Updates configs to allow for the third failure mode #64
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
Added activation checkpointing in the json itself
This illustrates the third combination: pipeline parallelism, activation checkpoints, and ZeRO Stage 1. Again, this works. Upgrading ZeRO to Stage 2 causes it to fail.
This illustrates the third combination: pipeline parallelism, activation checkpoints, and ZeRO Stage 1. Again, this works. Upgrading ZeRO to Stage 2 causes it to fail.
StellaAthena
requested review from
lucidrains and
AranKomat
and removed request for
a team
StellaAthena
added a commit
that referenced
this pull request
Jan 17, 2021
* Added reduce_bucket_size argument to optimizer * Started to revamp train.py * revised training function, version 1 * Update gpt_neox.py * Update train.sh * Update base_deepspeed.json * Create train_gpt3small_pipeline.sh * Update train_stella.py * Update train_stella.py * adding environment.yml file * Set zero to stage 1 * set ZeRO to stage 2 * turned off activation checkpointing * turn continuous gradients off * turn activation checkpointing on * removed reduce bucket size config * turn on cont. graadients * removed extraneous comma * Turn off masking * Turned on attention masking * Change name in prep for merging * Rename base_deepspeed.json to deepspeed_zero2.json * Create deepspeed_zero1.json * Updated with new config names * Update with new config names * Update with new config names * Update with new config names * Rename train_stella.py to train_pipeline.py * Update README.md * Changed training loop to be consistent with train_enwik8.py * Reorder code to match train_enwik8.py * Modified transformerblock to pass mask argument Minor tweak to `forward` to align better with the demo code * Add print statement for testing * Revert changes to avoid pulling dev work into main Forgot about the open PR and pushed dev code to this branch. * Updates configs to allow for the third failure mode (#64) * Pipeline + Checkpoint - ZeRO 2 This illustrates the third combination: pipeline parallelism, activation checkpoints, and ZeRO Stage 1. Again, this works. Upgrading ZeRO to Stage 2 causes it to fail. Co-authored-by: Shivanshu Purohit <[email protected]> * Update train_enwik8.sh * Update deepspeed_zero2.json Co-authored-by: sdtblck <[email protected]> Co-authored-by: Shivanshu Purohit <[email protected]>
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.
No description provided.