Support only-within-sequence attention for MosaicGPT #266
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
|
cc @jfrankle |
|
FYI, I think this PR has picked up some commits from the main branch that maybe weren't part of |
vchiley
requested changes
Mar 30, 2023
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Outdated suggestions cannot be applied.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.
Makes it possible to have attention restricted to tokens within the same source sequence when using pre-concatenated text dataloading.
eos_token_idandbos_token_idflags for thetextdataloader, which instructs it to add"sequence_id"to the batch using one of those tokens as the separator.attn_uses_sequence_idflag to MosaicGPT config/model andsequence_idto the forward arguments.sequence_idif the model config is set to use that input.125m comparison: https://wandb.ai/mosaic-ml/v004-125m
3b comparison: https://wandb.ai/mosaic-ml/v004-3b (for profiling; not a full convergence run)