Update torchtitan for proper bf16 & new quant APIs #281
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
joecummings
commented
Oct 1, 2025
| ) | ||
| use_vllm_builtin_load: bool = True | ||
| compile: Compile = field(default_factory=Compile) | ||
| float8: Float8Dense = field(default_factory=Float8Dense) |
There was a problem hiding this comment.
This was also a part of the updated Torchtitan package
|
nice wandb logging :D |
🫡 |
photomz
pushed a commit
to photomz/forge
that referenced
this pull request
Oct 25, 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Outdated suggestions cannot be applied.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.
@felipemello1 Correctly diagnosed that our memory was super high for our default Qwen3 8B script. The solution he proposed in #278 was to drop the
max_req_tokensandmax_res_tokens, which was very valid. But overall the high memory seems suspect. After looking into it, it was apparent that although we specified "bfloat16" in our Trainer/Ref sections of the configs, it was not being applied - hence the massive memory. The culprit was that we had an out of date Torchtitan package :/In this PR, I update the Torchtitan package so that bf16 is applied correctly.
Memory usage before (with 486 seq len):
Memory usage after (with 512 seq len):
sidenote: This has the fortunate side effect of halving the weight sync speed #impact