Split sharded Llama dataset exporting and loading in export scripts #327
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
rsuderman
requested changes
Oct 24, 2024
| llama_config.tensor_parallelism_size = attn_q_weight.shard_count | ||
| llama_config = LlamaModelConfig( | ||
| hp, | ||
| tensor_parallelism_size=args.tensor_parallelism_size, |
There was a problem hiding this comment.
This will not work as it assumes that the passed arg is guaranteed to match the irpa file. We should look into plumming sharding into the saved hyper parameters then extract. I also dislike the hack in line 84 however the solution is not to introduce mismatches.
There was a problem hiding this comment.
OK, I will add it to the exported irpa hyper parameters.
rsuderman
requested changes
Oct 24, 2024
| llama_config.tensor_parallelism_size = attn_q_weight.shard_count | ||
| llama_config = LlamaModelConfig( | ||
| hp, | ||
| tensor_parallelism_size=dataset.properties["tensor_parallelism_size"], |
There was a problem hiding this comment.
Add a check prior for when tensor_parallelism_size has no value and in those cases default to 1. We should maintain that the old non-sharded irpa files still work.
rsuderman
requested changes
Oct 25, 2024
| llama_config.tensor_parallelism_size = attn_q_weight.shard_count | ||
| llama_config = LlamaModelConfig( | ||
| hp, | ||
| tensor_parallelism_size=dataset.properties["tensor_parallelism_size"] |
There was a problem hiding this comment.
Just declare it outside of llama_config. Giant comprehensions are bad for readability.
sogartar
force-pushed
the
sharded-llama-dataset-exporting
branch
from
2c22d38
to
38c3410
Compare
rsuderman
approved these changes
Oct 25, 2024
Separate the 2 steps. We need exported irpa files for the IREE module anyway.
sogartar
force-pushed
the
sharded-llama-dataset-exporting
branch
from
3fbb144
to
1dc7897
Compare
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.
Separate the 2 steps. We need exported irpa files for the IREE module anyway.