Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[gguf] RE_GGUF_SHARD_FILE #601

Merged
merged 2 commits into from
Apr 3, 2024
Merged

[gguf] RE_GGUF_SHARD_FILE #601

merged 2 commits into from
Apr 3, 2024

Conversation

mishig25
Copy link
Collaborator

@mishig25 mishig25 commented Apr 3, 2024

I need on the frontend to detect which gguf file is sharded or not

cc: @ggerganov @madgetr

@mishig25 mishig25 marked this pull request as ready for review April 3, 2024 14:05
@mishig25 mishig25 requested a review from julien-c as a code owner April 3, 2024 14:05
@mishig25 mishig25 requested review from coyotte508 and gary149 and removed request for julien-c April 3, 2024 14:06
@mishig25 mishig25 merged commit e4d2d87 into main Apr 3, 2024
2 checks passed
@mishig25 mishig25 deleted the gguf_RE_GGUF_SHARD_FILE branch April 3, 2024 14:12
@mishig25 mishig25 requested a review from julien-c April 3, 2024 14:23
@madgetr
Copy link
Contributor

madgetr commented Apr 3, 2024

I need on the frontend to detect which gguf file is sharded or not

cc: @ggerganov @madgetr

Should these files be read as a single model? If so, which file is the authority? gguf doesn't seem to use an index.json file.

@julien-c
Copy link
Member

julien-c commented Apr 3, 2024

the first file of a sharded weights contains the complete metadata i.e. tokenizer etc. (cc @phymbert)

@madgetr
Copy link
Contributor

madgetr commented Apr 3, 2024

the first file of a sharded weights contains the complete metadata i.e. tokenizer etc. (cc @phymbert)

seems that the tensor shapes are scattered across the shards tho

@julien-c
Copy link
Member

julien-c commented Apr 3, 2024

true

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