prepare_input 4. unit tests
inplement low rank decomposed matrix multiplication; implement a tiny kernel fusion
exist a bug that make distributed and non-dist training not aligned and that will fixed later
…s done right directly from_local(reconstruction) will not all reduce across devices for reconstructions; now reconstrution also support data parallelism and aligned with sae there probably be minor bugs regarding decoder_norm and init, will be fixed later if necessary
…t inference and dist training Distinct model_parallel_size_training form model_parallel_size_running
MerlinRaptor
requested review from
Frankstein73,
Hzfinfdu and
dest1n1s
and removed request for
Hzfinfdu and
dest1n1s
dest1n1s
approved these changes
Sep 26, 2025
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Outdated suggestions cannot be applied.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.
Aside from mesh_shape assignment in src/lm_saes/runners/train.py (train_sae) 96 line, there should be no changes outside of molt, so this is expected to be a safe merge. Please let me know if you notice any unintended changes outside of molt.