Estimate adapter memory overhead in choose_num_blocks() #346
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
borzunov
reviewed
Jul 13, 2023
| block, block_index=0, adapter_name=adapter, peft_config=peft_config, peft_state_dict=peft_state_dict | ||
| ) | ||
| adapter_parameters = sum(p.numel() for p in block.parameters()) - base_block_parameters | ||
| bytes_per_parameter = torch.finfo(resolve_block_dtype(block_config, torch_dtype)).bits / 8 |
There was a problem hiding this comment.
Doesn't adapters have a different dtype from the base block?
There was a problem hiding this comment.
they should have the same, lest forward pass would cause an error
borzunov
reviewed
Jul 13, 2023
borzunov
reviewed
Jul 13, 2023
borzunov
changed the title
Co-authored-by: Alexander Borzunov <borzunov.alexander@gmail.com>
Co-authored-by: Alexander Borzunov <borzunov.alexander@gmail.com>
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Outdated suggestions cannot be applied.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.
This PR changes the way the server determines how many blocks it can handle if that server has at least one adapter. It does so by including the adapter parameters in the total block size that divides the free GPU memory.
ToDos: