fix INT8 prepare function #389
Merged
+16
−36
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
|
The documentation is not available anymore as the PR was closed or merged. |
younesbelkada
approved these changes
May 2, 2023
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
|
All the tests, single gpu, multi gpu, core and common pass |
4 tasks
4 tasks
Guy-Bilitski
pushed a commit
to Guy-Bilitski/peft
that referenced
this pull request
May 13, 2025
* fix INT8 prepare function * remove unused function args * fix related tests, examples and docs
cyyever
pushed a commit
to cyyever/peft
that referenced
this pull request
Sep 4, 2025
`dataloader must be a torch.utils.data.Dataset`: `dataloader` should be `dataset`
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Outdated suggestions cannot be applied.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.
What does this PR do?
a. When making modules trainable via
modules_to_save, it was resulting in loss beingNaNor0because those params are in fp16/bf16 which are unstable for training, in comparison trainable lora params are in FP32.b. half and float mismatches because certain layernorm were being converted in to FP32 and when making other trainable modules to be in FP32 manually, it was leading to mismatches as frozen layers were in half prescision and few in full precision.