Vram fix attempt #1164
Merged
Vram fix attempt #1164
Conversation
This file contains 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
winglian
force-pushed
the
vram-fix-attempt
branch
from
30f49b6 to
584074d
Compare
djsaunde
pushed a commit
that referenced
this pull request
Dec 17, 2024
* revert order of filter/drop_long step and handle calc for max_input_len only during preprocessing * revert some changes to preparing for packing to allow more flexibility * prepare dataset for packing during pre-processing step * prepare dataset hash based on sample packing too * enclose none check * just cast straight to string for ds hash
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Outdated suggestions cannot be applied.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.
Description
more permanently fixes issues from #1127 dealing with VRAM instability. reverts an ordering change stemming from #1066 that was causing VRAM spikes during training. Updated dataset len calculations to use numpy instead of pandas since pandas can use the GPU and added a fallback calculation so we can still call it before dropping rows longer than the max length.
reverted order of callingprocess_datasets_for_packingfrom #1141 which can cause issues when changing max sequence length and needing to re-preprocess the datasets.Motivation and Context
How has this been tested?
Screenshots (if appropriate)
Types of changes
Social Handles (Optional)