Make safetensor shard filename parsing more flexible #1817
Conversation
This file contains hidden or 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
kefranabg
force-pushed
the
improve-parse-safetensor-shard-name
branch
from
889215b to
5eec567
Compare
julien-c
reviewed
Oct 21, 2025
| export const RE_SAFETENSORS_INDEX_FILE = /\.safetensors\.index\.json$/; | ||
| export const RE_SAFETENSORS_SHARD_FILE = | ||
| /^(?<prefix>(?<basePrefix>.*?)[_-])(?<shard>\d{5})-of-(?<total>\d{5})\.safetensors$/; | ||
| /^(?<prefix>(?<basePrefix>.*?)[_-])(?<shard>\d+)-of-(?<total>\d+)\.safetensors$/; |
There was a problem hiding this comment.
at least enforce the same number of digits 🙏 (not sure how to do this in the regex, maybe it's an OR or the different digits for 5 and 6 digits)
There was a problem hiding this comment.
or we just don't care about the 5X and just allow regular numbers to avoid this issue again in the future?
There was a problem hiding this comment.
In that case maybe it's better to handle this edge case directly on hub side and fix the shard digit numbers before passing it to parseSafetensorsShardFilename no?
Like transforming 000163 to 00163 before parsing it
There was a problem hiding this comment.
hmm no i think here is fine (we use this hub-side anyways)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
Context:
A lot of deepseek models have unconventional number of digits for the total shard number in their filename, which leads to a filename parsing fail. It breaks the file info navigation on the hub:
Kapture.2025-10-21.at.11.35.17.mp4
here for instance,
model-00001-of-000163.safetensorsshould bemodel-00001-of-00163.safetensors(there's 1 extra "0" in "model-00001-of-.safetensors")Most of deepseek models with a lot of shards have this issue, since they are important models, I suggest we make the regex more flexible to accept a variable number of digits for shards in the filename.
Would that be acceptable?