Skip to content

add sharded loading for safetensors in AutoTP (#4854)#2

Closed
tjs-intel wants to merge 1 commit into
HabanaAI:mainfrom
tjs-intel:support-safetensors
Closed

add sharded loading for safetensors in AutoTP (#4854)#2
tjs-intel wants to merge 1 commit into
HabanaAI:mainfrom
tjs-intel:support-safetensors

Conversation

@tjs-intel
Copy link
Copy Markdown

Adds support for sharded loading of Safetensors. Commit from upstream microsoft/DeepSpeed.

Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
Co-authored-by: Logan Adams <114770087+loadams@users.noreply.github.com>
Co-authored-by: Michael Wyatt <michaelwyatt@microsoft.com>
@tjs-intel
Copy link
Copy Markdown
Author

Some context: this is required for the TinyLlama model to function because it only stores checkpoints in the safetensors format.

@tjs-intel
Copy link
Copy Markdown
Author

@tjs-intel
Copy link
Copy Markdown
Author

@nelyahu can you please tag the appropriate reviewers, or let me know how I should make this contribution?

@nelyahu
Copy link
Copy Markdown

nelyahu commented Feb 2, 2024

@tjs-intel - Thanks, we will apply to change the our dev branch, and it will be available on next release.

@tjs-intel
Copy link
Copy Markdown
Author

@nelyahu what is the timeline of the next release?

@tjs-intel
Copy link
Copy Markdown
Author

I am promised that this will be in the next release

@tjs-intel tjs-intel closed this Feb 8, 2024
@tjs-intel tjs-intel deleted the support-safetensors branch March 26, 2024 22:03
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants