url
string
repository_url
string
labels_url
string
comments_url
string
events_url
string
html_url
string
id
int64
node_id
string
number
int64
title
string
user
dict
labels
list
state
string
locked
bool
assignee
dict
assignees
list
milestone
null
comments
list
created_at
timestamp[ms]
updated_at
timestamp[ms]
closed_at
timestamp[ms]
author_association
string
type
dict
active_lock_reason
null
draft
bool
pull_request
dict
body
string
closed_by
dict
reactions
dict
timeline_url
string
performed_via_github_app
null
state_reason
string
sub_issues_summary
dict
issue_dependencies_summary
dict
is_pull_request
bool
is_closed
bool
https://api.github.com/repos/huggingface/transformers/issues/41442
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41442/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41442/comments
https://api.github.com/repos/huggingface/transformers/issues/41442/events
https://github.com/huggingface/transformers/pull/41442
3,494,909,707
PR_kwDOCUB6oc6srINg
41,442
Fix/perceptionlm auto alias
{ "login": "omsherikar", "id": 180152315, "node_id": "U_kgDOCrzn-w", "avatar_url": "https://avatars.githubusercontent.com/u/180152315?v=4", "gravatar_id": "", "url": "https://api.github.com/users/omsherikar", "html_url": "https://github.com/omsherikar", "followers_url": "https://api.github.com/users/omsherikar/followers", "following_url": "https://api.github.com/users/omsherikar/following{/other_user}", "gists_url": "https://api.github.com/users/omsherikar/gists{/gist_id}", "starred_url": "https://api.github.com/users/omsherikar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/omsherikar/subscriptions", "organizations_url": "https://api.github.com/users/omsherikar/orgs", "repos_url": "https://api.github.com/users/omsherikar/repos", "events_url": "https://api.github.com/users/omsherikar/events{/privacy}", "received_events_url": "https://api.github.com/users/omsherikar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T10:47:36
2025-10-08T13:23:46
2025-10-08T13:23:46
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41442", "html_url": "https://github.com/huggingface/transformers/pull/41442", "diff_url": "https://github.com/huggingface/transformers/pull/41442.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41442.patch", "merged_at": null }
### What does this PR do? Removes references to a non-existent base model from the Auto mappings so enumeration and lazy imports work as expected. - Remove `perception_encoder` entries from Auto mappings - `src/transformers/models/auto/modeling_auto.py` - `src/transformers/models/auto/configuration_auto.py` - Keep existing `perception_lm` entries intact - Regenerate modeling files (no runtime behavior change) Fixes: [#41387](https://github.com/huggingface/transformers/issues/41387) ### Motivation `MODEL_MAPPING.items()` failed because Auto mappings referenced `perception_encoder`, which doesn’t exist. Per maintainer guidance, we should remove non-existent models from auto-map rather than adding aliases. ### Dependencies None. ### Before submitting - [x] Minimal change; no behavior changes. - [x] Linked the issue. - [x] No docs changes required. - [x] Local check with `PYTHONPATH=src`: ```python from transformers.models.auto.modeling_auto import MODEL_MAPPING list(MODEL_MAPPING.items()) # succeeds ``` ### Who can review? - Model loading / auto mappings: @CyrilVallez - Multimodal: @zucchini-nlp
{ "login": "omsherikar", "id": 180152315, "node_id": "U_kgDOCrzn-w", "avatar_url": "https://avatars.githubusercontent.com/u/180152315?v=4", "gravatar_id": "", "url": "https://api.github.com/users/omsherikar", "html_url": "https://github.com/omsherikar", "followers_url": "https://api.github.com/users/omsherikar/followers", "following_url": "https://api.github.com/users/omsherikar/following{/other_user}", "gists_url": "https://api.github.com/users/omsherikar/gists{/gist_id}", "starred_url": "https://api.github.com/users/omsherikar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/omsherikar/subscriptions", "organizations_url": "https://api.github.com/users/omsherikar/orgs", "repos_url": "https://api.github.com/users/omsherikar/repos", "events_url": "https://api.github.com/users/omsherikar/events{/privacy}", "received_events_url": "https://api.github.com/users/omsherikar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41442/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41442/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41441
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41441/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41441/comments
https://api.github.com/repos/huggingface/transformers/issues/41441/events
https://github.com/huggingface/transformers/pull/41441
3,494,891,435
PR_kwDOCUB6oc6srELY
41,441
Enhance the handling of Union types in HfArgumentParser
{ "login": "cyyever", "id": 17618148, "node_id": "MDQ6VXNlcjE3NjE4MTQ4", "avatar_url": "https://avatars.githubusercontent.com/u/17618148?v=4", "gravatar_id": "", "url": "https://api.github.com/users/cyyever", "html_url": "https://github.com/cyyever", "followers_url": "https://api.github.com/users/cyyever/followers", "following_url": "https://api.github.com/users/cyyever/following{/other_user}", "gists_url": "https://api.github.com/users/cyyever/gists{/gist_id}", "starred_url": "https://api.github.com/users/cyyever/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/cyyever/subscriptions", "organizations_url": "https://api.github.com/users/cyyever/orgs", "repos_url": "https://api.github.com/users/cyyever/repos", "events_url": "https://api.github.com/users/cyyever/events{/privacy}", "received_events_url": "https://api.github.com/users/cyyever/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-08T10:41:51
2025-10-18T00:44:19
null
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41441", "html_url": "https://github.com/huggingface/transformers/pull/41441", "diff_url": "https://github.com/huggingface/transformers/pull/41441.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41441.patch", "merged_at": null }
# What does this PR do? This PR enhance the handling of Union types containing `str` by two rules: 1. If `str` in `Union` and `Union` has more than one other types pass `str` to `parser.add_argument`. The caller is responsible to convert the string argument to a more proper type. 2. Otherwise, `Union` must be `X | str` and pass `X` to `parser.add_argument`
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41441/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41441/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41440
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41440/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41440/comments
https://api.github.com/repos/huggingface/transformers/issues/41440/events
https://github.com/huggingface/transformers/pull/41440
3,494,721,771
PR_kwDOCUB6oc6sqfC1
41,440
Fix generate outputs and simplify cache tests
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T09:52:54
2025-10-08T17:04:20
2025-10-08T17:04:19
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41440", "html_url": "https://github.com/huggingface/transformers/pull/41440", "diff_url": "https://github.com/huggingface/transformers/pull/41440.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41440.patch", "merged_at": "2025-10-08T17:04:19" }
# What does this PR do? Improve the cache tests.`test_past_key_values_format` should really rely on `_check_past_key_values_for_generate` as they do the same, and we otherwise have to overwrite both for models with non-standard cache format, which is really annoying! This made me realize that `generate` does not correctly return the cache, if it's not called `past_key_values`. This fixes it as well. cc @gante
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41440/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41440/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41439
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41439/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41439/comments
https://api.github.com/repos/huggingface/transformers/issues/41439/events
https://github.com/huggingface/transformers/pull/41439
3,494,694,076
PR_kwDOCUB6oc6sqZA9
41,439
Patch MistralCommonTokenizer
{ "login": "juliendenize", "id": 40604584, "node_id": "MDQ6VXNlcjQwNjA0NTg0", "avatar_url": "https://avatars.githubusercontent.com/u/40604584?v=4", "gravatar_id": "", "url": "https://api.github.com/users/juliendenize", "html_url": "https://github.com/juliendenize", "followers_url": "https://api.github.com/users/juliendenize/followers", "following_url": "https://api.github.com/users/juliendenize/following{/other_user}", "gists_url": "https://api.github.com/users/juliendenize/gists{/gist_id}", "starred_url": "https://api.github.com/users/juliendenize/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/juliendenize/subscriptions", "organizations_url": "https://api.github.com/users/juliendenize/orgs", "repos_url": "https://api.github.com/users/juliendenize/repos", "events_url": "https://api.github.com/users/juliendenize/events{/privacy}", "received_events_url": "https://api.github.com/users/juliendenize/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T09:44:35
2025-10-14T11:13:19
2025-10-14T11:13:19
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41439", "html_url": "https://github.com/huggingface/transformers/pull/41439", "diff_url": "https://github.com/huggingface/transformers/pull/41439.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41439.patch", "merged_at": "2025-10-14T11:13:19" }
# What does this PR do? This PR patches the `MistralCommonTokenizer`: 1. [BUG FIX] spm now is correctly supported as previous usage would result in an error for `_piece_to_id` and `_is_control_token`. 2. [BUG FIX] previous implementation of `_piece_to_id` was incorrect for Tekkenizer. a) the special tokens were not supported b) the normal tokens were not shifted by adding the number of special tokens 4. [MAYBE BUG FIX] Changed `get_vocab` to a function that should better mimic what happens in `Transformers`: now the mapping is based on the real ids but some ids are missing due to conversion loss of some tokens in Tekken. 5. [FEATURE] `add_generation_prompt` has been added. This is to match signature of Transformers. In practice this value is ignored except if: a) `continue_final_message` and `add_generation_prompt` are `True` an error is raised. b) if `add_generation_prompt` is `True` and the last message is `assistant` then an error is raised as the user should have passed `continue_final_message`. 6. ~[FEATURE] Now the tokenizer is set to fast because:~ a) ~it is true~(Edit: not so sure about that after discussion) b) it removes annoying message when initializing the tokenizer. Edit 2: this has been reverted as it doesn't bring value and is misleading. 7. [OPTIMIZATION] now image tensors are converted without slowness warnings from torch. 8. [DOCS] updated some docs to remove unused args. Also added minimal tests to ensure SPM also works in the future. ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [x] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [x] Did you write any new necessary tests? ## Who can review? @patrickvonplaten @ArthurZucker @itazap
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41439/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41439/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41438
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41438/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41438/comments
https://api.github.com/repos/huggingface/transformers/issues/41438/events
https://github.com/huggingface/transformers/pull/41438
3,494,665,601
PR_kwDOCUB6oc6sqS2u
41,438
Update philosophy
{ "login": "molbap", "id": 39954772, "node_id": "MDQ6VXNlcjM5OTU0Nzcy", "avatar_url": "https://avatars.githubusercontent.com/u/39954772?v=4", "gravatar_id": "", "url": "https://api.github.com/users/molbap", "html_url": "https://github.com/molbap", "followers_url": "https://api.github.com/users/molbap/followers", "following_url": "https://api.github.com/users/molbap/following{/other_user}", "gists_url": "https://api.github.com/users/molbap/gists{/gist_id}", "starred_url": "https://api.github.com/users/molbap/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/molbap/subscriptions", "organizations_url": "https://api.github.com/users/molbap/orgs", "repos_url": "https://api.github.com/users/molbap/repos", "events_url": "https://api.github.com/users/molbap/events{/privacy}", "received_events_url": "https://api.github.com/users/molbap/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T09:35:42
2025-10-10T06:52:19
2025-10-10T06:52:18
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41438", "html_url": "https://github.com/huggingface/transformers/pull/41438", "diff_url": "https://github.com/huggingface/transformers/pull/41438.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41438.patch", "merged_at": "2025-10-10T06:52:18" }
# What does this PR do? Update our philosophy page to reflect developments in the recent [blog post](https://huggingface.co/spaces/transformers-community/Transformers-tenets). Explicitly mention tenets, include modular transformers, and link to the blog.
{ "login": "molbap", "id": 39954772, "node_id": "MDQ6VXNlcjM5OTU0Nzcy", "avatar_url": "https://avatars.githubusercontent.com/u/39954772?v=4", "gravatar_id": "", "url": "https://api.github.com/users/molbap", "html_url": "https://github.com/molbap", "followers_url": "https://api.github.com/users/molbap/followers", "following_url": "https://api.github.com/users/molbap/following{/other_user}", "gists_url": "https://api.github.com/users/molbap/gists{/gist_id}", "starred_url": "https://api.github.com/users/molbap/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/molbap/subscriptions", "organizations_url": "https://api.github.com/users/molbap/orgs", "repos_url": "https://api.github.com/users/molbap/repos", "events_url": "https://api.github.com/users/molbap/events{/privacy}", "received_events_url": "https://api.github.com/users/molbap/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41438/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41438/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41437
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41437/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41437/comments
https://api.github.com/repos/huggingface/transformers/issues/41437/events
https://github.com/huggingface/transformers/pull/41437
3,494,573,524
PR_kwDOCUB6oc6sp-37
41,437
perception_lm: remove perception_encoder from auto-maps
{ "login": "omsherikar", "id": 180152315, "node_id": "U_kgDOCrzn-w", "avatar_url": "https://avatars.githubusercontent.com/u/180152315?v=4", "gravatar_id": "", "url": "https://api.github.com/users/omsherikar", "html_url": "https://github.com/omsherikar", "followers_url": "https://api.github.com/users/omsherikar/followers", "following_url": "https://api.github.com/users/omsherikar/following{/other_user}", "gists_url": "https://api.github.com/users/omsherikar/gists{/gist_id}", "starred_url": "https://api.github.com/users/omsherikar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/omsherikar/subscriptions", "organizations_url": "https://api.github.com/users/omsherikar/orgs", "repos_url": "https://api.github.com/users/omsherikar/repos", "events_url": "https://api.github.com/users/omsherikar/events{/privacy}", "received_events_url": "https://api.github.com/users/omsherikar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T09:09:39
2025-10-09T12:12:17
2025-10-09T12:12:17
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41437", "html_url": "https://github.com/huggingface/transformers/pull/41437", "diff_url": "https://github.com/huggingface/transformers/pull/41437.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41437.patch", "merged_at": null }
### What does this PR do? Removes references to a non-existent base model from the Auto mappings so enumeration and lazy imports work as expected. - Remove `perception_encoder` entries from Auto mappings - `src/transformers/models/auto/modeling_auto.py` - `src/transformers/models/auto/configuration_auto.py` - Keep existing `perception_lm` entries intact - Regenerate modeling files (no runtime behavior change) Fixes: [#41387](https://github.com/huggingface/transformers/issues/41387) ### Motivation `MODEL_MAPPING.items()` failed because Auto mappings referenced `perception_encoder`, which doesn’t exist. Per maintainer guidance, we should remove non-existent models from auto-map rather than adding aliases. ### Dependencies None. ### Before submitting - [x] Minimal change; no behavior changes. - [x] Linked the issue. - [x] No docs changes required. - [x] Local check with `PYTHONPATH=src`: ```python from transformers.models.auto.modeling_auto import MODEL_MAPPING list(MODEL_MAPPING.items()) # succeeds ``` ### Who can review? - Model loading / auto mappings: @CyrilVallez - Multimodal: @zucchini-nlp
{ "login": "omsherikar", "id": 180152315, "node_id": "U_kgDOCrzn-w", "avatar_url": "https://avatars.githubusercontent.com/u/180152315?v=4", "gravatar_id": "", "url": "https://api.github.com/users/omsherikar", "html_url": "https://github.com/omsherikar", "followers_url": "https://api.github.com/users/omsherikar/followers", "following_url": "https://api.github.com/users/omsherikar/following{/other_user}", "gists_url": "https://api.github.com/users/omsherikar/gists{/gist_id}", "starred_url": "https://api.github.com/users/omsherikar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/omsherikar/subscriptions", "organizations_url": "https://api.github.com/users/omsherikar/orgs", "repos_url": "https://api.github.com/users/omsherikar/repos", "events_url": "https://api.github.com/users/omsherikar/events{/privacy}", "received_events_url": "https://api.github.com/users/omsherikar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41437/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41437/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41436
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41436/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41436/comments
https://api.github.com/repos/huggingface/transformers/issues/41436/events
https://github.com/huggingface/transformers/pull/41436
3,494,438,544
PR_kwDOCUB6oc6sphYF
41,436
v5 dev version
{ "login": "LysandreJik", "id": 30755778, "node_id": "MDQ6VXNlcjMwNzU1Nzc4", "avatar_url": "https://avatars.githubusercontent.com/u/30755778?v=4", "gravatar_id": "", "url": "https://api.github.com/users/LysandreJik", "html_url": "https://github.com/LysandreJik", "followers_url": "https://api.github.com/users/LysandreJik/followers", "following_url": "https://api.github.com/users/LysandreJik/following{/other_user}", "gists_url": "https://api.github.com/users/LysandreJik/gists{/gist_id}", "starred_url": "https://api.github.com/users/LysandreJik/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/LysandreJik/subscriptions", "organizations_url": "https://api.github.com/users/LysandreJik/orgs", "repos_url": "https://api.github.com/users/LysandreJik/repos", "events_url": "https://api.github.com/users/LysandreJik/events{/privacy}", "received_events_url": "https://api.github.com/users/LysandreJik/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T08:33:53
2025-10-08T08:45:35
2025-10-08T08:45:33
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41436", "html_url": "https://github.com/huggingface/transformers/pull/41436", "diff_url": "https://github.com/huggingface/transformers/pull/41436.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41436.patch", "merged_at": "2025-10-08T08:45:33" }
null
{ "login": "LysandreJik", "id": 30755778, "node_id": "MDQ6VXNlcjMwNzU1Nzc4", "avatar_url": "https://avatars.githubusercontent.com/u/30755778?v=4", "gravatar_id": "", "url": "https://api.github.com/users/LysandreJik", "html_url": "https://github.com/LysandreJik", "followers_url": "https://api.github.com/users/LysandreJik/followers", "following_url": "https://api.github.com/users/LysandreJik/following{/other_user}", "gists_url": "https://api.github.com/users/LysandreJik/gists{/gist_id}", "starred_url": "https://api.github.com/users/LysandreJik/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/LysandreJik/subscriptions", "organizations_url": "https://api.github.com/users/LysandreJik/orgs", "repos_url": "https://api.github.com/users/LysandreJik/repos", "events_url": "https://api.github.com/users/LysandreJik/events{/privacy}", "received_events_url": "https://api.github.com/users/LysandreJik/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41436/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41436/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41435
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41435/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41435/comments
https://api.github.com/repos/huggingface/transformers/issues/41435/events
https://github.com/huggingface/transformers/pull/41435
3,494,436,084
PR_kwDOCUB6oc6spg1E
41,435
Allow passing `tp_plan` in `from_pretrained` directly
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T08:33:19
2025-10-16T09:12:09
2025-10-16T09:12:07
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41435", "html_url": "https://github.com/huggingface/transformers/pull/41435", "diff_url": "https://github.com/huggingface/transformers/pull/41435.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41435.patch", "merged_at": "2025-10-16T09:12:07" }
# What does this PR do? As per the title. Also fixes the tp_plan internal for non-based models. I checked with the following, and it works nicely: ```python from transformers import AutoModelForCausalLM, AutoTokenizer import torch from torch.distributed.tensor.debug import visualize_sharding from torch.distributed.tensor import DTensor import os rank = int(os.environ["RANK"]) model_id = "meta-llama/Llama-3.1-8B-Instruct" # only shard attentions, but not mlps tp_plan = { "model.layers.*.self_attn.q_proj": "colwise", "model.layers.*.self_attn.k_proj": "colwise", "model.layers.*.self_attn.v_proj": "colwise", "model.layers.*.self_attn.o_proj": "rowwise", } tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained( model_id, dtype=torch.bfloat16, tp_plan=tp_plan, ) inputs = tokenizer(["How are you?"], return_tensors="pt").to(rank) out = model.generate(**inputs, max_new_tokens=10, do_sample=False) text = tokenizer.batch_decode(out[:,inputs.input_ids.shape[-1]:]) if rank == 0: print(text) print(model.tp_plan) print("SHARDING OF Q_PROJ") visualize_sharding(model.model.layers[0].self_attn.q_proj.weight) print("SHARDING OF UP_PROJ") if isinstance(model.model.layers[0].mlp.up_proj.weight, DTensor): visualize_sharding(model.model.layers[0].mlp.up_proj.weight) else: print("NOT SHARDED") torch.distributed.destroy_process_group() ```
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41435/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41435/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41434
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41434/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41434/comments
https://api.github.com/repos/huggingface/transformers/issues/41434/events
https://github.com/huggingface/transformers/pull/41434
3,494,164,413
PR_kwDOCUB6oc6somGb
41,434
Fix: Add Perceiver model registry test and correct auto mappings
{ "login": "aijadugar", "id": 139578960, "node_id": "U_kgDOCFHOUA", "avatar_url": "https://avatars.githubusercontent.com/u/139578960?v=4", "gravatar_id": "", "url": "https://api.github.com/users/aijadugar", "html_url": "https://github.com/aijadugar", "followers_url": "https://api.github.com/users/aijadugar/followers", "following_url": "https://api.github.com/users/aijadugar/following{/other_user}", "gists_url": "https://api.github.com/users/aijadugar/gists{/gist_id}", "starred_url": "https://api.github.com/users/aijadugar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/aijadugar/subscriptions", "organizations_url": "https://api.github.com/users/aijadugar/orgs", "repos_url": "https://api.github.com/users/aijadugar/repos", "events_url": "https://api.github.com/users/aijadugar/events{/privacy}", "received_events_url": "https://api.github.com/users/aijadugar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T07:08:54
2025-10-26T18:24:50
2025-10-26T18:24:50
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41434", "html_url": "https://github.com/huggingface/transformers/pull/41434", "diff_url": "https://github.com/huggingface/transformers/pull/41434.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41434.patch", "merged_at": null }
# Summary This PR adds testing and registry validation for the **Perceiver** model within the Hugging Face Transformers codebase. ## Changes Made - Tested a standalone `test_registry.py` to validate Perceiver model registration and tokenizer mapping. - Corrected auto model and tokenizer mappings to ensure `perceiver` is properly recognized in: - `modeling_auto.py` - `tokenization_auto.py` - Verified successful forward pass for `PerceiverModel` with output shape `torch.Size([1, 256, 1280])`. - Fixed missing config attributes (e.g., `input_channels`) to enable correct model instantiation. - Confirmed tokenizer functionality via `PerceiverTokenizer`. ## Verification <img width="1311" height="350" alt="Screenshot 2025-10-08 122041" src="https://github.com/user-attachments/assets/ff1a512e-fdf6-4fa3-beae-ef1e704afb79" /> <img width="1311" height="248" alt="Screenshot 2025-10-08 121126" src="https://github.com/user-attachments/assets/38ba7461-82b0-4be8-8403-02d3c64c19aa" /> <img width="1316" height="228" alt="Screenshot 2025-10-08 122210" src="https://github.com/user-attachments/assets/bc0a6575-3ce3-4247-9119-10375adcef48" />
{ "login": "aijadugar", "id": 139578960, "node_id": "U_kgDOCFHOUA", "avatar_url": "https://avatars.githubusercontent.com/u/139578960?v=4", "gravatar_id": "", "url": "https://api.github.com/users/aijadugar", "html_url": "https://github.com/aijadugar", "followers_url": "https://api.github.com/users/aijadugar/followers", "following_url": "https://api.github.com/users/aijadugar/following{/other_user}", "gists_url": "https://api.github.com/users/aijadugar/gists{/gist_id}", "starred_url": "https://api.github.com/users/aijadugar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/aijadugar/subscriptions", "organizations_url": "https://api.github.com/users/aijadugar/orgs", "repos_url": "https://api.github.com/users/aijadugar/repos", "events_url": "https://api.github.com/users/aijadugar/events{/privacy}", "received_events_url": "https://api.github.com/users/aijadugar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41434/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41434/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41433
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41433/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41433/comments
https://api.github.com/repos/huggingface/transformers/issues/41433/events
https://github.com/huggingface/transformers/pull/41433
3,493,954,967
PR_kwDOCUB6oc6sn4s4
41,433
Fix: remove invalid PerceptionEncoder and ParakeetCTCTokenizer registry entries (#41387)
{ "login": "aijadugar", "id": 139578960, "node_id": "U_kgDOCFHOUA", "avatar_url": "https://avatars.githubusercontent.com/u/139578960?v=4", "gravatar_id": "", "url": "https://api.github.com/users/aijadugar", "html_url": "https://github.com/aijadugar", "followers_url": "https://api.github.com/users/aijadugar/followers", "following_url": "https://api.github.com/users/aijadugar/following{/other_user}", "gists_url": "https://api.github.com/users/aijadugar/gists{/gist_id}", "starred_url": "https://api.github.com/users/aijadugar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/aijadugar/subscriptions", "organizations_url": "https://api.github.com/users/aijadugar/orgs", "repos_url": "https://api.github.com/users/aijadugar/repos", "events_url": "https://api.github.com/users/aijadugar/events{/privacy}", "received_events_url": "https://api.github.com/users/aijadugar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-08T05:55:48
2025-10-26T18:24:16
2025-10-26T18:24:16
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41433", "html_url": "https://github.com/huggingface/transformers/pull/41433", "diff_url": "https://github.com/huggingface/transformers/pull/41433.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41433.patch", "merged_at": null }
This PR fixes issue #41387 by commenting out invalid auto registry entries: - Removed non-existent `PerceptionEncoder` mapping in `modeling_auto.py` - Removed non-existent `ParakeetCTCTokenizer` mapping in `tokenization_auto.py` Local tests: <img width="1381" height="334" alt="image" src="https://github.com/user-attachments/assets/5947e69e-9f84-41c2-bd7c-e40dddc21b8a" />
{ "login": "aijadugar", "id": 139578960, "node_id": "U_kgDOCFHOUA", "avatar_url": "https://avatars.githubusercontent.com/u/139578960?v=4", "gravatar_id": "", "url": "https://api.github.com/users/aijadugar", "html_url": "https://github.com/aijadugar", "followers_url": "https://api.github.com/users/aijadugar/followers", "following_url": "https://api.github.com/users/aijadugar/following{/other_user}", "gists_url": "https://api.github.com/users/aijadugar/gists{/gist_id}", "starred_url": "https://api.github.com/users/aijadugar/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/aijadugar/subscriptions", "organizations_url": "https://api.github.com/users/aijadugar/orgs", "repos_url": "https://api.github.com/users/aijadugar/repos", "events_url": "https://api.github.com/users/aijadugar/events{/privacy}", "received_events_url": "https://api.github.com/users/aijadugar/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41433/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41433/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41432
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41432/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41432/comments
https://api.github.com/repos/huggingface/transformers/issues/41432/events
https://github.com/huggingface/transformers/pull/41432
3,493,637,633
PR_kwDOCUB6oc6sm2W8
41,432
[WIP] Refactor check_auto_docstring using AST
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-08T02:57:20
2025-10-08T02:57:20
null
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41432", "html_url": "https://github.com/huggingface/transformers/pull/41432", "diff_url": "https://github.com/huggingface/transformers/pull/41432.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41432.patch", "merged_at": null }
# What does this PR do? Use AST instead of parsing the raw code to check for missing args in docstrings of class/functions using auto_docstring
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41432/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41432/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41431
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41431/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41431/comments
https://api.github.com/repos/huggingface/transformers/issues/41431/events
https://github.com/huggingface/transformers/issues/41431
3,493,196,205
I_kwDOCUB6oc7QNfGt
41,431
gradient scaling occurs even though total gradient remains < max_grad_norm in trainer.py
{ "login": "lorsonblair", "id": 56801497, "node_id": "MDQ6VXNlcjU2ODAxNDk3", "avatar_url": "https://avatars.githubusercontent.com/u/56801497?v=4", "gravatar_id": "", "url": "https://api.github.com/users/lorsonblair", "html_url": "https://github.com/lorsonblair", "followers_url": "https://api.github.com/users/lorsonblair/followers", "following_url": "https://api.github.com/users/lorsonblair/following{/other_user}", "gists_url": "https://api.github.com/users/lorsonblair/gists{/gist_id}", "starred_url": "https://api.github.com/users/lorsonblair/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/lorsonblair/subscriptions", "organizations_url": "https://api.github.com/users/lorsonblair/orgs", "repos_url": "https://api.github.com/users/lorsonblair/repos", "events_url": "https://api.github.com/users/lorsonblair/events{/privacy}", "received_events_url": "https://api.github.com/users/lorsonblair/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-07T22:13:08
2025-10-09T17:16:20
null
NONE
null
null
null
null
Even though gradients remain < max_grad_norm throughout training, the gradient still goes through a scaling process. For instance, I set max_grad_norm = 1, and grad_norm consistently remains <= 0.33. Because the trainer takes you through the grad clip process if max_grad_norm > 0 or not None, this operation always gets executed within torch's clip function: `clip_coef = max_norm / (total_norm + 1e-6)`. Is there a way to prevent this? Thanks.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41431/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41431/timeline
null
null
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
false
https://api.github.com/repos/huggingface/transformers/issues/41430
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41430/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41430/comments
https://api.github.com/repos/huggingface/transformers/issues/41430/events
https://github.com/huggingface/transformers/pull/41430
3,493,038,725
PR_kwDOCUB6oc6sk61a
41,430
Fix Marian ONNX export with incorrect output (resolves #40122)
{ "login": "aaron-seq", "id": 96731649, "node_id": "U_kgDOBcQCAQ", "avatar_url": "https://avatars.githubusercontent.com/u/96731649?v=4", "gravatar_id": "", "url": "https://api.github.com/users/aaron-seq", "html_url": "https://github.com/aaron-seq", "followers_url": "https://api.github.com/users/aaron-seq/followers", "following_url": "https://api.github.com/users/aaron-seq/following{/other_user}", "gists_url": "https://api.github.com/users/aaron-seq/gists{/gist_id}", "starred_url": "https://api.github.com/users/aaron-seq/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/aaron-seq/subscriptions", "organizations_url": "https://api.github.com/users/aaron-seq/orgs", "repos_url": "https://api.github.com/users/aaron-seq/repos", "events_url": "https://api.github.com/users/aaron-seq/events{/privacy}", "received_events_url": "https://api.github.com/users/aaron-seq/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 9258341780, "node_id": "LA_kwDOCUB6oc8AAAACJ9cVlA", "url": "https://api.github.com/repos/huggingface/transformers/labels/Code%20agent%20slop", "name": "Code agent slop", "color": "C59579", "default": false, "description": "" } ]
closed
false
null
[]
null
[]
2025-10-07T21:06:32
2025-10-08T23:40:53
2025-10-08T12:35:17
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41430", "html_url": "https://github.com/huggingface/transformers/pull/41430", "diff_url": "https://github.com/huggingface/transformers/pull/41430.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41430.patch", "merged_at": null }
## Description This PR fixes the ONNX export issues for Marian translation models described in issue #40122. The original problem was that ONNX exported Marian models produced incorrect translation outputs compared to their PyTorch counterparts. ## Problem Summary The issue manifested when users exported OPUS (Marian) models to ONNX format: - **Wrong ONNX outputs**: Translations were incomplete or contained incorrect tokens - **Missing attention masks**: Decoder wasn't properly handling attention masks during export - **Token positioning errors**: Incorrect handling of token positions during autoregressive generation - **Generation logic bugs**: The inference code had issues in the token generation loop ### Example of the Issue **Input**: "Using handheld GPS devices and programs like Google Earth, members of the Trio Tribe, who live in the rainforests of southern Suriname, map out their ancestral lands to help strengthen their territorial claims." **Expected Output**: باستخدام أجهزة GPS المحمولة وبرامج مثل Google Earth ، يقوم أعضاء Trio Tribe ، الذين يعيشون في الغابات المطيرة في جنوب سورينام ، برسم خرائط لأراضي أجدادهم للمساعدة في تعزيز مطالبهم الإقليمية. **Broken ONNX Output**: باستخدام أجهزة GPS المحمولة وبرامج مثل جوجل أعضاء Tri، الذين يعيشون في الغابات جنوب سورينام رسم أراضي الفوركس للمساعدة تعزيز الإقليمية.. ## Solution This PR provides a comprehensive fix through: ### 1. Enhanced Decoder Wrapper (`MarianDecoderONNX`) - Proper handling of decoder attention masks during ONNX export - Correct application of causal masking for autoregressive generation - Maintains compatibility with the original PyTorch model behavior ### 2. Fixed ONNX Export Functions - `export_marian_encoder_to_onnx()`: Proper encoder export with correct input handling - `export_marian_decoder_to_onnx()`: Enhanced decoder export with attention mask fixes - Proper dynamic axes configuration for flexible input sizes ### 3. Corrected Inference Logic - Fixed token generation loop that matches PyTorch behavior - Proper initialization of decoder inputs with BOS tokens - Correct attention mask creation during sequential generation - Fixed token position indexing in the generation loop ### 4. Comprehensive Testing - Unit tests for encoder/decoder wrappers - Integration tests comparing PyTorch vs ONNX outputs - End-to-end translation validation - Edge case testing for various input scenarios ## Files Changed - **`examples/pytorch/seq2seq/marian_onnx_export.py`**: Main utility script with fixed export functions - **`tests/models/marian/test_marian_onnx_export.py`**: Comprehensive test suite - **`examples/pytorch/seq2seq/README_marian_onnx.md`**: Detailed documentation and usage guide ## Key Technical Fixes ### Attention Mask Handling ```python # Apply causal masking to decoder attention mask batch_size, seq_len = input_ids.shape causal_mask = torch.tril(torch.ones(seq_len, seq_len, device=input_ids.device, dtype=torch.bool)) expanded_causal_mask = causal_mask.unsqueeze(0).expand(batch_size, -1, -1) decoder_attention_mask = decoder_attention_mask.unsqueeze(1) & expanded_causal_mask ``` ### Fixed Generation Loop ```python for step in range(1, max_length): # Create current decoder attention mask (attend only to generated tokens so far) current_decoder_mask = np.zeros((batch_size, max_length), dtype=np.int64) current_decoder_mask[:, :step] = 1 # Get next token logits at the correct position next_token_logits = logits[0, step - 1, :] next_token_id = np.argmax(next_token_logits) ``` ## Supported Models Tested with various OPUS models: - `Helsinki-NLP/opus-mt-en-ar` (English to Arabic) - `Helsinki-NLP/opus-mt-en-de` (English to German) - `Helsinki-NLP/opus-mt-en-fr` (English to French) - `Helsinki-NLP/opus-mt-en-es` (English to Spanish) ## Breaking Changes None. This is a pure fix that maintains backward compatibility while correcting the ONNX export behavior. ## Usage Example ```python from transformers import MarianMTModel, MarianTokenizer from examples.pytorch.seq2seq.marian_onnx_export import export_marian_decoder_to_onnx # Load model model = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-en-ar") tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-ar") # Export with fix export_marian_decoder_to_onnx(model, tokenizer, "decoder_fixed.onnx") ``` ## Closes Closes #40122 ## Checklist - [x] I have tested this change on the original problematic example - [ ] I have added comprehensive tests - [ ] I have added documentation - [x] The fix maintains backward compatibility - [x] ONNX outputs now match PyTorch outputs - [x] All existing tests still pass
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41430/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41430/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41429
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41429/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41429/comments
https://api.github.com/repos/huggingface/transformers/issues/41429/events
https://github.com/huggingface/transformers/pull/41429
3,493,023,216
PR_kwDOCUB6oc6sk3fr
41,429
enhance patched_tearDown to support python 3.11+
{ "login": "yao-matrix", "id": 7245027, "node_id": "MDQ6VXNlcjcyNDUwMjc=", "avatar_url": "https://avatars.githubusercontent.com/u/7245027?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yao-matrix", "html_url": "https://github.com/yao-matrix", "followers_url": "https://api.github.com/users/yao-matrix/followers", "following_url": "https://api.github.com/users/yao-matrix/following{/other_user}", "gists_url": "https://api.github.com/users/yao-matrix/gists{/gist_id}", "starred_url": "https://api.github.com/users/yao-matrix/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yao-matrix/subscriptions", "organizations_url": "https://api.github.com/users/yao-matrix/orgs", "repos_url": "https://api.github.com/users/yao-matrix/repos", "events_url": "https://api.github.com/users/yao-matrix/events{/privacy}", "received_events_url": "https://api.github.com/users/yao-matrix/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T21:01:44
2025-10-09T15:57:25
2025-10-09T15:49:29
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41429", "html_url": "https://github.com/huggingface/transformers/pull/41429", "diff_url": "https://github.com/huggingface/transformers/pull/41429.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41429.patch", "merged_at": "2025-10-09T15:49:29" }
@ydshieh , I tried your `PATCH_TESTING_METHODS_TO_COLLECT_OUTPUTS=yes`, it's cool. And my env is Python 3.12, so the `_Outcome.errors` has changed to `_Outcome.result.errors`, so I do a minor enhancement to make it work in my env. @ydshieh Pls help review, thx very much.
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41429/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41429/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41428
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41428/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41428/comments
https://api.github.com/repos/huggingface/transformers/issues/41428/events
https://github.com/huggingface/transformers/pull/41428
3,492,946,448
PR_kwDOCUB6oc6skm08
41,428
enable some falcon-mamba uts on xpu
{ "login": "yao-matrix", "id": 7245027, "node_id": "MDQ6VXNlcjcyNDUwMjc=", "avatar_url": "https://avatars.githubusercontent.com/u/7245027?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yao-matrix", "html_url": "https://github.com/yao-matrix", "followers_url": "https://api.github.com/users/yao-matrix/followers", "following_url": "https://api.github.com/users/yao-matrix/following{/other_user}", "gists_url": "https://api.github.com/users/yao-matrix/gists{/gist_id}", "starred_url": "https://api.github.com/users/yao-matrix/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yao-matrix/subscriptions", "organizations_url": "https://api.github.com/users/yao-matrix/orgs", "repos_url": "https://api.github.com/users/yao-matrix/repos", "events_url": "https://api.github.com/users/yao-matrix/events{/privacy}", "received_events_url": "https://api.github.com/users/yao-matrix/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T20:35:22
2025-10-08T17:37:45
2025-10-08T16:48:05
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41428", "html_url": "https://github.com/huggingface/transformers/pull/41428", "diff_url": "https://github.com/huggingface/transformers/pull/41428.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41428.patch", "merged_at": "2025-10-08T16:48:05" }
enable some falcon-mamba uts on xpu. @ydshieh @SunMarc pls help review, thx very much
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41428/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41428/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41427
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41427/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41427/comments
https://api.github.com/repos/huggingface/transformers/issues/41427/events
https://github.com/huggingface/transformers/pull/41427
3,492,458,041
PR_kwDOCUB6oc6si9r6
41,427
[`FA`] Cleanup loading logic
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-07T17:51:09
2025-10-20T14:17:18
null
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41427", "html_url": "https://github.com/huggingface/transformers/pull/41427", "diff_url": "https://github.com/huggingface/transformers/pull/41427.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41427.patch", "merged_at": null }
We currently ignore anything passed through the interface and forced correct loading at attn setting time. This changes it to be more flexible where power users may customly switch it up by just setting it in the config.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41427/reactions", "total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41427/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41426
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41426/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41426/comments
https://api.github.com/repos/huggingface/transformers/issues/41426/events
https://github.com/huggingface/transformers/pull/41426
3,492,351,252
PR_kwDOCUB6oc6simON
41,426
fix gemma3n case failure
{ "login": "yao-matrix", "id": 7245027, "node_id": "MDQ6VXNlcjcyNDUwMjc=", "avatar_url": "https://avatars.githubusercontent.com/u/7245027?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yao-matrix", "html_url": "https://github.com/yao-matrix", "followers_url": "https://api.github.com/users/yao-matrix/followers", "following_url": "https://api.github.com/users/yao-matrix/following{/other_user}", "gists_url": "https://api.github.com/users/yao-matrix/gists{/gist_id}", "starred_url": "https://api.github.com/users/yao-matrix/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yao-matrix/subscriptions", "organizations_url": "https://api.github.com/users/yao-matrix/orgs", "repos_url": "https://api.github.com/users/yao-matrix/repos", "events_url": "https://api.github.com/users/yao-matrix/events{/privacy}", "received_events_url": "https://api.github.com/users/yao-matrix/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T17:21:58
2025-10-10T16:25:05
2025-10-10T09:15:28
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41426", "html_url": "https://github.com/huggingface/transformers/pull/41426", "diff_url": "https://github.com/huggingface/transformers/pull/41426.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41426.patch", "merged_at": "2025-10-10T09:15:28" }
# Problem `pytest -rA tests/models/gemma3n/test_modeling_gemma3n.py::Gemma3nIntegrationTest::test_generation_beyond_sliding_window_with_generation_config` generate different output even `do_sample=False` in `generation_config = GenerationConfig(max_new_tokens=20, do_sample=False)`, which makes UT fails # Root cause custom generation_config will be overode by model generation_config even it explicitly set, when the set value equals global default. # Fix Add a condition to make the overriding only happens when the custom_generation_config doesn't explicitly set the field. @SunMarc , pls help review, thx very much.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41426/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41426/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41425
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41425/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41425/comments
https://api.github.com/repos/huggingface/transformers/issues/41425/events
https://github.com/huggingface/transformers/pull/41425
3,492,319,152
PR_kwDOCUB6oc6sifK3
41,425
Remove deprecation warning
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T17:13:13
2025-10-07T17:27:34
2025-10-07T17:21:14
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41425", "html_url": "https://github.com/huggingface/transformers/pull/41425", "diff_url": "https://github.com/huggingface/transformers/pull/41425.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41425.patch", "merged_at": "2025-10-07T17:21:14" }
# What does this PR do? As per the title
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41425/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41425/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41424
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41424/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41424/comments
https://api.github.com/repos/huggingface/transformers/issues/41424/events
https://github.com/huggingface/transformers/pull/41424
3,492,173,335
PR_kwDOCUB6oc6siADe
41,424
Create pull_request_template.md
{ "login": "haleema-khatun", "id": 187542235, "node_id": "U_kgDOCy2q2w", "avatar_url": "https://avatars.githubusercontent.com/u/187542235?v=4", "gravatar_id": "", "url": "https://api.github.com/users/haleema-khatun", "html_url": "https://github.com/haleema-khatun", "followers_url": "https://api.github.com/users/haleema-khatun/followers", "following_url": "https://api.github.com/users/haleema-khatun/following{/other_user}", "gists_url": "https://api.github.com/users/haleema-khatun/gists{/gist_id}", "starred_url": "https://api.github.com/users/haleema-khatun/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/haleema-khatun/subscriptions", "organizations_url": "https://api.github.com/users/haleema-khatun/orgs", "repos_url": "https://api.github.com/users/haleema-khatun/repos", "events_url": "https://api.github.com/users/haleema-khatun/events{/privacy}", "received_events_url": "https://api.github.com/users/haleema-khatun/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-07T16:26:38
2025-10-07T16:26:38
null
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41424", "html_url": "https://github.com/huggingface/transformers/pull/41424", "diff_url": "https://github.com/huggingface/transformers/pull/41424.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41424.patch", "merged_at": null }
# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41424/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41424/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41423
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41423/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41423/comments
https://api.github.com/repos/huggingface/transformers/issues/41423/events
https://github.com/huggingface/transformers/pull/41423
3,492,114,440
PR_kwDOCUB6oc6shziC
41,423
[`JetMoe`] Fix KV head repetition and padding free
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T16:09:10
2025-10-08T12:27:26
2025-10-08T12:27:22
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41423", "html_url": "https://github.com/huggingface/transformers/pull/41423", "diff_url": "https://github.com/huggingface/transformers/pull/41423.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41423.patch", "merged_at": "2025-10-08T12:27:22" }
During the vllm refactor 2 small issues have been introduced - Jetmoe doesn't repeat the kv heads with `repeat_interleave` but uses `repeat`, i.e. different patterns - Position ids wasn't passed down leading to padding free not working Should hopefully align the tests again in #41377 (at least checked locally on our dgx machines that the outputs are matching again) cc @ydshieh
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41423/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 1, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41423/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41422
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41422/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41422/comments
https://api.github.com/repos/huggingface/transformers/issues/41422/events
https://github.com/huggingface/transformers/pull/41422
3,492,109,354
PR_kwDOCUB6oc6shyd4
41,422
Fix tests fsdp
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 8103865784, "node_id": "LA_kwDOCUB6oc8AAAAB4wctuA", "url": "https://api.github.com/repos/huggingface/transformers/labels/for%20patch", "name": "for patch", "color": "D93F0B", "default": false, "description": "Tag issues / labels that should be included in the next patch" } ]
closed
false
null
[]
null
[]
2025-10-07T16:07:31
2025-10-09T12:09:54
2025-10-09T12:09:52
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41422", "html_url": "https://github.com/huggingface/transformers/pull/41422", "diff_url": "https://github.com/huggingface/transformers/pull/41422.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41422.patch", "merged_at": "2025-10-09T12:09:52" }
# What does this PR do? This PR fixes all fsdp tests as there were some regressions: - fsdp typing changed so hf parser had issues - `fused_adamw_torch` (new default for users with torch >=2.8) + fp16 + FSDP don't work cc @winglian @qgallouedec did you face this issue ? - Fix cmd by using `fsdp_config`, this was broken for way too long ## Run tests with 2 gpus `RUN_SLOW=True pytest tests/fsdp/test_fsdp.py`
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41422/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41422/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41421
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41421/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41421/comments
https://api.github.com/repos/huggingface/transformers/issues/41421/events
https://github.com/huggingface/transformers/pull/41421
3,492,096,025
PR_kwDOCUB6oc6shvm_
41,421
Restore cuda graphs to continuous batching
{ "login": "remi-or", "id": 83456801, "node_id": "MDQ6VXNlcjgzNDU2ODAx", "avatar_url": "https://avatars.githubusercontent.com/u/83456801?v=4", "gravatar_id": "", "url": "https://api.github.com/users/remi-or", "html_url": "https://github.com/remi-or", "followers_url": "https://api.github.com/users/remi-or/followers", "following_url": "https://api.github.com/users/remi-or/following{/other_user}", "gists_url": "https://api.github.com/users/remi-or/gists{/gist_id}", "starred_url": "https://api.github.com/users/remi-or/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/remi-or/subscriptions", "organizations_url": "https://api.github.com/users/remi-or/orgs", "repos_url": "https://api.github.com/users/remi-or/repos", "events_url": "https://api.github.com/users/remi-or/events{/privacy}", "received_events_url": "https://api.github.com/users/remi-or/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T16:03:25
2025-10-13T09:57:57
2025-10-13T09:57:57
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41421", "html_url": "https://github.com/huggingface/transformers/pull/41421", "diff_url": "https://github.com/huggingface/transformers/pull/41421.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41421.patch", "merged_at": "2025-10-13T09:57:57" }
This PR restores cuda graphs in continuous batching. The main changes associated with this are: 1. the logic of how to generate tokens have been moved to the CB processor, which also handles the cuda graphs 2. the generation step automatically slices the tensors to remove all padding unless cuda graphs are activated 3. cuda graphs are captured on padded shapes, which is 25%, 50%, 75% or 100% of the queries axis and 1/8, ... 8/8 of the keys values axis, to strike a balance between the amount of padding and the quantity of cuda graphs Documentation is kind of lacking but will be added in next commits, I am opening the PR so @ArthurZucker can test stuff out - [x] Add more documentation - [x] Test it out and add performance numbers with / without on AMD / Nvidia with the three main attn implementations
{ "login": "remi-or", "id": 83456801, "node_id": "MDQ6VXNlcjgzNDU2ODAx", "avatar_url": "https://avatars.githubusercontent.com/u/83456801?v=4", "gravatar_id": "", "url": "https://api.github.com/users/remi-or", "html_url": "https://github.com/remi-or", "followers_url": "https://api.github.com/users/remi-or/followers", "following_url": "https://api.github.com/users/remi-or/following{/other_user}", "gists_url": "https://api.github.com/users/remi-or/gists{/gist_id}", "starred_url": "https://api.github.com/users/remi-or/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/remi-or/subscriptions", "organizations_url": "https://api.github.com/users/remi-or/orgs", "repos_url": "https://api.github.com/users/remi-or/repos", "events_url": "https://api.github.com/users/remi-or/events{/privacy}", "received_events_url": "https://api.github.com/users/remi-or/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41421/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 1 }
https://api.github.com/repos/huggingface/transformers/issues/41421/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41420
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41420/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41420/comments
https://api.github.com/repos/huggingface/transformers/issues/41420/events
https://github.com/huggingface/transformers/pull/41420
3,492,092,306
PR_kwDOCUB6oc6shu1i
41,420
[Qwen3VLMoe] Fixed: Expected self.dtype to be equal to src.dtype - routing_weights casting
{ "login": "danielquintas8", "id": 72402095, "node_id": "MDQ6VXNlcjcyNDAyMDk1", "avatar_url": "https://avatars.githubusercontent.com/u/72402095?v=4", "gravatar_id": "", "url": "https://api.github.com/users/danielquintas8", "html_url": "https://github.com/danielquintas8", "followers_url": "https://api.github.com/users/danielquintas8/followers", "following_url": "https://api.github.com/users/danielquintas8/following{/other_user}", "gists_url": "https://api.github.com/users/danielquintas8/gists{/gist_id}", "starred_url": "https://api.github.com/users/danielquintas8/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/danielquintas8/subscriptions", "organizations_url": "https://api.github.com/users/danielquintas8/orgs", "repos_url": "https://api.github.com/users/danielquintas8/repos", "events_url": "https://api.github.com/users/danielquintas8/events{/privacy}", "received_events_url": "https://api.github.com/users/danielquintas8/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T16:02:14
2025-10-14T11:15:15
2025-10-14T11:14:49
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41420", "html_url": "https://github.com/huggingface/transformers/pull/41420", "diff_url": "https://github.com/huggingface/transformers/pull/41420.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41420.patch", "merged_at": "2025-10-14T11:14:49" }
## What does this PR do? Fixed: Expected self.dtype to be equal to src.dtype - routing_weights casting ## Related issue Fixes [#41418](https://github.com/huggingface/transformers/issues/41418)
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41420/reactions", "total_count": 8, "+1": 8, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41420/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41419
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41419/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41419/comments
https://api.github.com/repos/huggingface/transformers/issues/41419/events
https://github.com/huggingface/transformers/pull/41419
3,492,047,320
PR_kwDOCUB6oc6shlOf
41,419
First QAT for Finegrained FP8
{ "login": "MekkCyber", "id": 93391238, "node_id": "U_kgDOBZEJhg", "avatar_url": "https://avatars.githubusercontent.com/u/93391238?v=4", "gravatar_id": "", "url": "https://api.github.com/users/MekkCyber", "html_url": "https://github.com/MekkCyber", "followers_url": "https://api.github.com/users/MekkCyber/followers", "following_url": "https://api.github.com/users/MekkCyber/following{/other_user}", "gists_url": "https://api.github.com/users/MekkCyber/gists{/gist_id}", "starred_url": "https://api.github.com/users/MekkCyber/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/MekkCyber/subscriptions", "organizations_url": "https://api.github.com/users/MekkCyber/orgs", "repos_url": "https://api.github.com/users/MekkCyber/repos", "events_url": "https://api.github.com/users/MekkCyber/events{/privacy}", "received_events_url": "https://api.github.com/users/MekkCyber/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-07T15:48:18
2025-10-07T15:58:28
null
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41419", "html_url": "https://github.com/huggingface/transformers/pull/41419", "diff_url": "https://github.com/huggingface/transformers/pull/41419.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41419.patch", "merged_at": null }
# What does this PR do? Adds QAT support for finegrained FP8, simply calling `prepare_model` should prepare the model for qat: ``` quantization_config = FineGrainedFP8Config( weight_block_size=(128, 128), ) model = AutoModelForMaskedLM.from_pretrained(model_name, device_map="auto", dtype=torch.bfloat16) quantizer = FineGrainedFP8HfQuantizer(quantization_config) model = quantizer.prepare_model(model) ```
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41419/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41419/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41418
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41418/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41418/comments
https://api.github.com/repos/huggingface/transformers/issues/41418/events
https://github.com/huggingface/transformers/issues/41418
3,492,032,779
I_kwDOCUB6oc7QJDEL
41,418
Qwen3 VL Moe: Expected self.dtype to be equal to src.dtype
{ "login": "danielquintas8", "id": 72402095, "node_id": "MDQ6VXNlcjcyNDAyMDk1", "avatar_url": "https://avatars.githubusercontent.com/u/72402095?v=4", "gravatar_id": "", "url": "https://api.github.com/users/danielquintas8", "html_url": "https://github.com/danielquintas8", "followers_url": "https://api.github.com/users/danielquintas8/followers", "following_url": "https://api.github.com/users/danielquintas8/following{/other_user}", "gists_url": "https://api.github.com/users/danielquintas8/gists{/gist_id}", "starred_url": "https://api.github.com/users/danielquintas8/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/danielquintas8/subscriptions", "organizations_url": "https://api.github.com/users/danielquintas8/orgs", "repos_url": "https://api.github.com/users/danielquintas8/repos", "events_url": "https://api.github.com/users/danielquintas8/events{/privacy}", "received_events_url": "https://api.github.com/users/danielquintas8/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 3817266200, "node_id": "MDU6TGFiZWwzODE3MjY2MjAw", "url": "https://api.github.com/repos/huggingface/transformers/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": null }, { "id": 5769473378, "node_id": "LA_kwDOCUB6oc8AAAABV-MtYg", "url": "https://api.github.com/repos/huggingface/transformers/labels/Vision", "name": "Vision", "color": "C079EF", "default": false, "description": "" }, { "id": 7510456769, "node_id": "LA_kwDOCUB6oc8AAAABv6h5wQ", "url": "https://api.github.com/repos/huggingface/transformers/labels/Mixture%20of%20Experts", "name": "Mixture of Experts", "color": "DDB5D0", "default": false, "description": "" } ]
closed
false
null
[]
null
[]
2025-10-07T15:44:23
2025-10-14T11:14:50
2025-10-14T11:14:50
CONTRIBUTOR
null
null
null
null
### System Info - `transformers` version: 4.57.0 - Platform: Linux-5.14.0-452.el9.x86_64-x86_64-with-glibc2.34 - Python version: 3.12.11 - Huggingface_hub version: 0.35.3 - Safetensors version: 0.6.2 - Accelerate version: 1.10.1 - Accelerate config: not found - DeepSpeed version: 0.17.6 - PyTorch version (accelerator?): 2.8.0+cu128 (CUDA) - Tensorflow version (GPU?): not installed (NA) - Flax version (CPU?/GPU?/TPU?): not installed (NA) - Jax version: not installed - JaxLib version: not installed - Using distributed or parallel set-up in script?: Yes - Using GPU in script?: Yes - GPU type: NVIDIA A100 80GB PCIe ### Who can help? _No response_ ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below) ### Reproduction ```python import os import torch from transformers import EarlyStoppingCallback, Qwen3VLMoeForConditionalGeneration, AutoProcessor from trl import SFTTrainer, SFTConfig from peft import LoraConfig from datasets import load_dataset # Prepare dataset def formatting_prompts_func(examples): images = [[image] for image in examples["image"]] prompt_template =[ { "role": "system", "content": [ { "type": "text", "text": """You are an helpful assistant""" }, ] }, { "role": "user", "content": [ { "type": "image", }, { "type": "text", "text": "Process the provided images." } ] } ] completion_template = [ { "role": "assistant", "content": [ { "type": "text", "text": "example" ## just a text field to simulate } ] } ] prompts = [prompt_template] * len(examples) completions = [completion_template] * len(examples) return {"images": images, "prompt": prompts, "completion": completions} epochs = 15 learning_rate = 1e-4 per_device_batch_size = 1 gradient_accumulation_steps = 2 weight_decay = 0.01 warmup_steps = 0 lora_rank = 8 adapter_path = f"adapters/reproduce" output_path = f"{adapter_path}/outputs/" # Load tokenizer model_path = "Qwen/Qwen3-VL-30B-A3B-Instruct" processor = AutoProcessor.from_pretrained(model_path) model = Qwen3VLMoeForConditionalGeneration.from_pretrained( model_path, dtype=torch.bfloat16, ) training_args = SFTConfig( max_length=None, # padding="longest", seed=3407, output_dir=output_path, num_train_epochs=epochs, learning_rate=learning_rate, lr_scheduler_type="cosine", per_device_train_batch_size=per_device_batch_size, gradient_accumulation_steps=gradient_accumulation_steps, per_device_eval_batch_size=1, eval_accumulation_steps=1, warmup_steps=warmup_steps, logging_steps=10, logging_strategy="steps", do_eval=True, bf16_full_eval=True, eval_strategy="epoch", save_strategy="epoch", packing=False, fp16=not torch.cuda.is_bf16_supported(), bf16=torch.cuda.is_bf16_supported(), gradient_checkpointing=True, weight_decay=weight_decay, load_best_model_at_end=True, metric_for_best_model="eval_loss", greater_is_better=False, completion_only_loss=True, eos_token=processor.tokenizer.eos_token, ) # LoRA configuration; peft_config = LoraConfig( r=lora_rank, lora_alpha=lora_rank * 2, lora_dropout=0.05, bias="none", task_type="CAUSAL_LM", target_modules=[ "self_attn.q_proj", "self_attn.k_proj", "self_attn.v_proj", "self_attn.o_proj", ], ) model.enable_input_require_grads() # Load dataset dataset = load_dataset("huggingface/documentation-images", split="train[:1]") ## just an image/text dataset to simulate formatted_dataset = dataset.map( formatting_prompts_func, batched=True, remove_columns=dataset.column_names, ) # Initialize trainer trainer = SFTTrainer( model=model, peft_config=peft_config, processing_class=processor, args=training_args, train_dataset=formatted_dataset, eval_dataset=formatted_dataset, ) trainer.train() trainer.save_model(output_dir=adapter_path) ``` ```bash $ accelerate launch --mixed_precision bf16 --use_fsdp --fsdp_sharding_strategy 1 --fsdp_backward_prefetch NO_PREFETCH --fsdp_offload_params true --fsdp_auto_wrap_policy TRANSFORMER_BASED_WRAP --fsdp_transformer_layer_cls_to_wrap Qwen3VLMoeVisionBlock,Qwen3VLMoeTextDecoderLayer --num_machines 1 --num_processes 4 main.py ``` ``` [rank0]: Traceback (most recent call last): [rank0]: File "/home/quintas/workspace/reproduce.py", line 142, in <module> [rank0]: # accelerate launch --mixed_precision bf16 --use_fsdp --fsdp_sharding_strategy 1 --fsdp_backward_prefetch NO_PREFETCH --fsdp_offload_params true --fsdp_auto_wrap_policy TRANSFORMER_BASED_WRAP --fsdp_transformer_layer_cls_to_wrap Qwen3VLMoeVisionBlock,Qwen3VLMoeTextDecoderLayer --num_machines 1 --num_processes 4 Qwen3-VL-30B-A3B-Instruct/adapters/invoice_reading/train.py [rank0]: ^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 2325, in train [rank0]: return inner_training_loop( [rank0]: ^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 2790, in _inner_training_loop [rank0]: self._maybe_log_save_evaluate( [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 3221, in _maybe_log_save_evaluate [rank0]: metrics = self._evaluate(trial, ignore_keys_for_eval) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 3170, in _evaluate [rank0]: metrics = self.evaluate(ignore_keys=ignore_keys_for_eval) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 4489, in evaluate [rank0]: output = eval_loop( [rank0]: ^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 4685, in evaluation_loop [rank0]: losses, logits, labels = self.prediction_step(model, inputs, prediction_loss_only, ignore_keys=ignore_keys) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 4902, in prediction_step [rank0]: loss, outputs = self.compute_loss( [rank0]: ^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/trl/trainer/sft_trainer.py", line 1096, in compute_loss [rank0]: (loss, outputs) = super().compute_loss( [rank0]: ^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/trainer.py", line 4110, in compute_loss [rank0]: outputs = model(**inputs) [rank0]: ^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/accelerate/utils/operations.py", line 818, in forward [rank0]: return model_forward(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/accelerate/utils/operations.py", line 806, in __call__ [rank0]: return convert_to_fp32(self.model_forward(*args, **kwargs)) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/amp/autocast_mode.py", line 44, in decorate_autocast [rank0]: return func(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py", line 854, in forward [rank0]: output = self._fsdp_wrapped_module(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/accelerate/utils/operations.py", line 818, in forward [rank0]: return model_forward(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/accelerate/utils/operations.py", line 806, in __call__ [rank0]: return convert_to_fp32(self.model_forward(*args, **kwargs)) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/amp/autocast_mode.py", line 44, in decorate_autocast [rank0]: return func(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/peft/peft_model.py", line 1850, in forward [rank0]: return self.base_model( [rank0]: ^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/peft/tuners/tuners_utils.py", line 222, in forward [rank0]: return self.model.forward(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/utils/generic.py", line 1064, in wrapper [rank0]: outputs = func(self, *args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", line 1601, in forward [rank0]: outputs = self.model( [rank0]: ^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/utils/generic.py", line 1064, in wrapper [rank0]: outputs = func(self, *args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", line 1389, in forward [rank0]: outputs = self.language_model( [rank0]: ^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/utils/generic.py", line 1064, in wrapper [rank0]: outputs = func(self, *args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", line 962, in forward [rank0]: layer_outputs = decoder_layer( [rank0]: ^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py", line 854, in forward [rank0]: output = self._fsdp_wrapped_module(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/modeling_layers.py", line 94, in __call__ [rank0]: return super().__call__(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/utils/deprecation.py", line 172, in wrapped_func [rank0]: return func(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", line 391, in forward [rank0]: hidden_states = self.mlp(hidden_states) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/home/quintas/miniconda3/envs/Comudel/lib/python3.12/site-packages/transformers/models/qwen3_vl_moe/modeling_qwen3_vl_moe.py", line 149, in forward [rank0]: router_weights = torch.zeros_like(router_logits).scatter_(1, router_indices, routing_weights) [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: RuntimeError: scatter(): Expected self.dtype to be equal to src.dtype ``` ### Expected behavior Train successful
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41418/reactions", "total_count": 9, "+1": 9, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41418/timeline
null
completed
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
true
https://api.github.com/repos/huggingface/transformers/issues/41417
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41417/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41417/comments
https://api.github.com/repos/huggingface/transformers/issues/41417/events
https://github.com/huggingface/transformers/pull/41417
3,492,010,898
PR_kwDOCUB6oc6shdMn
41,417
🚨 [v5] Prune `prune_heads`
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T15:38:47
2025-10-08T09:25:17
2025-10-08T09:25:13
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41417", "html_url": "https://github.com/huggingface/transformers/pull/41417", "diff_url": "https://github.com/huggingface/transformers/pull/41417.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41417.patch", "merged_at": "2025-10-08T09:25:13" }
# What does this PR do? Spiritual successor to #41076 -- removes `prune_heads` and related code/tests/docs. `prune_heads` was a feature that we stopped including in models a while ago. For instance, the `test_head_pruning_save_load_from_config_init` test passes in 28 model classes, but is skipped in 444.
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41417/reactions", "total_count": 2, "+1": 0, "-1": 0, "laugh": 0, "hooray": 2, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41417/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41416
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41416/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41416/comments
https://api.github.com/repos/huggingface/transformers/issues/41416/events
https://github.com/huggingface/transformers/pull/41416
3,491,992,004
PR_kwDOCUB6oc6shZHm
41,416
Stop collecting all model parameters to save models when using DeepSpeed and LoRA
{ "login": "Taise228", "id": 95762401, "node_id": "U_kgDOBbU34Q", "avatar_url": "https://avatars.githubusercontent.com/u/95762401?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Taise228", "html_url": "https://github.com/Taise228", "followers_url": "https://api.github.com/users/Taise228/followers", "following_url": "https://api.github.com/users/Taise228/following{/other_user}", "gists_url": "https://api.github.com/users/Taise228/gists{/gist_id}", "starred_url": "https://api.github.com/users/Taise228/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Taise228/subscriptions", "organizations_url": "https://api.github.com/users/Taise228/orgs", "repos_url": "https://api.github.com/users/Taise228/repos", "events_url": "https://api.github.com/users/Taise228/events{/privacy}", "received_events_url": "https://api.github.com/users/Taise228/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-07T15:32:40
2025-10-07T15:32:40
null
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41416", "html_url": "https://github.com/huggingface/transformers/pull/41416", "diff_url": "https://github.com/huggingface/transformers/pull/41416.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41416.patch", "merged_at": null }
# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes #41344 ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [x] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. @SunMarc (Let me tag you since you replied to the issue. Thanks for taking time!) <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41416/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41416/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41415
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41415/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41415/comments
https://api.github.com/repos/huggingface/transformers/issues/41415/events
https://github.com/huggingface/transformers/pull/41415
3,491,982,532
PR_kwDOCUB6oc6shXA9
41,415
Fix bnb fsdp loading for pre-quantized checkpoint
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 8103865784, "node_id": "LA_kwDOCUB6oc8AAAAB4wctuA", "url": "https://api.github.com/repos/huggingface/transformers/labels/for%20patch", "name": "for patch", "color": "D93F0B", "default": false, "description": "Tag issues / labels that should be included in the next patch" } ]
closed
false
null
[]
null
[]
2025-10-07T15:30:31
2025-10-09T16:05:38
2025-10-09T16:05:36
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41415", "html_url": "https://github.com/huggingface/transformers/pull/41415", "diff_url": "https://github.com/huggingface/transformers/pull/41415.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41415.patch", "merged_at": "2025-10-09T16:05:35" }
# What does this PR do? This PR fixes bnb loading when using FSDP for pre-quantized checkpoints. This happened because we changed how we load quantized checkpoints as we need to cache all the quantized stats before creating the quantized weight.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41415/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41415/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41414
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41414/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41414/comments
https://api.github.com/repos/huggingface/transformers/issues/41414/events
https://github.com/huggingface/transformers/pull/41414
3,491,815,425
PR_kwDOCUB6oc6sgzK_
41,414
Fixing comments in __init__ file
{ "login": "MekkCyber", "id": 93391238, "node_id": "U_kgDOBZEJhg", "avatar_url": "https://avatars.githubusercontent.com/u/93391238?v=4", "gravatar_id": "", "url": "https://api.github.com/users/MekkCyber", "html_url": "https://github.com/MekkCyber", "followers_url": "https://api.github.com/users/MekkCyber/followers", "following_url": "https://api.github.com/users/MekkCyber/following{/other_user}", "gists_url": "https://api.github.com/users/MekkCyber/gists{/gist_id}", "starred_url": "https://api.github.com/users/MekkCyber/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/MekkCyber/subscriptions", "organizations_url": "https://api.github.com/users/MekkCyber/orgs", "repos_url": "https://api.github.com/users/MekkCyber/repos", "events_url": "https://api.github.com/users/MekkCyber/events{/privacy}", "received_events_url": "https://api.github.com/users/MekkCyber/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T14:47:54
2025-10-08T10:07:29
2025-10-08T10:07:27
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41414", "html_url": "https://github.com/huggingface/transformers/pull/41414", "diff_url": "https://github.com/huggingface/transformers/pull/41414.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41414.patch", "merged_at": "2025-10-08T10:07:27" }
# What does this PR do? Fixes comments in global __init__ file
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41414/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41414/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41413
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41413/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41413/comments
https://api.github.com/repos/huggingface/transformers/issues/41413/events
https://github.com/huggingface/transformers/pull/41413
3,491,754,189
PR_kwDOCUB6oc6sgl8U
41,413
:facepalm: CB nit!
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T14:31:37
2025-10-10T11:08:04
2025-10-08T11:36:27
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41413", "html_url": "https://github.com/huggingface/transformers/pull/41413", "diff_url": "https://github.com/huggingface/transformers/pull/41413.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41413.patch", "merged_at": "2025-10-08T11:36:27" }
# What does this PR do? Fix a small typo
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41413/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41413/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41412
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41412/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41412/comments
https://api.github.com/repos/huggingface/transformers/issues/41412/events
https://github.com/huggingface/transformers/pull/41412
3,491,647,903
PR_kwDOCUB6oc6sgO_3
41,412
[Parakeet] unnecessary warning & auto mapping
{ "login": "eustlb", "id": 94853470, "node_id": "U_kgDOBadZXg", "avatar_url": "https://avatars.githubusercontent.com/u/94853470?v=4", "gravatar_id": "", "url": "https://api.github.com/users/eustlb", "html_url": "https://github.com/eustlb", "followers_url": "https://api.github.com/users/eustlb/followers", "following_url": "https://api.github.com/users/eustlb/following{/other_user}", "gists_url": "https://api.github.com/users/eustlb/gists{/gist_id}", "starred_url": "https://api.github.com/users/eustlb/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/eustlb/subscriptions", "organizations_url": "https://api.github.com/users/eustlb/orgs", "repos_url": "https://api.github.com/users/eustlb/repos", "events_url": "https://api.github.com/users/eustlb/events{/privacy}", "received_events_url": "https://api.github.com/users/eustlb/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T14:04:48
2025-10-10T09:00:15
2025-10-10T09:00:15
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41412", "html_url": "https://github.com/huggingface/transformers/pull/41412", "diff_url": "https://github.com/huggingface/transformers/pull/41412.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41412.patch", "merged_at": "2025-10-10T09:00:15" }
# What does this PR do? - fixes #41387 - and fixes warning ``` 🚨 Config not found for parakeet. You can manually add it to HARDCODED_CONFIG_FOR_MODELS in utils/auto_docstring.py ``` - and fixes auto tokenizer mapping that was not updated when moving from `ParakeetCTCTokenizer` to `ParakeetTokenizerFast` --- What triggers the unexpected warning: `@auto_doctring` on a class (not on a method) (here `ParakeetPreTrainedModel`, `ParakeetEncoder`, `ParakeetForCTC`) triggers `auto_method_docstring` call **with parent_class kwargs setted**, that will trigger looking for a config class with the model name `parakeet` (extracted from `modeling_parakeet.py`). Yet we don't have a config class for `parakeet` alone, we have `ParakeetEncoderConfig` and `ParakeetCTCConfig`. To solve, let's add `"parakeet": "ParakeetCTCConfig"` to `HARDCODED_CONFIG_FOR_MODELS`. >[!NOTE] > When other Parakeet variants will be integrated `ParakeetForTDT` and `ParakeetForRNTT`, then we might want to add a global ParakeetConfig that encapsulate them all
{ "login": "eustlb", "id": 94853470, "node_id": "U_kgDOBadZXg", "avatar_url": "https://avatars.githubusercontent.com/u/94853470?v=4", "gravatar_id": "", "url": "https://api.github.com/users/eustlb", "html_url": "https://github.com/eustlb", "followers_url": "https://api.github.com/users/eustlb/followers", "following_url": "https://api.github.com/users/eustlb/following{/other_user}", "gists_url": "https://api.github.com/users/eustlb/gists{/gist_id}", "starred_url": "https://api.github.com/users/eustlb/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/eustlb/subscriptions", "organizations_url": "https://api.github.com/users/eustlb/orgs", "repos_url": "https://api.github.com/users/eustlb/repos", "events_url": "https://api.github.com/users/eustlb/events{/privacy}", "received_events_url": "https://api.github.com/users/eustlb/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41412/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41412/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41411
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41411/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41411/comments
https://api.github.com/repos/huggingface/transformers/issues/41411/events
https://github.com/huggingface/transformers/pull/41411
3,491,638,874
PR_kwDOCUB6oc6sgNED
41,411
Small Fix for imports
{ "login": "MekkCyber", "id": 93391238, "node_id": "U_kgDOBZEJhg", "avatar_url": "https://avatars.githubusercontent.com/u/93391238?v=4", "gravatar_id": "", "url": "https://api.github.com/users/MekkCyber", "html_url": "https://github.com/MekkCyber", "followers_url": "https://api.github.com/users/MekkCyber/followers", "following_url": "https://api.github.com/users/MekkCyber/following{/other_user}", "gists_url": "https://api.github.com/users/MekkCyber/gists{/gist_id}", "starred_url": "https://api.github.com/users/MekkCyber/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/MekkCyber/subscriptions", "organizations_url": "https://api.github.com/users/MekkCyber/orgs", "repos_url": "https://api.github.com/users/MekkCyber/repos", "events_url": "https://api.github.com/users/MekkCyber/events{/privacy}", "received_events_url": "https://api.github.com/users/MekkCyber/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T14:02:22
2025-10-27T12:21:41
2025-10-20T15:21:04
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41411", "html_url": "https://github.com/huggingface/transformers/pull/41411", "diff_url": "https://github.com/huggingface/transformers/pull/41411.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41411.patch", "merged_at": "2025-10-20T15:21:04" }
# What does this PR do? When using `importlib.util.find_spec(pkg_name)`, this might return some packages that are not importable but still exist. In this pr we simply check if they can be loaded
{ "login": "MekkCyber", "id": 93391238, "node_id": "U_kgDOBZEJhg", "avatar_url": "https://avatars.githubusercontent.com/u/93391238?v=4", "gravatar_id": "", "url": "https://api.github.com/users/MekkCyber", "html_url": "https://github.com/MekkCyber", "followers_url": "https://api.github.com/users/MekkCyber/followers", "following_url": "https://api.github.com/users/MekkCyber/following{/other_user}", "gists_url": "https://api.github.com/users/MekkCyber/gists{/gist_id}", "starred_url": "https://api.github.com/users/MekkCyber/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/MekkCyber/subscriptions", "organizations_url": "https://api.github.com/users/MekkCyber/orgs", "repos_url": "https://api.github.com/users/MekkCyber/repos", "events_url": "https://api.github.com/users/MekkCyber/events{/privacy}", "received_events_url": "https://api.github.com/users/MekkCyber/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41411/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41411/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41410
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41410/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41410/comments
https://api.github.com/repos/huggingface/transformers/issues/41410/events
https://github.com/huggingface/transformers/pull/41410
3,491,487,739
PR_kwDOCUB6oc6sfsoz
41,410
Remove Python 3.9 classifier
{ "login": "cyyever", "id": 17618148, "node_id": "MDQ6VXNlcjE3NjE4MTQ4", "avatar_url": "https://avatars.githubusercontent.com/u/17618148?v=4", "gravatar_id": "", "url": "https://api.github.com/users/cyyever", "html_url": "https://github.com/cyyever", "followers_url": "https://api.github.com/users/cyyever/followers", "following_url": "https://api.github.com/users/cyyever/following{/other_user}", "gists_url": "https://api.github.com/users/cyyever/gists{/gist_id}", "starred_url": "https://api.github.com/users/cyyever/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/cyyever/subscriptions", "organizations_url": "https://api.github.com/users/cyyever/orgs", "repos_url": "https://api.github.com/users/cyyever/repos", "events_url": "https://api.github.com/users/cyyever/events{/privacy}", "received_events_url": "https://api.github.com/users/cyyever/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T13:23:50
2025-10-08T13:39:26
2025-10-08T12:20:36
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41410", "html_url": "https://github.com/huggingface/transformers/pull/41410", "diff_url": "https://github.com/huggingface/transformers/pull/41410.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41410.patch", "merged_at": "2025-10-08T12:20:36" }
# What does this PR do? Follows recent clean-up.
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41410/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41410/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41409
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41409/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41409/comments
https://api.github.com/repos/huggingface/transformers/issues/41409/events
https://github.com/huggingface/transformers/pull/41409
3,491,480,337
PR_kwDOCUB6oc6sfrBb
41,409
Fix early CUDA initialisation
{ "login": "hmellor", "id": 19981378, "node_id": "MDQ6VXNlcjE5OTgxMzc4", "avatar_url": "https://avatars.githubusercontent.com/u/19981378?v=4", "gravatar_id": "", "url": "https://api.github.com/users/hmellor", "html_url": "https://github.com/hmellor", "followers_url": "https://api.github.com/users/hmellor/followers", "following_url": "https://api.github.com/users/hmellor/following{/other_user}", "gists_url": "https://api.github.com/users/hmellor/gists{/gist_id}", "starred_url": "https://api.github.com/users/hmellor/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/hmellor/subscriptions", "organizations_url": "https://api.github.com/users/hmellor/orgs", "repos_url": "https://api.github.com/users/hmellor/repos", "events_url": "https://api.github.com/users/hmellor/events{/privacy}", "received_events_url": "https://api.github.com/users/hmellor/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T13:21:43
2025-10-07T13:37:21
2025-10-07T13:37:18
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41409", "html_url": "https://github.com/huggingface/transformers/pull/41409", "diff_url": "https://github.com/huggingface/transformers/pull/41409.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41409.patch", "merged_at": "2025-10-07T13:37:18" }
This import added in https://github.com/huggingface/transformers/pull/41370 was identified to be causing CUDA to initialise early. It was identified in vLLM's Transformers nightly test https://buildkite.com/vllm/ci/builds/33777/steps/canvas?sid=0199bcd4-8cf7-442a-a94d-df2969d65bde
{ "login": "hmellor", "id": 19981378, "node_id": "MDQ6VXNlcjE5OTgxMzc4", "avatar_url": "https://avatars.githubusercontent.com/u/19981378?v=4", "gravatar_id": "", "url": "https://api.github.com/users/hmellor", "html_url": "https://github.com/hmellor", "followers_url": "https://api.github.com/users/hmellor/followers", "following_url": "https://api.github.com/users/hmellor/following{/other_user}", "gists_url": "https://api.github.com/users/hmellor/gists{/gist_id}", "starred_url": "https://api.github.com/users/hmellor/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/hmellor/subscriptions", "organizations_url": "https://api.github.com/users/hmellor/orgs", "repos_url": "https://api.github.com/users/hmellor/repos", "events_url": "https://api.github.com/users/hmellor/events{/privacy}", "received_events_url": "https://api.github.com/users/hmellor/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41409/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41409/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41408
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41408/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41408/comments
https://api.github.com/repos/huggingface/transformers/issues/41408/events
https://github.com/huggingface/transformers/pull/41408
3,491,437,750
PR_kwDOCUB6oc6sfh0j
41,408
Benchmark overhaul
{ "login": "remi-or", "id": 83456801, "node_id": "MDQ6VXNlcjgzNDU2ODAx", "avatar_url": "https://avatars.githubusercontent.com/u/83456801?v=4", "gravatar_id": "", "url": "https://api.github.com/users/remi-or", "html_url": "https://github.com/remi-or", "followers_url": "https://api.github.com/users/remi-or/followers", "following_url": "https://api.github.com/users/remi-or/following{/other_user}", "gists_url": "https://api.github.com/users/remi-or/gists{/gist_id}", "starred_url": "https://api.github.com/users/remi-or/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/remi-or/subscriptions", "organizations_url": "https://api.github.com/users/remi-or/orgs", "repos_url": "https://api.github.com/users/remi-or/repos", "events_url": "https://api.github.com/users/remi-or/events{/privacy}", "received_events_url": "https://api.github.com/users/remi-or/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
{ "login": "remi-or", "id": 83456801, "node_id": "MDQ6VXNlcjgzNDU2ODAx", "avatar_url": "https://avatars.githubusercontent.com/u/83456801?v=4", "gravatar_id": "", "url": "https://api.github.com/users/remi-or", "html_url": "https://github.com/remi-or", "followers_url": "https://api.github.com/users/remi-or/followers", "following_url": "https://api.github.com/users/remi-or/following{/other_user}", "gists_url": "https://api.github.com/users/remi-or/gists{/gist_id}", "starred_url": "https://api.github.com/users/remi-or/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/remi-or/subscriptions", "organizations_url": "https://api.github.com/users/remi-or/orgs", "repos_url": "https://api.github.com/users/remi-or/repos", "events_url": "https://api.github.com/users/remi-or/events{/privacy}", "received_events_url": "https://api.github.com/users/remi-or/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "login": "remi-or", "id": 83456801, "node_id": "MDQ6VXNlcjgzNDU2ODAx", "avatar_url": "https://avatars.githubusercontent.com/u/83456801?v=4", "gravatar_id": "", "url": "https://api.github.com/users/remi-or", "html_url": "https://github.com/remi-or", "followers_url": "https://api.github.com/users/remi-or/followers", "following_url": "https://api.github.com/users/remi-or/following{/other_user}", "gists_url": "https://api.github.com/users/remi-or/gists{/gist_id}", "starred_url": "https://api.github.com/users/remi-or/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/remi-or/subscriptions", "organizations_url": "https://api.github.com/users/remi-or/orgs", "repos_url": "https://api.github.com/users/remi-or/repos", "events_url": "https://api.github.com/users/remi-or/events{/privacy}", "received_events_url": "https://api.github.com/users/remi-or/received_events", "type": "User", "user_view_type": "public", "site_admin": false } ]
null
[]
2025-10-07T13:10:05
2025-10-14T19:41:43
2025-10-14T19:41:43
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41408", "html_url": "https://github.com/huggingface/transformers/pull/41408", "diff_url": "https://github.com/huggingface/transformers/pull/41408.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41408.patch", "merged_at": "2025-10-14T19:41:43" }
This PR overhauls the benchmarking suite that is included in transformers. The benchmarking suite is now based around three main components: - `BenchmarkingConfig` is a dataclass-like object which contains everything needed to reproduce a benchmark on the same machine: input length, generation length, whether to use `kernels` or `compile`, attention implementation, etc. (subject to name change) - `BenchmarkRunner` is the class that runs the benchmarks defined by the configs, with a given number of measurement iterations, warmup iterations, and a model-id. The runner takes care of setting up the runs in a way that ensures no run interacts with the downstream ones: the model is reloaded, the cache is emptied and the GPU memory is flushed. It also saves the results, the config, and any additional metadata needed to reproduce the benchmark, like hardware information and package versions. - The created results files, which contain enough informations to induces (to my knowledge) most of the metrics used to evaluate a model: e2e_atency, tpot, ttft, even inter-token latency. Results also include a sample of what has been generated, which is useful to check if it was gibberish. The results files are in json format and are made to be easily created from the dataclass-like objects and vice versa. For now, the new benchmarking suite replaces the `benchmark_v2` part of `transformers` but it could also overwrite the `benchmark` (v1) part. It would be good to make that decision in this PR. And update the CI workflows that rely on the current `benchmark_v2` (putting the PR in draft mode until then). An example of how to use the new benchmarking suite can be found in `run_benchmarks.py`. The format of the results file can (and may be bound to) change as we develop tools to analyze them. If there is a metric you want to see measured in `transformers`, please leave a comment before this is merged :slightly_smiling_face:
{ "login": "remi-or", "id": 83456801, "node_id": "MDQ6VXNlcjgzNDU2ODAx", "avatar_url": "https://avatars.githubusercontent.com/u/83456801?v=4", "gravatar_id": "", "url": "https://api.github.com/users/remi-or", "html_url": "https://github.com/remi-or", "followers_url": "https://api.github.com/users/remi-or/followers", "following_url": "https://api.github.com/users/remi-or/following{/other_user}", "gists_url": "https://api.github.com/users/remi-or/gists{/gist_id}", "starred_url": "https://api.github.com/users/remi-or/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/remi-or/subscriptions", "organizations_url": "https://api.github.com/users/remi-or/orgs", "repos_url": "https://api.github.com/users/remi-or/repos", "events_url": "https://api.github.com/users/remi-or/events{/privacy}", "received_events_url": "https://api.github.com/users/remi-or/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41408/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41408/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41407
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41407/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41407/comments
https://api.github.com/repos/huggingface/transformers/issues/41407/events
https://github.com/huggingface/transformers/pull/41407
3,491,327,213
PR_kwDOCUB6oc6sfKM7
41,407
Fix local start time handling in trainer.py
{ "login": "pythongiant", "id": 13624560, "node_id": "MDQ6VXNlcjEzNjI0NTYw", "avatar_url": "https://avatars.githubusercontent.com/u/13624560?v=4", "gravatar_id": "", "url": "https://api.github.com/users/pythongiant", "html_url": "https://github.com/pythongiant", "followers_url": "https://api.github.com/users/pythongiant/followers", "following_url": "https://api.github.com/users/pythongiant/following{/other_user}", "gists_url": "https://api.github.com/users/pythongiant/gists{/gist_id}", "starred_url": "https://api.github.com/users/pythongiant/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/pythongiant/subscriptions", "organizations_url": "https://api.github.com/users/pythongiant/orgs", "repos_url": "https://api.github.com/users/pythongiant/repos", "events_url": "https://api.github.com/users/pythongiant/events{/privacy}", "received_events_url": "https://api.github.com/users/pythongiant/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T12:39:44
2025-10-07T12:40:49
2025-10-07T12:40:43
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41407", "html_url": "https://github.com/huggingface/transformers/pull/41407", "diff_url": "https://github.com/huggingface/transformers/pull/41407.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41407.patch", "merged_at": null }
# What does this PR do? Fixes #40560 This PR resolves an issue in `Trainer` where continuing training from a checkpoint (`--resume_from_checkpoint`) causes unrealistically high **tokens per second** values at the beginning of resumed runs. ### Root Cause When resuming training, the `start_time` passed to `_maybe_log_save_evaluate()` originates from the *initial* training session rather than the resumed one. This leads to inflated throughput metrics since elapsed time is computed from the first session’s start instead of the current resume point. ### Fix Implemented - Reset the local timer used for performance logging when resuming from a checkpoint. - Ensured `_maybe_log_save_evaluate()` uses a localized `start_time` (`_local_start_time`) to compute logging metrics accurately for each resumed session. - Prevented accumulation of prior global step timing data that caused misleading throughput spikes. The change ensures that throughput metrics reflect only the current training run’s performance rather than cumulative global time. ### Example Reproduction Run: ```bash python examples/pytorch/language-modeling/run_clm.py \ --model_name_or_path gpt2 \ --dataset_name wikitext \ --do_train \ --output_dir ./clm_test --overwrite_output_dir ```` Then resume: ```bash python examples/pytorch/language-modeling/run_clm.py \ --model_name_or_path ./clm_test \ --resume_from_checkpoint ./clm_test \ --output_dir ./clm_test_resume ``` Before the fix, tokens/sec skyrockets at the start. After the fix, the metric remains consistent across resumed and fresh runs. --- ## Before submitting * [x] This PR fixes an existing issue (`#40560`). * [x] Discussed and reproduced via issue thread and local tests. * [x] Code updated to prevent metric carryover across resumed runs. * [x] Verified correct tokens/sec logging after resume. * [x] No docs changes needed (affects internal logging only). * [ ] New test could be added under `tests/trainer/test_trainer_utils.py`. --- ## Who can review? Tagging maintainers for `Trainer`: * @zach-huggingface * @SunMarc
{ "login": "pythongiant", "id": 13624560, "node_id": "MDQ6VXNlcjEzNjI0NTYw", "avatar_url": "https://avatars.githubusercontent.com/u/13624560?v=4", "gravatar_id": "", "url": "https://api.github.com/users/pythongiant", "html_url": "https://github.com/pythongiant", "followers_url": "https://api.github.com/users/pythongiant/followers", "following_url": "https://api.github.com/users/pythongiant/following{/other_user}", "gists_url": "https://api.github.com/users/pythongiant/gists{/gist_id}", "starred_url": "https://api.github.com/users/pythongiant/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/pythongiant/subscriptions", "organizations_url": "https://api.github.com/users/pythongiant/orgs", "repos_url": "https://api.github.com/users/pythongiant/repos", "events_url": "https://api.github.com/users/pythongiant/events{/privacy}", "received_events_url": "https://api.github.com/users/pythongiant/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41407/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41407/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41406
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41406/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41406/comments
https://api.github.com/repos/huggingface/transformers/issues/41406/events
https://github.com/huggingface/transformers/pull/41406
3,490,932,630
PR_kwDOCUB6oc6sd0EW
41,406
🚨 [v5] Remove deprecated cache classes
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-07T10:37:49
2025-10-07T12:35:54
null
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41406", "html_url": "https://github.com/huggingface/transformers/pull/41406", "diff_url": "https://github.com/huggingface/transformers/pull/41406.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41406.patch", "merged_at": null }
# What does this PR do? 🚨🚨🚨 DO NOT MERGE 🚨🚨🚨 Merging this PR will break `from transformers import Trainer` on the latest PEFT release (`0.17.1`), which in turn breaks our `make fixup` when PEFT is installed. PEFT `main` already sorted this issue ([here](https://github.com/huggingface/peft/commit/41c07f0445879da0e73ca3c682a0c359cd027f50)), but it is yet unreleased as of 7-oct-2025. cc @BenjaminBossan (chatted on slack -- when `transformers` has a concrete timeline for v5, we can best evaluate an adequate timing for a PEFT release) cc @Cyrilvallez FYI _______________________________ This PR removes the deprecated cache classes in `cache_utils.py`.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41406/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41406/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41405
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41405/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41405/comments
https://api.github.com/repos/huggingface/transformers/issues/41405/events
https://github.com/huggingface/transformers/pull/41405
3,490,890,378
PR_kwDOCUB6oc6sdq50
41,405
🚨 [v5] remove deprecated cache tuple input support
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T10:24:53
2025-10-07T12:52:21
2025-10-07T12:52:21
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41405", "html_url": "https://github.com/huggingface/transformers/pull/41405", "diff_url": "https://github.com/huggingface/transformers/pull/41405.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41405.patch", "merged_at": null }
# What does this PR do? This PR: - Removes the last remnants (🤞) of tuple cache input support, scheduled to be removed in v4.58 - Removes other minor items scheduled to be removed in v4.58, caught in the search - Updates a few incorrect `past_key_values` type hints - ⚠️ EDIT: add cache initialization in `forward` on models that were missing it
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41405/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41405/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41404
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41404/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41404/comments
https://api.github.com/repos/huggingface/transformers/issues/41404/events
https://github.com/huggingface/transformers/pull/41404
3,490,865,878
PR_kwDOCUB6oc6sdlr6
41,404
Remove deprecated args in Trainer for v5
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T10:18:35
2025-10-09T16:10:17
2025-10-09T16:10:14
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41404", "html_url": "https://github.com/huggingface/transformers/pull/41404", "diff_url": "https://github.com/huggingface/transformers/pull/41404.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41404.patch", "merged_at": "2025-10-09T16:10:14" }
# What does this PR do? This PR removes `half_precision_backend` and `include_tokens_per_second` as we move to v5.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41404/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41404/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41403
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41403/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41403/comments
https://api.github.com/repos/huggingface/transformers/issues/41403/events
https://github.com/huggingface/transformers/pull/41403
3,490,829,944
PR_kwDOCUB6oc6sdeCm
41,403
[Trainer] deprecate ray scope
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T10:09:17
2025-10-09T16:50:02
2025-10-09T16:50:00
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41403", "html_url": "https://github.com/huggingface/transformers/pull/41403", "diff_url": "https://github.com/huggingface/transformers/pull/41403.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41403.patch", "merged_at": "2025-10-09T16:50:00" }
# What does this PR do? This PR removes `ray_scope` from `TrainingArguments`. Instead, users will have to set the env var RAY_SCOPE if they want to change its value.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41403/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41403/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41402
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41402/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41402/comments
https://api.github.com/repos/huggingface/transformers/issues/41402/events
https://github.com/huggingface/transformers/pull/41402
3,490,761,333
PR_kwDOCUB6oc6sdPWr
41,402
[v5] rm `utils/tf_ops/`
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T09:49:24
2025-10-09T09:27:51
2025-10-09T09:27:48
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41402", "html_url": "https://github.com/huggingface/transformers/pull/41402", "diff_url": "https://github.com/huggingface/transformers/pull/41402.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41402.patch", "merged_at": "2025-10-09T09:27:48" }
# What does this PR do? TF-related folder that flew under the radar :) (seems unused in the lib, latest commit was 4 years ago)
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41402/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41402/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41401
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41401/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41401/comments
https://api.github.com/repos/huggingface/transformers/issues/41401/events
https://github.com/huggingface/transformers/pull/41401
3,490,705,592
PR_kwDOCUB6oc6sdDZj
41,401
[Model] Lfm2Moe
{ "login": "paulpak58", "id": 52512091, "node_id": "MDQ6VXNlcjUyNTEyMDkx", "avatar_url": "https://avatars.githubusercontent.com/u/52512091?v=4", "gravatar_id": "", "url": "https://api.github.com/users/paulpak58", "html_url": "https://github.com/paulpak58", "followers_url": "https://api.github.com/users/paulpak58/followers", "following_url": "https://api.github.com/users/paulpak58/following{/other_user}", "gists_url": "https://api.github.com/users/paulpak58/gists{/gist_id}", "starred_url": "https://api.github.com/users/paulpak58/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/paulpak58/subscriptions", "organizations_url": "https://api.github.com/users/paulpak58/orgs", "repos_url": "https://api.github.com/users/paulpak58/repos", "events_url": "https://api.github.com/users/paulpak58/events{/privacy}", "received_events_url": "https://api.github.com/users/paulpak58/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T09:32:13
2025-10-07T13:09:58
2025-10-07T13:09:58
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41401", "html_url": "https://github.com/huggingface/transformers/pull/41401", "diff_url": "https://github.com/huggingface/transformers/pull/41401.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41401.patch", "merged_at": "2025-10-07T13:09:58" }
# What does this PR do? This PR implements [LFM2-8B-A1B](https://huggingface.co/LiquidAI/LFM2-8B-A1B), a hybrid Mixture-of-Experts architecture variant of [LFM2](https://huggingface.co/collections/LiquidAI/lfm2-686d721927015b2ad73eaa38). The LFM2 family is optimized for on-device inference by combining short‑range, input‑aware gated convolutions with grouped‑query attention (GQA). LFM2‑MoE keeps this fast backbone and introduces sparse MoE feed‑forward networks to add representational capacity without significantly increasing the active compute path. ## Before submitting - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [x] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [x] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41401/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41401/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41400
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41400/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41400/comments
https://api.github.com/repos/huggingface/transformers/issues/41400/events
https://github.com/huggingface/transformers/pull/41400
3,490,653,619
PR_kwDOCUB6oc6sc4JL
41,400
Automatically infer attention_mask when pad_token_id == eos_token_id
{ "login": "tanuj-rai", "id": 84439872, "node_id": "MDQ6VXNlcjg0NDM5ODcy", "avatar_url": "https://avatars.githubusercontent.com/u/84439872?v=4", "gravatar_id": "", "url": "https://api.github.com/users/tanuj-rai", "html_url": "https://github.com/tanuj-rai", "followers_url": "https://api.github.com/users/tanuj-rai/followers", "following_url": "https://api.github.com/users/tanuj-rai/following{/other_user}", "gists_url": "https://api.github.com/users/tanuj-rai/gists{/gist_id}", "starred_url": "https://api.github.com/users/tanuj-rai/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/tanuj-rai/subscriptions", "organizations_url": "https://api.github.com/users/tanuj-rai/orgs", "repos_url": "https://api.github.com/users/tanuj-rai/repos", "events_url": "https://api.github.com/users/tanuj-rai/events{/privacy}", "received_events_url": "https://api.github.com/users/tanuj-rai/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T09:18:00
2025-10-08T16:27:25
2025-10-08T16:27:25
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41400", "html_url": "https://github.com/huggingface/transformers/pull/41400", "diff_url": "https://github.com/huggingface/transformers/pull/41400.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41400.patch", "merged_at": null }
# What does this PR do? This PR updates the _prepare_special_tokens in the generation utilities to automatically infer a correct attention mask when none is provided and pad_token_id == eos_token_id. <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes ##41392 ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [X] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? @ArthurZucker @vasqu <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41400/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41400/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41399
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41399/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41399/comments
https://api.github.com/repos/huggingface/transformers/issues/41399/events
https://github.com/huggingface/transformers/issues/41399
3,490,638,919
I_kwDOCUB6oc7QDuxH
41,399
Filter models by Original Provider
{ "login": "HarrySoteriou", "id": 25437880, "node_id": "MDQ6VXNlcjI1NDM3ODgw", "avatar_url": "https://avatars.githubusercontent.com/u/25437880?v=4", "gravatar_id": "", "url": "https://api.github.com/users/HarrySoteriou", "html_url": "https://github.com/HarrySoteriou", "followers_url": "https://api.github.com/users/HarrySoteriou/followers", "following_url": "https://api.github.com/users/HarrySoteriou/following{/other_user}", "gists_url": "https://api.github.com/users/HarrySoteriou/gists{/gist_id}", "starred_url": "https://api.github.com/users/HarrySoteriou/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/HarrySoteriou/subscriptions", "organizations_url": "https://api.github.com/users/HarrySoteriou/orgs", "repos_url": "https://api.github.com/users/HarrySoteriou/repos", "events_url": "https://api.github.com/users/HarrySoteriou/events{/privacy}", "received_events_url": "https://api.github.com/users/HarrySoteriou/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 2648621985, "node_id": "MDU6TGFiZWwyNjQ4NjIxOTg1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Feature%20request", "name": "Feature request", "color": "FBCA04", "default": false, "description": "Request for a new feature" } ]
open
false
null
[]
null
[]
2025-10-07T09:13:49
2025-10-08T09:24:22
null
NONE
null
null
null
null
### Feature request Add a filter that only displays official model releases. ### Motivation When looking to compare various models I would like to start from the original model provider (Qwen, Kimi, DeepSeek, etc...) and then check out quantizations, finetunes, etc... With the volume of models being pushed on HuggingFace it feels impossible to collect the best 5-6 latest models so i can test them. ### Your contribution I can only contribute the though process: A dictionary with model providers (keys) and the names of their model releases (values). When enabling the filter for original releases it should only display the models that are the values of that dictionary.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41399/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41399/timeline
null
null
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
false
https://api.github.com/repos/huggingface/transformers/issues/41398
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41398/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41398/comments
https://api.github.com/repos/huggingface/transformers/issues/41398/events
https://github.com/huggingface/transformers/issues/41398
3,490,426,762
I_kwDOCUB6oc7QC6-K
41,398
OwlViT Process has error 'numpy.ndarray' object has no attribute 'to'
{ "login": "PinedTD2025", "id": 232501780, "node_id": "U_kgDODduyFA", "avatar_url": "https://avatars.githubusercontent.com/u/232501780?v=4", "gravatar_id": "", "url": "https://api.github.com/users/PinedTD2025", "html_url": "https://github.com/PinedTD2025", "followers_url": "https://api.github.com/users/PinedTD2025/followers", "following_url": "https://api.github.com/users/PinedTD2025/following{/other_user}", "gists_url": "https://api.github.com/users/PinedTD2025/gists{/gist_id}", "starred_url": "https://api.github.com/users/PinedTD2025/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/PinedTD2025/subscriptions", "organizations_url": "https://api.github.com/users/PinedTD2025/orgs", "repos_url": "https://api.github.com/users/PinedTD2025/repos", "events_url": "https://api.github.com/users/PinedTD2025/events{/privacy}", "received_events_url": "https://api.github.com/users/PinedTD2025/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 1862634478, "node_id": "MDU6TGFiZWwxODYyNjM0NDc4", "url": "https://api.github.com/repos/huggingface/transformers/labels/Should%20Fix", "name": "Should Fix", "color": "FF0000", "default": false, "description": "This has been identified as a bug and should be fixed." }, { "id": 3817266200, "node_id": "MDU6TGFiZWwzODE3MjY2MjAw", "url": "https://api.github.com/repos/huggingface/transformers/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": null }, { "id": 5769473378, "node_id": "LA_kwDOCUB6oc8AAAABV-MtYg", "url": "https://api.github.com/repos/huggingface/transformers/labels/Vision", "name": "Vision", "color": "C079EF", "default": false, "description": "" } ]
closed
false
null
[]
null
[]
2025-10-07T08:10:31
2025-10-08T07:15:22
2025-10-08T07:15:22
NONE
null
null
null
null
### System Info - `transformers` version: 4.57.0.dev0 - Platform: macOS-15.7.1-x86_64-i386-64bit - Python version: 3.12.10 - Huggingface_hub version: 1.0.0.rc2 - Safetensors version: 0.6.2 - Accelerate version: not installed - Accelerate config: not found - DeepSpeed version: not installed - PyTorch version (accelerator?): 2.2.2 (NA) ### Who can help? @zucchini-nlp ### Information - [x] The official example scripts - [x] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below) ### Reproduction I try to do zero-shot learning by using model = OwlViTForObjectDetection.from_pretrained("google/owlvit-base-patch32") processor = OwlViTProcessor.from_pretrained("google/owlvit-base-patch32") I also follow the tutorial image = Image.open('1.png').convert("RGB") text_queries = ["product"] inputs = processor(text=text_queries, images=image, return_tensors="pt").to(device) But I get error when I forward # Get predictions with torch.no_grad(): outputs = model(**inputs) <img width="1270" height="179" alt="Image" src="https://github.com/user-attachments/assets/35c22297-74da-4f0f-95f8-9b386d978803" /> I get this error both from my local macbook or colab (both cpu/gpu) ### Expected behavior it should able to forward model
{ "login": "PinedTD2025", "id": 232501780, "node_id": "U_kgDODduyFA", "avatar_url": "https://avatars.githubusercontent.com/u/232501780?v=4", "gravatar_id": "", "url": "https://api.github.com/users/PinedTD2025", "html_url": "https://github.com/PinedTD2025", "followers_url": "https://api.github.com/users/PinedTD2025/followers", "following_url": "https://api.github.com/users/PinedTD2025/following{/other_user}", "gists_url": "https://api.github.com/users/PinedTD2025/gists{/gist_id}", "starred_url": "https://api.github.com/users/PinedTD2025/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/PinedTD2025/subscriptions", "organizations_url": "https://api.github.com/users/PinedTD2025/orgs", "repos_url": "https://api.github.com/users/PinedTD2025/repos", "events_url": "https://api.github.com/users/PinedTD2025/events{/privacy}", "received_events_url": "https://api.github.com/users/PinedTD2025/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41398/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41398/timeline
null
completed
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
true
https://api.github.com/repos/huggingface/transformers/issues/41397
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41397/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41397/comments
https://api.github.com/repos/huggingface/transformers/issues/41397/events
https://github.com/huggingface/transformers/issues/41397
3,489,779,207
I_kwDOCUB6oc7QAc4H
41,397
ViTForMaskedImageModeling (vit-base-patch16-224-in21k) distributed training warning
{ "login": "willxxy", "id": 90741489, "node_id": "MDQ6VXNlcjkwNzQxNDg5", "avatar_url": "https://avatars.githubusercontent.com/u/90741489?v=4", "gravatar_id": "", "url": "https://api.github.com/users/willxxy", "html_url": "https://github.com/willxxy", "followers_url": "https://api.github.com/users/willxxy/followers", "following_url": "https://api.github.com/users/willxxy/following{/other_user}", "gists_url": "https://api.github.com/users/willxxy/gists{/gist_id}", "starred_url": "https://api.github.com/users/willxxy/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/willxxy/subscriptions", "organizations_url": "https://api.github.com/users/willxxy/orgs", "repos_url": "https://api.github.com/users/willxxy/repos", "events_url": "https://api.github.com/users/willxxy/events{/privacy}", "received_events_url": "https://api.github.com/users/willxxy/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 3817266200, "node_id": "MDU6TGFiZWwzODE3MjY2MjAw", "url": "https://api.github.com/repos/huggingface/transformers/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": null }, { "id": 5769473378, "node_id": "LA_kwDOCUB6oc8AAAABV-MtYg", "url": "https://api.github.com/repos/huggingface/transformers/labels/Vision", "name": "Vision", "color": "C079EF", "default": false, "description": "" } ]
open
false
null
[]
null
[]
2025-10-07T03:13:35
2025-10-13T23:54:47
null
NONE
null
null
null
null
### System Info I am having some strange errors/warnings when running the vit model. I am on version `4.56.0`. 2 A6000 NVIDIA GPUS. Ubuntu 20.04.6 LTS. Python 3.10.18. torch version 2.6.0+cu124. ### Who can help? @molbap @yonigozlan 1. When I use the `AutoModel.from_pretrained` class for the `vit-base-patch16-224-in21k` checkpoint and wrap the model with `DDP` and run with `torchrun`, I get this error ``` [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/utils/generic.py", line 1064, in wrapper [rank1]: outputs = func(self, *args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/models/vit/modeling_vit.py", line 482, in forward [rank1]: embedding_output = self.embeddings( [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank1]: return self._call_impl(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank1]: return forward_call(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/models/vit/modeling_vit.py", line 112, in forward [rank1]: mask_tokens = self.mask_token.expand(batch_size, seq_length, -1) [rank1]: AttributeError: 'NoneType' object has no attribute 'expand' ``` However, when I use the `ViTForMaskedImageModeling.from_pretrained` class I get this warning ``` /home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/autograd/graph.py:823: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance. grad.sizes() = [768, 768, 1, 1], strides() = [768, 1, 768, 768] bucket_view.sizes() = [768, 768, 1, 1], strides() = [768, 1, 1, 1] (Triggered internally at /pytorch/torch/csrc/distributed/c10d/reducer.cpp:327.) return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass /home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/autograd/graph.py:823: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance. grad.sizes() = [768, 768, 1, 1], strides() = [768, 1, 768, 768] bucket_view.sizes() = [768, 768, 1, 1], strides() = [768, 1, 1, 1] (Triggered internally at /pytorch/torch/csrc/distributed/c10d/reducer.cpp:327.) return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass ``` Why does this happen? I found maybe one related issue [here](https://github.com/huggingface/transformers/issues/32614). I found [this too](https://github.com/huggingface/transformers/issues/23832) but I don't think this one is related. ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [ ] My own task or dataset (give details below) ### Reproduction run `CUDA_VISIBLE_DEVICES=1,2 torchrun --nproc_per_node=2 --master_port=12355 ecg_bench/test.py` ``` import os import torch import torch.distributed as dist from torch.nn.parallel import DistributedDataParallel as DDP from transformers import ViTForMaskedImageModeling, AutoModel, AutoImageProcessor from ecg_bench.configs.constants import HF_CACHE_DIR def setup_ddp(): if not dist.is_initialized(): dist.init_process_group(backend="nccl") local_rank = int(os.environ.get("LOCAL_RANK", 0)) torch.cuda.set_device(local_rank) return local_rank def test_model(model_class): model = model_class.from_pretrained("google/vit-base-patch16-224-in21k", cache_dir=HF_CACHE_DIR) processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k", cache_dir=HF_CACHE_DIR) local_rank = setup_ddp() model = DDP(model.to(f"cuda:{local_rank}"), device_ids=[local_rank]) images = torch.rand(2, 3, 224, 224) inputs = processor(images=images, return_tensors="pt") pixel_values = inputs["pixel_values"].to(f"cuda:{local_rank}") bool_masked_pos = torch.zeros(2, 196, dtype=torch.bool, device=f"cuda:{local_rank}") for i in range(2): bool_masked_pos[i, torch.randperm(196)[:29]] = True outputs = model(pixel_values=pixel_values, bool_masked_pos=bool_masked_pos) loss = outputs.loss loss.backward() print(f"Loss: {loss.item()}") if __name__ == "__main__": test_model(ViTForMaskedImageModeling) test_model(AutoModel) ``` ### Expected behavior ``` (ecgv2) william@gpu2:~/ecg_dev/ECG-Bench-v2$ bash scripts/test_ddp_reproduction.sh Running minimal reproduction test with DDP... W1007 03:11:37.695000 3870938 site-packages/torch/distributed/run.py:792] W1007 03:11:37.695000 3870938 site-packages/torch/distributed/run.py:792] ***************************************** W1007 03:11:37.695000 3870938 site-packages/torch/distributed/run.py:792] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. W1007 03:11:37.695000 3870938 site-packages/torch/distributed/run.py:792] ***************************************** Some weights of ViTForMaskedImageModeling were not initialized from the model checkpoint at google/vit-base-patch16-224-in21k and are newly initialized: ['decoder.0.bias', 'decoder.0.weight', 'vit.embeddings.mask_token'] You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. Some weights of ViTForMaskedImageModeling were not initialized from the model checkpoint at google/vit-base-patch16-224-in21k and are newly initialized: ['decoder.0.bias', 'decoder.0.weight', 'vit.embeddings.mask_token'] You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. Fetching 1 files: 100%|██████████████████████████████████████████| 1/1 [00:00<00:00, 16070.13it/s] Fetching 1 files: 100%|██████████████████████████████████████████| 1/1 [00:00<00:00, 19239.93it/s] Fast image processor class <class 'transformers.models.vit.image_processing_vit_fast.ViTImageProcessorFast'> is available for this model. Using slow image processor class. To use the fast image processor class set `use_fast=True`. Fast image processor class <class 'transformers.models.vit.image_processing_vit_fast.ViTImageProcessorFast'> is available for this model. Using slow image processor class. To use the fast image processor class set `use_fast=True`. Fetching 1 files: 100%|██████████████████████████████████████████| 1/1 [00:00<00:00, 13443.28it/s] Fetching 1 files: 100%|███████████████████████████████████████████| 1/1 [00:00<00:00, 8473.34it/s] It looks like you are trying to rescale already rescaled images. If the input images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again. It looks like you are trying to rescale already rescaled images. If the input images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again. /home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/autograd/graph.py:823: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance. grad.sizes() = [768, 768, 1, 1], strides() = [768, 1, 768, 768] bucket_view.sizes() = [768, 768, 1, 1], strides() = [768, 1, 1, 1] (Triggered internally at /pytorch/torch/csrc/distributed/c10d/reducer.cpp:327.) return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass /home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/autograd/graph.py:823: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance. grad.sizes() = [768, 768, 1, 1], strides() = [768, 1, 768, 768] bucket_view.sizes() = [768, 768, 1, 1], strides() = [768, 1, 1, 1] (Triggered internally at /pytorch/torch/csrc/distributed/c10d/reducer.cpp:327.) return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass Loss: 1.00250244140625Loss: 1.002082109451294 Fetching 1 files: 100%|███████████████████████████████████████████| 1/1 [00:00<00:00, 9664.29it/s] Fetching 1 files: 100%|███████████████████████████████████████████| 1/1 [00:00<00:00, 6364.65it/s] Fast image processor class <class 'transformers.models.vit.image_processing_vit_fast.ViTImageProcessorFast'> is available for this model. Using slow image processor class. To use the fast image processor class set `use_fast=True`. Fast image processor class <class 'transformers.models.vit.image_processing_vit_fast.ViTImageProcessorFast'> is available for this model. Using slow image processor class. To use the fast image processor class set `use_fast=True`. Fetching 1 files: 100%|██████████████████████████████████████████| 1/1 [00:00<00:00, 18001.30it/s] Fetching 1 files: 100%|██████████████████████████████████████████| 1/1 [00:00<00:00, 22192.08it/s] [rank1]: Traceback (most recent call last): [rank1]: File "/home/william/ecg_dev/ECG-Bench-v2/ecg_bench/test.py", line 37, in <module> [rank1]: test_model(AutoModel) [rank1]: File "/home/william/ecg_dev/ECG-Bench-v2/ecg_bench/test.py", line 29, in test_model [rank1]: outputs = model(pixel_values=pixel_values, bool_masked_pos=bool_masked_pos) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank1]: return self._call_impl(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank1]: return forward_call(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/parallel/distributed.py", line 1643, in forward [rank1]: else self._run_ddp_forward(*inputs, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/parallel/distributed.py", line 1459, in _run_ddp_forward [rank1]: return self.module(*inputs, **kwargs) # type: ignore[index] [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank1]: return self._call_impl(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank1]: return forward_call(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/utils/generic.py", line 1064, in wrapper [rank1]: outputs = func(self, *args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/models/vit/modeling_vit.py", line 482, in forward [rank1]: embedding_output = self.embeddings( [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank1]: return self._call_impl(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank1]: return forward_call(*args, **kwargs) [rank1]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/models/vit/modeling_vit.py", line 112, in forward [rank1]: mask_tokens = self.mask_token.expand(batch_size, seq_length, -1) [rank1]: AttributeError: 'NoneType' object has no attribute 'expand' [rank0]: Traceback (most recent call last): [rank0]: File "/home/william/ecg_dev/ECG-Bench-v2/ecg_bench/test.py", line 37, in <module> [rank0]: test_model(AutoModel) [rank0]: File "/home/william/ecg_dev/ECG-Bench-v2/ecg_bench/test.py", line 29, in test_model [rank0]: outputs = model(pixel_values=pixel_values, bool_masked_pos=bool_masked_pos) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/parallel/distributed.py", line 1643, in forward [rank0]: else self._run_ddp_forward(*inputs, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/parallel/distributed.py", line 1459, in _run_ddp_forward [rank0]: return self.module(*inputs, **kwargs) # type: ignore[index] [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/utils/generic.py", line 1064, in wrapper [rank0]: outputs = func(self, *args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/models/vit/modeling_vit.py", line 482, in forward [rank0]: embedding_output = self.embeddings( [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl [rank0]: return self._call_impl(*args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl [rank0]: return forward_call(*args, **kwargs) [rank0]: File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/transformers/models/vit/modeling_vit.py", line 112, in forward [rank0]: mask_tokens = self.mask_token.expand(batch_size, seq_length, -1) [rank0]: AttributeError: 'NoneType' object has no attribute 'expand' [rank0]:[W1007 03:11:47.275383972 ProcessGroupNCCL.cpp:1496] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator()) W1007 03:11:48.425000 3870938 site-packages/torch/distributed/elastic/multiprocessing/api.py:897] Sending process 3871025 closing signal SIGTERM E1007 03:11:48.589000 3870938 site-packages/torch/distributed/elastic/multiprocessing/api.py:869] failed (exitcode: 1) local_rank: 1 (pid: 3871026) of binary: /home/william/miniconda3/envs/ecgv2/bin/python3 Traceback (most recent call last): File "/home/william/miniconda3/envs/ecgv2/bin/torchrun", line 10, in <module> sys.exit(main()) File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper return f(*args, **kwargs) File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/distributed/run.py", line 918, in main run(args) File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/distributed/run.py", line 909, in run elastic_launch( File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 138, in __call__ return launch_agent(self._config, self._entrypoint, list(args)) File "/home/william/miniconda3/envs/ecgv2/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 269, in launch_agent raise ChildFailedError( torch.distributed.elastic.multiprocessing.errors.ChildFailedError: ============================================================ ecg_bench/test.py FAILED ------------------------------------------------------------ Failures: <NO_OTHER_FAILURES> ------------------------------------------------------------ Root Cause (first observed failure): [0]: time : 2025-10-07_03:11:48 host : gpu2 rank : 1 (local_rank: 1) exitcode : 1 (pid: 3871026) error_file: <N/A> traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html ============================================================ Test completed! ```
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41397/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41397/timeline
null
null
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
false
https://api.github.com/repos/huggingface/transformers/issues/41396
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41396/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41396/comments
https://api.github.com/repos/huggingface/transformers/issues/41396/events
https://github.com/huggingface/transformers/pull/41396
3,489,724,001
PR_kwDOCUB6oc6sZyAY
41,396
[Cache] lfm2 cache: allocate empty kv layers during init
{ "login": "paulpak58", "id": 52512091, "node_id": "MDQ6VXNlcjUyNTEyMDkx", "avatar_url": "https://avatars.githubusercontent.com/u/52512091?v=4", "gravatar_id": "", "url": "https://api.github.com/users/paulpak58", "html_url": "https://github.com/paulpak58", "followers_url": "https://api.github.com/users/paulpak58/followers", "following_url": "https://api.github.com/users/paulpak58/following{/other_user}", "gists_url": "https://api.github.com/users/paulpak58/gists{/gist_id}", "starred_url": "https://api.github.com/users/paulpak58/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/paulpak58/subscriptions", "organizations_url": "https://api.github.com/users/paulpak58/orgs", "repos_url": "https://api.github.com/users/paulpak58/repos", "events_url": "https://api.github.com/users/paulpak58/events{/privacy}", "received_events_url": "https://api.github.com/users/paulpak58/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T02:36:16
2025-10-07T12:01:31
2025-10-07T12:01:31
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41396", "html_url": "https://github.com/huggingface/transformers/pull/41396", "diff_url": "https://github.com/huggingface/transformers/pull/41396.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41396.patch", "merged_at": "2025-10-07T12:01:31" }
# What does this PR do? The current Lfm2Cache adds empty kv-cache tensors when it detects a `layer_idx` that corresponds to an attention layer, but this ends up making the length of Lfm2Cache < `num_hidden_layers` if the last layer is not an attention layer. This change makes Lfm2Cache follow convention to the default cache by pre-allocating empty tensors during init, so we can keep the total length fixed to the number of hidden layers. <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41396/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41396/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41395
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41395/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41395/comments
https://api.github.com/repos/huggingface/transformers/issues/41395/events
https://github.com/huggingface/transformers/pull/41395
3,489,548,620
PR_kwDOCUB6oc6sZNbY
41,395
Enable FURB rules in ruff
{ "login": "cyyever", "id": 17618148, "node_id": "MDQ6VXNlcjE3NjE4MTQ4", "avatar_url": "https://avatars.githubusercontent.com/u/17618148?v=4", "gravatar_id": "", "url": "https://api.github.com/users/cyyever", "html_url": "https://github.com/cyyever", "followers_url": "https://api.github.com/users/cyyever/followers", "following_url": "https://api.github.com/users/cyyever/following{/other_user}", "gists_url": "https://api.github.com/users/cyyever/gists{/gist_id}", "starred_url": "https://api.github.com/users/cyyever/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/cyyever/subscriptions", "organizations_url": "https://api.github.com/users/cyyever/orgs", "repos_url": "https://api.github.com/users/cyyever/repos", "events_url": "https://api.github.com/users/cyyever/events{/privacy}", "received_events_url": "https://api.github.com/users/cyyever/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-07T00:33:10
2025-10-17T15:15:38
2025-10-17T15:00:40
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41395", "html_url": "https://github.com/huggingface/transformers/pull/41395", "diff_url": "https://github.com/huggingface/transformers/pull/41395.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41395.patch", "merged_at": "2025-10-17T15:00:40" }
# What does this PR do? ruff contains the below `FURB` rules: ``` FURB101 open and read should be replaced by Path({filename}).{suggestion} FURB103 open and write should be replaced by Path({filename}).{suggestion} FURB105 Unnecessary empty string passed to print FURB110 Replace ternary if expression with or operator FURB113 Use {suggestion} instead of repeatedly calling {name}.append() FURB116 Replace {function_name} call with {display} FURB118 Use operator.{operator} instead of defining a {target} FURB122 Use of {}.write in a for loop FURB129 Instead of calling readlines(), iterate over file object directly FURB131 Prefer clear over deleting a full slice FURB132 Use {suggestion} instead of check and remove FURB136 Replace if expression with {min_max} call FURB140 Use itertools.starmap instead of the generator FURB142 Use of set.{}() in a for loop FURB145 Prefer copy method over slicing FURB148 enumerate value is unused, use for x in range(len(y)) instead FURB152 Replace {literal} with math.{constant} FURB154 Use of repeated consecutive {} FURB156 Use of hardcoded string charset FURB157 Verbose expression in Decimal constructor FURB161 Use of bin({existing}).count('1') FURB162 Unnecessary timezone replacement with zero offset FURB163 Prefer math.{log_function}({arg}) over math.log with a redundant base FURB164 Verbose method {method_name} in {constructor} construction FURB166 Use of int with explicit base={base} after removing prefix FURB167 Use of regular expression alias re.{} FURB168 Prefer is operator over isinstance to check if an object is None FURB169 When checking against None, use {} instead of comparison with type(None) FURB171 Membership test against single-item container FURB177 Prefer Path.cwd() over Path().resolve() for current-directory lookups FURB180 Use of metaclass=abc.ABCMeta to define abstract base class FURB181 Use of hashlib's .digest().hex() FURB187 Use of assignment of reversed on list {name} FURB188 Prefer str.removeprefix() over conditionally replacing with slice. FURB189 Subclassing {subclass} can be error prone, use collections.{replacement} instead FURB192 Prefer min over sorted() to compute the minimum value in a sequence ``` In my opinion these rules are useful and many of their fixed have been applied ion previous PRs.
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41395/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41395/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41394
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41394/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41394/comments
https://api.github.com/repos/huggingface/transformers/issues/41394/events
https://github.com/huggingface/transformers/pull/41394
3,489,336,468
PR_kwDOCUB6oc6sYf8r
41,394
Adding superglue fast image processing
{ "login": "AlphaOrOmega", "id": 103359692, "node_id": "U_kgDOBikkzA", "avatar_url": "https://avatars.githubusercontent.com/u/103359692?v=4", "gravatar_id": "", "url": "https://api.github.com/users/AlphaOrOmega", "html_url": "https://github.com/AlphaOrOmega", "followers_url": "https://api.github.com/users/AlphaOrOmega/followers", "following_url": "https://api.github.com/users/AlphaOrOmega/following{/other_user}", "gists_url": "https://api.github.com/users/AlphaOrOmega/gists{/gist_id}", "starred_url": "https://api.github.com/users/AlphaOrOmega/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/AlphaOrOmega/subscriptions", "organizations_url": "https://api.github.com/users/AlphaOrOmega/orgs", "repos_url": "https://api.github.com/users/AlphaOrOmega/repos", "events_url": "https://api.github.com/users/AlphaOrOmega/events{/privacy}", "received_events_url": "https://api.github.com/users/AlphaOrOmega/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T22:29:36
2025-10-16T19:34:10
2025-10-16T19:34:10
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41394", "html_url": "https://github.com/huggingface/transformers/pull/41394", "diff_url": "https://github.com/huggingface/transformers/pull/41394.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41394.patch", "merged_at": "2025-10-16T19:34:10" }
# What does this PR do? TLDR : - Implement fast processor for SuperGlue - About 3 times faster This PR aims to translate the features of the class `SuperGlueImageProcessor` in the fast equivalent class `SuperGlueImageProcessorFast`. The implementation heavily follows the standard implementation but reduces memory consumption and about 3 times the execution speed on my hardware. The implementation mostly refactor the image formatting in the `preprocessing` step, notably by using torch tensors instead of PIL or Numpy. ## Test Performed RUN_SLOW=1 python -m pytest tests/models/superglue/test_image_processing_superglue.py With an additional test based on the default processor tester (this test has not to be included in the repo) : ```python @require_vision @require_torch def test_fast_is_faster_than_slow(self): if not self.test_slow_image_processor or not self.test_fast_image_processor: self.skipTest(reason="Skipping speed test") if self.image_processing_class is None or self.fast_image_processing_class is None: self.skipTest(reason="Skipping speed test as one of the image processors is not defined") def measure_time(image_processor, image): # Warmup for _ in range(5): _ = image_processor(image, return_tensors="pt") all_times = [] for _ in range(10): start = time.time() _ = image_processor(image, return_tensors="pt") all_times.append(time.time() - start) # Take the average of the fastest 3 runs avg_time = sum(sorted(all_times[:3])) / 3.0 return avg_time dummy_images = self.image_processor_tester.prepare_image_inputs(equal_resolution=False, torchify=True) image_processor_slow = self.image_processing_class(**self.image_processor_dict) image_processor_fast = self.fast_image_processing_class(**self.image_processor_dict) fast_time = measure_time(image_processor_fast, dummy_images) slow_time = measure_time(image_processor_slow, dummy_images) self.assertLessEqual(fast_time, slow_time) ``` By reviewing the flame graph, I noticed the improvement in every `__calls__` made to the fast version. Callers of the old processor, and the full execution time of the method: <img width="1239" height="391" alt="image" src="https://github.com/user-attachments/assets/0fedbb2f-991c-481e-9425-b041b4f31767" /> The equivalent but with the fast processor: <img width="1255" height="385" alt="image" src="https://github.com/user-attachments/assets/ff658450-0c8c-459d-b58c-6da727555558" /> Some calls made during the test passes directly to the preprocess function, without passing by the `__call__` one, I am including them as well: Slow <img width="1150" height="209" alt="image" src="https://github.com/user-attachments/assets/b763e4c8-0336-42fe-a1de-3611dc2b0f66" /> Fast <img width="1230" height="176" alt="image" src="https://github.com/user-attachments/assets/9979f600-e03b-4732-85eb-91a9ded018ee" /> ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ X ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ X ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. link : [Contributions Welcome] Add Fast Image Processors https://github.com/huggingface/transformers/issues/36978#issue-2947632853 - [ X ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Thank you for reviewing my PR @yonigozlan (or anyone else :) )
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41394/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41394/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41393
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41393/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41393/comments
https://api.github.com/repos/huggingface/transformers/issues/41393/events
https://github.com/huggingface/transformers/pull/41393
3,489,270,880
PR_kwDOCUB6oc6sYRpY
41,393
Fix MaskFormer/Mask2Former fast image processors
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-06T21:58:56
2025-10-20T12:32:43
null
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41393", "html_url": "https://github.com/huggingface/transformers/pull/41393", "diff_url": "https://github.com/huggingface/transformers/pull/41393.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41393.patch", "merged_at": null }
# What does this PR do? Depends on https://github.com/huggingface/transformers/pull/41391. These two fast image processors had issues and were not properly tested: - There was an issue where the processors would crash if `do_resize-False` - After conversion to binary masks, the grouped masks cannot be stacked anymore, as their channels dimensions are not the same. This fix uses the method introduced in https://github.com/huggingface/transformers/pull/41391 to group the masks according to the shapes of the corresponding images. This PR fixes the issues and ensure that the integration tests are also ran with the fast image processors
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41393/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41393/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41392
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41392/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41392/comments
https://api.github.com/repos/huggingface/transformers/issues/41392/events
https://github.com/huggingface/transformers/issues/41392
3,489,222,243
I_kwDOCUB6oc7P-U5j
41,392
automaticlly setting attention mask when pas token is equal to the eos token
{ "login": "yonikremer", "id": 76044840, "node_id": "MDQ6VXNlcjc2MDQ0ODQw", "avatar_url": "https://avatars.githubusercontent.com/u/76044840?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonikremer", "html_url": "https://github.com/yonikremer", "followers_url": "https://api.github.com/users/yonikremer/followers", "following_url": "https://api.github.com/users/yonikremer/following{/other_user}", "gists_url": "https://api.github.com/users/yonikremer/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonikremer/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonikremer/subscriptions", "organizations_url": "https://api.github.com/users/yonikremer/orgs", "repos_url": "https://api.github.com/users/yonikremer/repos", "events_url": "https://api.github.com/users/yonikremer/events{/privacy}", "received_events_url": "https://api.github.com/users/yonikremer/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 1314768611, "node_id": "MDU6TGFiZWwxMzE0NzY4NjEx", "url": "https://api.github.com/repos/huggingface/transformers/labels/wontfix", "name": "wontfix", "color": "ffffff", "default": true, "description": null }, { "id": 2648621985, "node_id": "MDU6TGFiZWwyNjQ4NjIxOTg1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Feature%20request", "name": "Feature request", "color": "FBCA04", "default": false, "description": "Request for a new feature" } ]
closed
false
null
[]
null
[]
2025-10-06T21:41:37
2025-10-08T16:26:44
2025-10-08T16:26:44
NONE
null
null
null
null
### Feature request In AutoModelForCausalLM If the model eos token and pad token is the same, automaticly set the attention mask by treating the first eos token as a normal eos token and the rest as padding tokens. ### Motivation hi, I am using qwen3-0.6b (infrence only, no training) and got the message: 'The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's `attention_mask` to obtain reliable results' I solved it for my specific case by treating the first eos of each token as an eos token and the rest as padding tokens. It can be solved like that for all cases of AutoModelForCausalLM where the tokenizer padding size if left. ### Your contribution Unfortunently I work full time and probably won't have the time for the overhead of contributing to a new project in the near future.
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41392/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41392/timeline
null
completed
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
true
https://api.github.com/repos/huggingface/transformers/issues/41391
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41391/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41391/comments
https://api.github.com/repos/huggingface/transformers/issues/41391/events
https://github.com/huggingface/transformers/pull/41391
3,489,050,754
PR_kwDOCUB6oc6sXgVA
41,391
Add MLlama fast image processor
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T20:51:49
2025-10-13T09:16:06
2025-10-13T09:16:06
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41391", "html_url": "https://github.com/huggingface/transformers/pull/41391", "diff_url": "https://github.com/huggingface/transformers/pull/41391.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41391.patch", "merged_at": "2025-10-13T09:16:06" }
# What does this PR do? Finishes https://github.com/huggingface/transformers/pull/37539 It also adds support to group other inputs that mirrors the shape of images in `group_images_by_shape`, it could also be useful for some other future video processors @zucchini-nlp ;).
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41391/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41391/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41390
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41390/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41390/comments
https://api.github.com/repos/huggingface/transformers/issues/41390/events
https://github.com/huggingface/transformers/pull/41390
3,488,981,636
PR_kwDOCUB6oc6sXROs
41,390
docs: replace twitter.com with x.com
{ "login": "cseas", "id": 29686866, "node_id": "MDQ6VXNlcjI5Njg2ODY2", "avatar_url": "https://avatars.githubusercontent.com/u/29686866?v=4", "gravatar_id": "", "url": "https://api.github.com/users/cseas", "html_url": "https://github.com/cseas", "followers_url": "https://api.github.com/users/cseas/followers", "following_url": "https://api.github.com/users/cseas/following{/other_user}", "gists_url": "https://api.github.com/users/cseas/gists{/gist_id}", "starred_url": "https://api.github.com/users/cseas/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/cseas/subscriptions", "organizations_url": "https://api.github.com/users/cseas/orgs", "repos_url": "https://api.github.com/users/cseas/repos", "events_url": "https://api.github.com/users/cseas/events{/privacy}", "received_events_url": "https://api.github.com/users/cseas/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T20:31:31
2025-10-07T17:28:41
2025-10-07T12:40:16
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41390", "html_url": "https://github.com/huggingface/transformers/pull/41390", "diff_url": "https://github.com/huggingface/transformers/pull/41390.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41390.patch", "merged_at": null }
# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> ## Before submitting - [x] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41390/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41390/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41389
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41389/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41389/comments
https://api.github.com/repos/huggingface/transformers/issues/41389/events
https://github.com/huggingface/transformers/pull/41389
3,488,913,778
PR_kwDOCUB6oc6sXCDa
41,389
Modified Whisper to have the FFN referenced as mlp instead of fc1 and fc2
{ "login": "TheRealSal", "id": 75696327, "node_id": "MDQ6VXNlcjc1Njk2MzI3", "avatar_url": "https://avatars.githubusercontent.com/u/75696327?v=4", "gravatar_id": "", "url": "https://api.github.com/users/TheRealSal", "html_url": "https://github.com/TheRealSal", "followers_url": "https://api.github.com/users/TheRealSal/followers", "following_url": "https://api.github.com/users/TheRealSal/following{/other_user}", "gists_url": "https://api.github.com/users/TheRealSal/gists{/gist_id}", "starred_url": "https://api.github.com/users/TheRealSal/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/TheRealSal/subscriptions", "organizations_url": "https://api.github.com/users/TheRealSal/orgs", "repos_url": "https://api.github.com/users/TheRealSal/repos", "events_url": "https://api.github.com/users/TheRealSal/events{/privacy}", "received_events_url": "https://api.github.com/users/TheRealSal/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T20:13:51
2025-10-06T20:15:00
2025-10-06T20:14:00
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41389", "html_url": "https://github.com/huggingface/transformers/pull/41389", "diff_url": "https://github.com/huggingface/transformers/pull/41389.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41389.patch", "merged_at": null }
# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "TheRealSal", "id": 75696327, "node_id": "MDQ6VXNlcjc1Njk2MzI3", "avatar_url": "https://avatars.githubusercontent.com/u/75696327?v=4", "gravatar_id": "", "url": "https://api.github.com/users/TheRealSal", "html_url": "https://github.com/TheRealSal", "followers_url": "https://api.github.com/users/TheRealSal/followers", "following_url": "https://api.github.com/users/TheRealSal/following{/other_user}", "gists_url": "https://api.github.com/users/TheRealSal/gists{/gist_id}", "starred_url": "https://api.github.com/users/TheRealSal/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/TheRealSal/subscriptions", "organizations_url": "https://api.github.com/users/TheRealSal/orgs", "repos_url": "https://api.github.com/users/TheRealSal/repos", "events_url": "https://api.github.com/users/TheRealSal/events{/privacy}", "received_events_url": "https://api.github.com/users/TheRealSal/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41389/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41389/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41388
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41388/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41388/comments
https://api.github.com/repos/huggingface/transformers/issues/41388/events
https://github.com/huggingface/transformers/pull/41388
3,488,687,728
PR_kwDOCUB6oc6sWPew
41,388
Default to fast image processors for all models
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-06T19:11:56
2025-10-15T10:08:29
null
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41388", "html_url": "https://github.com/huggingface/transformers/pull/41388", "diff_url": "https://github.com/huggingface/transformers/pull/41388.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41388.patch", "merged_at": null }
# What does this PR do? Following the trial testing with Qwen_VL image processors, this extends defaulting to fast image processors even for checkpoints saved with a slow one to all models. Also made sure that all processors use AutoImageProcessor to instantiate their image_processor_class. On that point, defining default subclass in processors feels a bit redundant, as we basically already have that in auto classes. It would be nice to get rid of this for v5, wdyt @molbap @zucchini-nlp @ArthurZucker ? I'll open a PR for that too.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41388/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 1 }
https://api.github.com/repos/huggingface/transformers/issues/41388/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41387
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41387/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41387/comments
https://api.github.com/repos/huggingface/transformers/issues/41387/events
https://github.com/huggingface/transformers/issues/41387
3,488,640,475
I_kwDOCUB6oc7P8G3b
41,387
Parakeet and PerceptionLM are not registered correctly for auto
{ "login": "fschlatt", "id": 23191892, "node_id": "MDQ6VXNlcjIzMTkxODky", "avatar_url": "https://avatars.githubusercontent.com/u/23191892?v=4", "gravatar_id": "", "url": "https://api.github.com/users/fschlatt", "html_url": "https://github.com/fschlatt", "followers_url": "https://api.github.com/users/fschlatt/followers", "following_url": "https://api.github.com/users/fschlatt/following{/other_user}", "gists_url": "https://api.github.com/users/fschlatt/gists{/gist_id}", "starred_url": "https://api.github.com/users/fschlatt/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/fschlatt/subscriptions", "organizations_url": "https://api.github.com/users/fschlatt/orgs", "repos_url": "https://api.github.com/users/fschlatt/repos", "events_url": "https://api.github.com/users/fschlatt/events{/privacy}", "received_events_url": "https://api.github.com/users/fschlatt/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 3817266200, "node_id": "MDU6TGFiZWwzODE3MjY2MjAw", "url": "https://api.github.com/repos/huggingface/transformers/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": null } ]
closed
false
null
[]
null
[]
2025-10-06T18:58:51
2025-10-09T12:08:04
2025-10-09T12:08:04
CONTRIBUTOR
null
null
null
null
### System Info - `transformers` version: 4.57.0.dev0 - Platform: Linux-5.19.0-50-generic-x86_64-with-glibc2.35 - Python version: 3.13.7 - Huggingface_hub version: 1.0.0.rc2 - Safetensors version: 0.6.2 - Accelerate version: not installed - Accelerate config: not found - DeepSpeed version: not installed - PyTorch version (accelerator?): 2.8.0+cpu (NA) ### Who can help? @Cyrilvallez ### Information - [ ] The official example scripts - [x] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...) - [x] My own task or dataset (give details below) ### Reproduction When trying to iterate over the auto-registered mappings as below, the PerceptionLM models throw errors because it is incorrectly registered. ```py from transformers import MODEL_MAPPING MODEL_MAPPING.items() ``` ``` Traceback (most recent call last): File "/home/fschlatt/Code/uni/transformers/tmp.py", line 4, in <module> MODEL_MAPPING.items() ~~~~~~~~~~~~~~~~~~~^^ File "/home/fschlatt/Code/uni/transformers/src/transformers/models/auto/auto_factory.py", line 628, in items self._load_attr_from_module(key, self._model_mapping[key]), ~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/fschlatt/Code/uni/transformers/src/transformers/models/auto/auto_factory.py", line 597, in _load_attr_from_module return getattribute_from_module(self._modules[module_name], attr) File "/home/fschlatt/Code/uni/transformers/src/transformers/models/auto/auto_factory.py", line 519, in getattribute_from_module raise ValueError(f"Could not find {attr} neither in {module} nor in {transformers_module}!") ValueError: Could not find PerceptionEncoder neither in <module 'transformers.models.perception_lm' from '/home/fschlatt/Code/uni/transformers/src/transformers/models/perception_lm/__init__.py'> nor in <module 'transformers' from '/home/fschlatt/Code/uni/transformers/src/transformers/__init__.py'>! ``` ### Expected behavior It should be possible to iterate over the mappings
{ "login": "zucchini-nlp", "id": 100715397, "node_id": "U_kgDOBgDLhQ", "avatar_url": "https://avatars.githubusercontent.com/u/100715397?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zucchini-nlp", "html_url": "https://github.com/zucchini-nlp", "followers_url": "https://api.github.com/users/zucchini-nlp/followers", "following_url": "https://api.github.com/users/zucchini-nlp/following{/other_user}", "gists_url": "https://api.github.com/users/zucchini-nlp/gists{/gist_id}", "starred_url": "https://api.github.com/users/zucchini-nlp/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/zucchini-nlp/subscriptions", "organizations_url": "https://api.github.com/users/zucchini-nlp/orgs", "repos_url": "https://api.github.com/users/zucchini-nlp/repos", "events_url": "https://api.github.com/users/zucchini-nlp/events{/privacy}", "received_events_url": "https://api.github.com/users/zucchini-nlp/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41387/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41387/timeline
null
completed
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
true
https://api.github.com/repos/huggingface/transformers/issues/41386
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41386/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41386/comments
https://api.github.com/repos/huggingface/transformers/issues/41386/events
https://github.com/huggingface/transformers/pull/41386
3,488,487,086
PR_kwDOCUB6oc6sVjDu
41,386
enable new model uts to xpu and fix some failures on xpu
{ "login": "yao-matrix", "id": 7245027, "node_id": "MDQ6VXNlcjcyNDUwMjc=", "avatar_url": "https://avatars.githubusercontent.com/u/7245027?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yao-matrix", "html_url": "https://github.com/yao-matrix", "followers_url": "https://api.github.com/users/yao-matrix/followers", "following_url": "https://api.github.com/users/yao-matrix/following{/other_user}", "gists_url": "https://api.github.com/users/yao-matrix/gists{/gist_id}", "starred_url": "https://api.github.com/users/yao-matrix/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yao-matrix/subscriptions", "organizations_url": "https://api.github.com/users/yao-matrix/orgs", "repos_url": "https://api.github.com/users/yao-matrix/repos", "events_url": "https://api.github.com/users/yao-matrix/events{/privacy}", "received_events_url": "https://api.github.com/users/yao-matrix/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T18:15:54
2025-10-08T15:01:20
2025-10-08T10:14:51
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41386", "html_url": "https://github.com/huggingface/transformers/pull/41386", "diff_url": "https://github.com/huggingface/transformers/pull/41386.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41386.patch", "merged_at": "2025-10-08T10:14:50" }
covered modeling: - cohere2_vesion - ernie4_5 - internvl - llava - llava_next - phi - qwen2.5_omni - whisper - xglm @SunMarc @ydshieh , pls help review, thx very much.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41386/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41386/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41385
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41385/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41385/comments
https://api.github.com/repos/huggingface/transformers/issues/41385/events
https://github.com/huggingface/transformers/pull/41385
3,488,429,518
PR_kwDOCUB6oc6sVWi3
41,385
[`Ernie 4.5 Moe`] Fix Moe and offloading
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T18:00:27
2025-10-16T11:59:12
2025-10-16T11:59:02
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41385", "html_url": "https://github.com/huggingface/transformers/pull/41385", "diff_url": "https://github.com/huggingface/transformers/pull/41385.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41385.patch", "merged_at": "2025-10-16T11:59:02" }
The model is broken - Moe was initialized with wrong dims (loading the models already failed) - Gate needs the autocast decorator as well (we need to guarantee fp32) - Not sure why but using the forward with the moe statics works with offloading while retrieving the parameter etc does not - The router seems like an artefact
{ "login": "vasqu", "id": 73884904, "node_id": "MDQ6VXNlcjczODg0OTA0", "avatar_url": "https://avatars.githubusercontent.com/u/73884904?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vasqu", "html_url": "https://github.com/vasqu", "followers_url": "https://api.github.com/users/vasqu/followers", "following_url": "https://api.github.com/users/vasqu/following{/other_user}", "gists_url": "https://api.github.com/users/vasqu/gists{/gist_id}", "starred_url": "https://api.github.com/users/vasqu/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/vasqu/subscriptions", "organizations_url": "https://api.github.com/users/vasqu/orgs", "repos_url": "https://api.github.com/users/vasqu/repos", "events_url": "https://api.github.com/users/vasqu/events{/privacy}", "received_events_url": "https://api.github.com/users/vasqu/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41385/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41385/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41384
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41384/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41384/comments
https://api.github.com/repos/huggingface/transformers/issues/41384/events
https://github.com/huggingface/transformers/pull/41384
3,488,190,932
PR_kwDOCUB6oc6sUhsG
41,384
Remove `past_index`
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T16:51:42
2025-10-09T16:06:49
2025-10-09T16:06:46
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41384", "html_url": "https://github.com/huggingface/transformers/pull/41384", "diff_url": "https://github.com/huggingface/transformers/pull/41384.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41384.patch", "merged_at": "2025-10-09T16:06:46" }
# What does this PR do? This PR removes `past_index`. This was supposed to allow training a few old + specific arch models like transformersxl but I don't think anyone actually did it as there is no documentation at all. I think it is fine to remove it without any deprecation msg but happy to add it if needed. Maybe some users just copy paste Trainer code and try to maintain that.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41384/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41384/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41383
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41383/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41383/comments
https://api.github.com/repos/huggingface/transformers/issues/41383/events
https://github.com/huggingface/transformers/pull/41383
3,488,159,883
PR_kwDOCUB6oc6sUazX
41,383
remove `tpu_num_cores`
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T16:43:31
2025-10-10T13:53:29
2025-10-10T13:53:28
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41383", "html_url": "https://github.com/huggingface/transformers/pull/41383", "diff_url": "https://github.com/huggingface/transformers/pull/41383.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41383.patch", "merged_at": "2025-10-10T13:53:28" }
# What does this PR do? This PR removes `tpu_num_cores` arguments in `TrainingArguments`. This arg is somewhat confusing because the value is either `1` or `None` which means that we are using all devices. Instead, we remove the arg to make it clearer. Users who wants to control the number of devices can use `TPU_NUM_DEVICES` as specified in xla docs. Note that this was only used in example scripts
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41383/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41383/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41382
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41382/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41382/comments
https://api.github.com/repos/huggingface/transformers/issues/41382/events
https://github.com/huggingface/transformers/pull/41382
3,488,043,479
PR_kwDOCUB6oc6sUA7h
41,382
Remove `local_rank` arg from `TrainingArguments`
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T16:16:29
2025-10-10T09:34:14
2025-10-09T16:54:12
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41382", "html_url": "https://github.com/huggingface/transformers/pull/41382", "diff_url": "https://github.com/huggingface/transformers/pull/41382.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41382.patch", "merged_at": "2025-10-09T16:54:12" }
# What does this PR do? This PR removes `local_rank` from `TrainingArguments`. This arg can be safely replaced by `local_process_index`. This arg was never meant to be set and was supposed to be there to be overwritten by torch distributed.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41382/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41382/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41381
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41381/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41381/comments
https://api.github.com/repos/huggingface/transformers/issues/41381/events
https://github.com/huggingface/transformers/pull/41381
3,487,987,109
PR_kwDOCUB6oc6sT0XT
41,381
Fix overriding common_kwargs defaults in processor calls
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T16:04:05
2025-10-08T03:13:57
2025-10-08T03:13:56
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41381", "html_url": "https://github.com/huggingface/transformers/pull/41381", "diff_url": "https://github.com/huggingface/transformers/pull/41381.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41381.patch", "merged_at": "2025-10-08T03:13:56" }
# What does this PR do? After https://github.com/huggingface/transformers/pull/40931, the `common_kwargs` (`return_tensors` really) defaults were set after updating them with call kwargs, which mean it's impossible to override the defaults. This fixes it by setting the common kwargs defaults before updating them with call kwargs, and fixes a test in `test_image_processing_owlv2.py` that was failing because of it (as Owlv2 sets return_tensors to `np` by default for some reasons) Cc @zucchini-nlp to make sure I'm not missing anything!
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41381/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41381/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41380
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41380/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41380/comments
https://api.github.com/repos/huggingface/transformers/issues/41380/events
https://github.com/huggingface/transformers/pull/41380
3,487,941,425
PR_kwDOCUB6oc6sTqcM
41,380
RT-Detr correct 2d positional embeddings for non-square images
{ "login": "konstantinos-p", "id": 23555352, "node_id": "MDQ6VXNlcjIzNTU1MzUy", "avatar_url": "https://avatars.githubusercontent.com/u/23555352?v=4", "gravatar_id": "", "url": "https://api.github.com/users/konstantinos-p", "html_url": "https://github.com/konstantinos-p", "followers_url": "https://api.github.com/users/konstantinos-p/followers", "following_url": "https://api.github.com/users/konstantinos-p/following{/other_user}", "gists_url": "https://api.github.com/users/konstantinos-p/gists{/gist_id}", "starred_url": "https://api.github.com/users/konstantinos-p/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/konstantinos-p/subscriptions", "organizations_url": "https://api.github.com/users/konstantinos-p/orgs", "repos_url": "https://api.github.com/users/konstantinos-p/repos", "events_url": "https://api.github.com/users/konstantinos-p/events{/privacy}", "received_events_url": "https://api.github.com/users/konstantinos-p/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T15:52:19
2025-10-09T15:58:22
2025-10-09T15:58:22
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41380", "html_url": "https://github.com/huggingface/transformers/pull/41380", "diff_url": "https://github.com/huggingface/transformers/pull/41380.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41380.patch", "merged_at": "2025-10-09T15:58:22" }
# What does this PR do? This PR solves a subtle bug in how 2d positional embeddings are generated in the RT-Detr architecture. After flattening the meshgrid that generated the 2d positional embeddings, and for non-square images, the positional embeddings no longer encode the correct 2d positional information. For an image with width=4 and height=3 we get: ```python import torch # Create 2d grid x = torch.tensor([0, 1, 2, 3]) y = torch.tensor([0, 1, 2]) grid_x, grid_y = torch.meshgrid(x, y, indexing='ij') print(grid_x) #tensor([[0, 0, 0], # [1, 1, 1], # [2, 2, 2], # [3, 3, 3]]) print(grid_y) #tensor([[0, 1, 2], # [0, 1, 2], # [0, 1, 2], # [0, 1, 2]]) # Create image image = torch.range(0,11).reshape((3,4)).int() print(image) #tensor([[ 0, 1, 2, 3], # [ 4, 5, 6, 7], # [ 8, 9, 10, 11]], dtype=torch.int32) #Print flattened grid and image print(grid_x.flatten()) # tensor([0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3]) print(grid_y.flatten()) # tensor([0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2]) print(image.flatten()) # tensor([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]) ``` Note how the first four pixels 0,1,2,3 get x positions 0,0,0,1 instead of the intended 0,0,0,0. # Solution The proposed solution is consistent with the current way that positional embeddings are generated, while generalizing correctly to non-square shapes. Fixes # (issue) #41379 ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. Models: - vision models: @yonigozlan @molbap @qubvel
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41380/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41380/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41379
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41379/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41379/comments
https://api.github.com/repos/huggingface/transformers/issues/41379/events
https://github.com/huggingface/transformers/issues/41379
3,487,913,434
I_kwDOCUB6oc7P5VXa
41,379
RT-Detr 2d positional embedding bug
{ "login": "konstantinos-p", "id": 23555352, "node_id": "MDQ6VXNlcjIzNTU1MzUy", "avatar_url": "https://avatars.githubusercontent.com/u/23555352?v=4", "gravatar_id": "", "url": "https://api.github.com/users/konstantinos-p", "html_url": "https://github.com/konstantinos-p", "followers_url": "https://api.github.com/users/konstantinos-p/followers", "following_url": "https://api.github.com/users/konstantinos-p/following{/other_user}", "gists_url": "https://api.github.com/users/konstantinos-p/gists{/gist_id}", "starred_url": "https://api.github.com/users/konstantinos-p/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/konstantinos-p/subscriptions", "organizations_url": "https://api.github.com/users/konstantinos-p/orgs", "repos_url": "https://api.github.com/users/konstantinos-p/repos", "events_url": "https://api.github.com/users/konstantinos-p/events{/privacy}", "received_events_url": "https://api.github.com/users/konstantinos-p/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 3817266200, "node_id": "MDU6TGFiZWwzODE3MjY2MjAw", "url": "https://api.github.com/repos/huggingface/transformers/labels/bug", "name": "bug", "color": "d73a4a", "default": true, "description": null }, { "id": 5769473378, "node_id": "LA_kwDOCUB6oc8AAAABV-MtYg", "url": "https://api.github.com/repos/huggingface/transformers/labels/Vision", "name": "Vision", "color": "C079EF", "default": false, "description": "" } ]
open
false
null
[]
null
[]
2025-10-06T15:45:07
2025-10-08T05:34:59
null
CONTRIBUTOR
null
null
null
null
I believe that there is a subtle bug in how 2d positional embeddings are generated for the rt_detr model. Specifically, when generating a meshgrid and then flattening, the positional embeddings no longer match the intended image pixels. We can see this in the script below ```python import torch # Create 2d grid x = torch.tensor([0, 1, 2, 3]) y = torch.tensor([0, 1, 2]) grid_x, grid_y = torch.meshgrid(x, y, indexing='ij') print(grid_x) #tensor([[0, 0, 0], # [1, 1, 1], # [2, 2, 2], # [3, 3, 3]]) print(grid_y) #tensor([[0, 1, 2], # [0, 1, 2], # [0, 1, 2], # [0, 1, 2]]) # Create image image = torch.range(0,11).reshape((3,4)).int() print(image) #tensor([[ 0, 1, 2, 3], # [ 4, 5, 6, 7], # [ 8, 9, 10, 11]], dtype=torch.int32) #Print flattened grid and image print(grid_x.flatten()) # tensor([0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3]) print(grid_y.flatten()) # tensor([0, 1, 2, 0, 1, 2, 0, 1, 2, 0, 1, 2]) print(image.flatten()) # tensor([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]) ``` This means that for images that are not square the embeddings don't encode the intended spatial information. This issue comes from the following line. I believe setting indexing='xy' would have been the correct choice. https://github.com/huggingface/transformers/blob/fa36c973fc5f2f473f58013097e21ed74025ea42/src/transformers/models/rt_detr/modeling_rt_detr.py#L1154 While RT-Detr was intended by the authors to work on square images, generalizing correctly to non-square images could be important for future research. I will make a PR with a simple fix, that is consistent with the current positional embeddings but generalizes correctly to variable image sizes.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41379/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41379/timeline
null
null
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
false
https://api.github.com/repos/huggingface/transformers/issues/41378
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41378/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41378/comments
https://api.github.com/repos/huggingface/transformers/issues/41378/events
https://github.com/huggingface/transformers/pull/41378
3,487,910,998
PR_kwDOCUB6oc6sTjsP
41,378
🚨🚨 Remove all traces of legacy cache format
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T15:44:30
2025-10-08T09:14:45
2025-10-08T09:14:44
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41378", "html_url": "https://github.com/huggingface/transformers/pull/41378", "diff_url": "https://github.com/huggingface/transformers/pull/41378.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41378.patch", "merged_at": "2025-10-08T09:14:44" }
# What does this PR do? As per the title. Remove all legacy format and deprecated classes/methods.
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41378/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41378/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41377
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41377/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41377/comments
https://api.github.com/repos/huggingface/transformers/issues/41377/events
https://github.com/huggingface/transformers/pull/41377
3,487,898,000
PR_kwDOCUB6oc6sTgzn
41,377
[testing] Fix `JetMoeIntegrationTest`
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T15:41:10
2025-10-08T13:11:54
2025-10-08T13:11:54
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41377", "html_url": "https://github.com/huggingface/transformers/pull/41377", "diff_url": "https://github.com/huggingface/transformers/pull/41377.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41377.patch", "merged_at": "2025-10-08T13:11:54" }
# What does this PR do? > tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeIntegrationTest::test_model_8b_generation cause the pytest process being killed (and sometimes hangs for hours), see the log below. It loads the model on `cpu`. Let's use "auto". The outputs changes after #41324 (which is a fix of #40132), so the tests still fail but not causing the pytest process being killed and job hanging. There is a new warning showing > The attention mask is not set and cannot be inferred from input because pad token is same as eos token. As a consequence, you may observe unexpected behavior. Please pass your input's `attention_mask` to obtain reliable results. cc @ArthurZucker for this output change issue. Log when running before this PR: > 2025-09-30T16:33:22.8891725Z tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeIntegrationTest::test_model_8b_batched_generation 2025-09-30T16:33:22.8892381Z -------------------------------- live log call --------------------------------- 2025-09-30T16:33:22.8893314Z WARNING transformers.generation.configuration_utils:logging.py:328 The following generation flags are not valid and may be ignored: ['temperature']. Set `TRANSFORMERS_VERBOSITY=info` for more details. 2025-09-30T16:33:23.3206200Z FAILED [ 99%] 2025-09-30T16:39:21.4805460Z Killed 2025-09-30T16:39:21.4847256Z tests/models/jetmoe/test_modeling_jetmoe.py::JetMoeIntegrationTest::test_model_8b_generation
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41377/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41377/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41376
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41376/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41376/comments
https://api.github.com/repos/huggingface/transformers/issues/41376/events
https://github.com/huggingface/transformers/pull/41376
3,487,837,985
PR_kwDOCUB6oc6sTTZ4
41,376
deprecate `jit_mode_eval`
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T15:25:58
2025-10-09T16:50:48
2025-10-09T16:50:46
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41376", "html_url": "https://github.com/huggingface/transformers/pull/41376", "diff_url": "https://github.com/huggingface/transformers/pull/41376.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41376.patch", "merged_at": "2025-10-09T16:50:46" }
# What does this PR do? This PR deprecates `jit_mode_eval` arguments. Going on, it is better to fully rely on torch compile.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41376/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41376/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41375
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41375/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41375/comments
https://api.github.com/repos/huggingface/transformers/issues/41375/events
https://github.com/huggingface/transformers/pull/41375
3,487,725,383
PR_kwDOCUB6oc6sS6f_
41,375
`report_to` default changed to "none" + cleaning deprecated env var
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T15:00:27
2025-10-29T13:04:01
2025-10-09T16:28:48
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41375", "html_url": "https://github.com/huggingface/transformers/pull/41375", "diff_url": "https://github.com/huggingface/transformers/pull/41375.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41375.patch", "merged_at": "2025-10-09T16:28:48" }
# What does this PR do? This PR changes the default for reporting to "none" as we move to v5. Moreover, we clean some deprecations messages around some env env + we deprecate `logging_dir` that was solely used for tensorboard. Instead the user needs to pass a global variable.
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41375/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41375/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41374
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41374/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41374/comments
https://api.github.com/repos/huggingface/transformers/issues/41374/events
https://github.com/huggingface/transformers/pull/41374
3,487,542,260
PR_kwDOCUB6oc6sSRqY
41,374
Correct numerical regression in vision embeddings
{ "login": "i3hz", "id": 144821361, "node_id": "U_kgDOCKHMcQ", "avatar_url": "https://avatars.githubusercontent.com/u/144821361?v=4", "gravatar_id": "", "url": "https://api.github.com/users/i3hz", "html_url": "https://github.com/i3hz", "followers_url": "https://api.github.com/users/i3hz/followers", "following_url": "https://api.github.com/users/i3hz/following{/other_user}", "gists_url": "https://api.github.com/users/i3hz/gists{/gist_id}", "starred_url": "https://api.github.com/users/i3hz/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/i3hz/subscriptions", "organizations_url": "https://api.github.com/users/i3hz/orgs", "repos_url": "https://api.github.com/users/i3hz/repos", "events_url": "https://api.github.com/users/i3hz/events{/privacy}", "received_events_url": "https://api.github.com/users/i3hz/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T14:21:06
2025-10-07T11:43:24
2025-10-07T11:43:24
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41374", "html_url": "https://github.com/huggingface/transformers/pull/41374", "diff_url": "https://github.com/huggingface/transformers/pull/41374.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41374.patch", "merged_at": "2025-10-07T11:43:24" }
# What does this PR do? This PR fixes a numerical regression bug in the vision positional embedding calculation that was introduced between `transformers` versions v4.54.1 and v4.55.1. The original change was made to improve exportability but resulted in a slight floating point difference. The fix was applied to the base modular files (`modular_idefics2.py` and `modular_idefics3.py`) and then propagated to all dependent models via `make fix-copies`. Fixes #41190 ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [x] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. https://github.com/huggingface/transformers/issues/41190 - [ ] Did you make sure to update the documentation with your changes? - [ ] Did you write any new necessary tests? ## Who can review @zucchini-nlp
{ "login": "zucchini-nlp", "id": 100715397, "node_id": "U_kgDOBgDLhQ", "avatar_url": "https://avatars.githubusercontent.com/u/100715397?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zucchini-nlp", "html_url": "https://github.com/zucchini-nlp", "followers_url": "https://api.github.com/users/zucchini-nlp/followers", "following_url": "https://api.github.com/users/zucchini-nlp/following{/other_user}", "gists_url": "https://api.github.com/users/zucchini-nlp/gists{/gist_id}", "starred_url": "https://api.github.com/users/zucchini-nlp/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/zucchini-nlp/subscriptions", "organizations_url": "https://api.github.com/users/zucchini-nlp/orgs", "repos_url": "https://api.github.com/users/zucchini-nlp/repos", "events_url": "https://api.github.com/users/zucchini-nlp/events{/privacy}", "received_events_url": "https://api.github.com/users/zucchini-nlp/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41374/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41374/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41373
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41373/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41373/comments
https://api.github.com/repos/huggingface/transformers/issues/41373/events
https://github.com/huggingface/transformers/pull/41373
3,487,496,797
PR_kwDOCUB6oc6sSHm1
41,373
[testing] reduce runtime of `HunYuanMoEV1IntegrationTest:test_model_generation`
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T14:10:46
2025-10-10T20:27:03
2025-10-10T20:27:01
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41373", "html_url": "https://github.com/huggingface/transformers/pull/41373", "diff_url": "https://github.com/huggingface/transformers/pull/41373.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41373.patch", "merged_at": "2025-10-10T20:27:01" }
# What does this PR do? This test takes 26 minutes to run, way too much to be helpful. Let's reduce `max_new_tokens` to `10` and runs in 7 minutes.
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41373/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41373/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41372
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41372/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41372/comments
https://api.github.com/repos/huggingface/transformers/issues/41372/events
https://github.com/huggingface/transformers/pull/41372
3,487,471,349
PR_kwDOCUB6oc6sSB-C
41,372
Enable modular files from other libraries
{ "login": "regisss", "id": 15324346, "node_id": "MDQ6VXNlcjE1MzI0MzQ2", "avatar_url": "https://avatars.githubusercontent.com/u/15324346?v=4", "gravatar_id": "", "url": "https://api.github.com/users/regisss", "html_url": "https://github.com/regisss", "followers_url": "https://api.github.com/users/regisss/followers", "following_url": "https://api.github.com/users/regisss/following{/other_user}", "gists_url": "https://api.github.com/users/regisss/gists{/gist_id}", "starred_url": "https://api.github.com/users/regisss/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/regisss/subscriptions", "organizations_url": "https://api.github.com/users/regisss/orgs", "repos_url": "https://api.github.com/users/regisss/repos", "events_url": "https://api.github.com/users/regisss/events{/privacy}", "received_events_url": "https://api.github.com/users/regisss/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T14:05:08
2025-10-13T11:48:39
2025-10-13T11:48:32
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41372", "html_url": "https://github.com/huggingface/transformers/pull/41372", "diff_url": "https://github.com/huggingface/transformers/pull/41372.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41372.patch", "merged_at": "2025-10-13T11:48:32" }
# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> This PR extends the modular model converter so that one can use a modular file in an external library. The generated files (modeling, configuration, etc) will located in the same folder as the modular file. This requires users to use the new `--source-library` flag. Another PR will come so that one can use model classes that are outside of the Transformers library in their modular file. ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "regisss", "id": 15324346, "node_id": "MDQ6VXNlcjE1MzI0MzQ2", "avatar_url": "https://avatars.githubusercontent.com/u/15324346?v=4", "gravatar_id": "", "url": "https://api.github.com/users/regisss", "html_url": "https://github.com/regisss", "followers_url": "https://api.github.com/users/regisss/followers", "following_url": "https://api.github.com/users/regisss/following{/other_user}", "gists_url": "https://api.github.com/users/regisss/gists{/gist_id}", "starred_url": "https://api.github.com/users/regisss/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/regisss/subscriptions", "organizations_url": "https://api.github.com/users/regisss/orgs", "repos_url": "https://api.github.com/users/regisss/repos", "events_url": "https://api.github.com/users/regisss/events{/privacy}", "received_events_url": "https://api.github.com/users/regisss/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41372/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 1, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41372/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41371
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41371/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41371/comments
https://api.github.com/repos/huggingface/transformers/issues/41371/events
https://github.com/huggingface/transformers/pull/41371
3,487,300,353
PR_kwDOCUB6oc6sRbzv
41,371
Correcting numerical regression in vision embeddings
{ "login": "i3hz", "id": 144821361, "node_id": "U_kgDOCKHMcQ", "avatar_url": "https://avatars.githubusercontent.com/u/144821361?v=4", "gravatar_id": "", "url": "https://api.github.com/users/i3hz", "html_url": "https://github.com/i3hz", "followers_url": "https://api.github.com/users/i3hz/followers", "following_url": "https://api.github.com/users/i3hz/following{/other_user}", "gists_url": "https://api.github.com/users/i3hz/gists{/gist_id}", "starred_url": "https://api.github.com/users/i3hz/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/i3hz/subscriptions", "organizations_url": "https://api.github.com/users/i3hz/orgs", "repos_url": "https://api.github.com/users/i3hz/repos", "events_url": "https://api.github.com/users/i3hz/events{/privacy}", "received_events_url": "https://api.github.com/users/i3hz/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T13:27:38
2025-10-06T13:33:23
2025-10-06T13:30:34
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41371", "html_url": "https://github.com/huggingface/transformers/pull/41371", "diff_url": "https://github.com/huggingface/transformers/pull/41371.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41371.patch", "merged_at": null }
# What does this PR do? This PR fixes a numerical regression bug in the vision positional embedding calculation that was introduced between `transformers` versions v4.54.1 and v4.55.1. The original change was made to improve exportability but resulted in a slight floating point difference, causing models like `Idefics2`, `Idefics3`, and `SmolVLM` to produce different outputs than the previous version. The fix was applied to the base modular files (`modular_idefics2.py` and `modular_idefics3.py`) and then propagated to all dependent models via `make fix-copies`. Fixes #41190
{ "login": "i3hz", "id": 144821361, "node_id": "U_kgDOCKHMcQ", "avatar_url": "https://avatars.githubusercontent.com/u/144821361?v=4", "gravatar_id": "", "url": "https://api.github.com/users/i3hz", "html_url": "https://github.com/i3hz", "followers_url": "https://api.github.com/users/i3hz/followers", "following_url": "https://api.github.com/users/i3hz/following{/other_user}", "gists_url": "https://api.github.com/users/i3hz/gists{/gist_id}", "starred_url": "https://api.github.com/users/i3hz/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/i3hz/subscriptions", "organizations_url": "https://api.github.com/users/i3hz/orgs", "repos_url": "https://api.github.com/users/i3hz/repos", "events_url": "https://api.github.com/users/i3hz/events{/privacy}", "received_events_url": "https://api.github.com/users/i3hz/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41371/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41371/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41370
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41370/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41370/comments
https://api.github.com/repos/huggingface/transformers/issues/41370/events
https://github.com/huggingface/transformers/pull/41370
3,487,261,079
PR_kwDOCUB6oc6sRTLF
41,370
[`CB`] Refactors the way we access paged
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T13:18:41
2025-10-06T15:55:33
2025-10-06T15:55:32
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41370", "html_url": "https://github.com/huggingface/transformers/pull/41370", "diff_url": "https://github.com/huggingface/transformers/pull/41370.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41370.patch", "merged_at": "2025-10-06T15:55:32" }
# What does this PR do? The user decides on the function, but CB handles itself which "interface" wrapper to use. This should make stuff easier
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41370/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41370/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41369
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41369/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41369/comments
https://api.github.com/repos/huggingface/transformers/issues/41369/events
https://github.com/huggingface/transformers/pull/41369
3,487,210,499
PR_kwDOCUB6oc6sRH6g
41,369
Add logits_to_keep parameter to BertForSequenceClassification
{ "login": "Kraga922", "id": 138029547, "node_id": "U_kgDOCDop6w", "avatar_url": "https://avatars.githubusercontent.com/u/138029547?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Kraga922", "html_url": "https://github.com/Kraga922", "followers_url": "https://api.github.com/users/Kraga922/followers", "following_url": "https://api.github.com/users/Kraga922/following{/other_user}", "gists_url": "https://api.github.com/users/Kraga922/gists{/gist_id}", "starred_url": "https://api.github.com/users/Kraga922/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Kraga922/subscriptions", "organizations_url": "https://api.github.com/users/Kraga922/orgs", "repos_url": "https://api.github.com/users/Kraga922/repos", "events_url": "https://api.github.com/users/Kraga922/events{/privacy}", "received_events_url": "https://api.github.com/users/Kraga922/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T13:08:07
2025-10-06T13:46:29
2025-10-06T13:46:28
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41369", "html_url": "https://github.com/huggingface/transformers/pull/41369", "diff_url": "https://github.com/huggingface/transformers/pull/41369.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41369.patch", "merged_at": null }
## Summary This PR adds support for the `logits_to_keep` parameter to `BertForSequenceClassification` ## Motivation The `logits_to_keep` parameter enables memory-efficient inference by computing logits only for the last N token positions. This optimization is particularly useful for: - Generation tasks where only recent tokens are needed - Memory-constrained environments - Processing long sequences efficiently ## Implementation Details - **Added** `logits_to_keep` parameter to the `forward()` method signature - **Implemented** token slicing logic that activates when parameter is specified - **Updated** docstring with comprehensive parameter documentation - **Maintains** full backward compatibility (parameter defaults to `None`) ## Code Changes The implementation adds approximately 10 lines of code: 1. New optional parameter in function signature 2. Conditional slicing logic for sequence outputs 3. Documentation in docstring ## Testing Comprehensive local testing performed with `bert-base-uncased`: - Single input and batch processing verified - Multiple `logits_to_keep` values tested (1, 2, 3, 5, 10, None) - Edge cases handled (0, negative values, values > sequence length) - Training compatibility confirmed (gradient flow works) - Output structure validation passed - Small performance improvement observed (~0.6% faster) ## Backward Compatibility - Parameter is optional and defaults to `None` - When `None`, behavior is identical to current implementation - No breaking changes to existing code ## Related Issues Inspired by #40984
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41369/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41369/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41368
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41368/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41368/comments
https://api.github.com/repos/huggingface/transformers/issues/41368/events
https://github.com/huggingface/transformers/pull/41368
3,487,131,977
PR_kwDOCUB6oc6sQ2cX
41,368
[testing] update `test_longcat_generation_cpu`
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T12:51:12
2025-10-06T13:21:30
2025-10-06T13:21:29
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41368", "html_url": "https://github.com/huggingface/transformers/pull/41368", "diff_url": "https://github.com/huggingface/transformers/pull/41368.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41368.patch", "merged_at": "2025-10-06T13:21:29" }
# What does this PR do? Use "auto" instead of "cpu" to avoid the pytest process being killed. (#40132 has issue with using "auto" for this model, and we still need an fix for that) Also using `max_new_tokens=3` to run this test within 8 minutes (on a single A10 CI runner) The expected output is not updated yet: will do it after the above mentioned issue is fixed.
{ "login": "ydshieh", "id": 2521628, "node_id": "MDQ6VXNlcjI1MjE2Mjg=", "avatar_url": "https://avatars.githubusercontent.com/u/2521628?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ydshieh", "html_url": "https://github.com/ydshieh", "followers_url": "https://api.github.com/users/ydshieh/followers", "following_url": "https://api.github.com/users/ydshieh/following{/other_user}", "gists_url": "https://api.github.com/users/ydshieh/gists{/gist_id}", "starred_url": "https://api.github.com/users/ydshieh/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ydshieh/subscriptions", "organizations_url": "https://api.github.com/users/ydshieh/orgs", "repos_url": "https://api.github.com/users/ydshieh/repos", "events_url": "https://api.github.com/users/ydshieh/events{/privacy}", "received_events_url": "https://api.github.com/users/ydshieh/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41368/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41368/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41367
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41367/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41367/comments
https://api.github.com/repos/huggingface/transformers/issues/41367/events
https://github.com/huggingface/transformers/pull/41367
3,487,111,841
PR_kwDOCUB6oc6sQyBL
41,367
🚨 Remove BetterTransformer
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T12:46:20
2025-10-06T13:18:14
2025-10-06T13:18:12
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41367", "html_url": "https://github.com/huggingface/transformers/pull/41367", "diff_url": "https://github.com/huggingface/transformers/pull/41367.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41367.patch", "merged_at": "2025-10-06T13:18:12" }
# What does this PR do? This API is completely out-of-date and superseded by everything we did on attention computation
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41367/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 1, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41367/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41366
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41366/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41366/comments
https://api.github.com/repos/huggingface/transformers/issues/41366/events
https://github.com/huggingface/transformers/pull/41366
3,487,093,565
PR_kwDOCUB6oc6sQt9b
41,366
Add in-out modalities as class attribute per model
{ "login": "zucchini-nlp", "id": 100715397, "node_id": "U_kgDOBgDLhQ", "avatar_url": "https://avatars.githubusercontent.com/u/100715397?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zucchini-nlp", "html_url": "https://github.com/zucchini-nlp", "followers_url": "https://api.github.com/users/zucchini-nlp/followers", "following_url": "https://api.github.com/users/zucchini-nlp/following{/other_user}", "gists_url": "https://api.github.com/users/zucchini-nlp/gists{/gist_id}", "starred_url": "https://api.github.com/users/zucchini-nlp/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/zucchini-nlp/subscriptions", "organizations_url": "https://api.github.com/users/zucchini-nlp/orgs", "repos_url": "https://api.github.com/users/zucchini-nlp/repos", "events_url": "https://api.github.com/users/zucchini-nlp/events{/privacy}", "received_events_url": "https://api.github.com/users/zucchini-nlp/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T12:41:23
2025-10-16T15:11:06
2025-10-16T15:11:06
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41366", "html_url": "https://github.com/huggingface/transformers/pull/41366", "diff_url": "https://github.com/huggingface/transformers/pull/41366.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41366.patch", "merged_at": "2025-10-16T15:11:06" }
# What does this PR do? Branches out from https://github.com/huggingface/transformers/pull/40884#discussion_r2400964015 to make review and merge faster Adds a class attr for each model to indicate the supported input and output modalities. Out modalities will be `None` in case the model is not generative and "text" in most other cases. We have only a few models that can generate audio and image in the output. Note that for encoder decoder models that whisper the input modalities will contain both encoder ("audio") and decoder ("text") modalities This will be used firstly for the pipeline and we can extend usage later to better testing suite and in preparing inputs better in generation with multimodal LLMs (e.g. if we move multimodal encoding to `GenrationMixin._prepare_multimodal_encodings`). No test added at this point, because there is nothing to test
{ "login": "zucchini-nlp", "id": 100715397, "node_id": "U_kgDOBgDLhQ", "avatar_url": "https://avatars.githubusercontent.com/u/100715397?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zucchini-nlp", "html_url": "https://github.com/zucchini-nlp", "followers_url": "https://api.github.com/users/zucchini-nlp/followers", "following_url": "https://api.github.com/users/zucchini-nlp/following{/other_user}", "gists_url": "https://api.github.com/users/zucchini-nlp/gists{/gist_id}", "starred_url": "https://api.github.com/users/zucchini-nlp/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/zucchini-nlp/subscriptions", "organizations_url": "https://api.github.com/users/zucchini-nlp/orgs", "repos_url": "https://api.github.com/users/zucchini-nlp/repos", "events_url": "https://api.github.com/users/zucchini-nlp/events{/privacy}", "received_events_url": "https://api.github.com/users/zucchini-nlp/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41366/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41366/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41365
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41365/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41365/comments
https://api.github.com/repos/huggingface/transformers/issues/41365/events
https://github.com/huggingface/transformers/pull/41365
3,486,856,239
PR_kwDOCUB6oc6sP59o
41,365
Update CONTRIBUTING.md
{ "login": "Karthikg998", "id": 178138088, "node_id": "U_kgDOCp4r6A", "avatar_url": "https://avatars.githubusercontent.com/u/178138088?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Karthikg998", "html_url": "https://github.com/Karthikg998", "followers_url": "https://api.github.com/users/Karthikg998/followers", "following_url": "https://api.github.com/users/Karthikg998/following{/other_user}", "gists_url": "https://api.github.com/users/Karthikg998/gists{/gist_id}", "starred_url": "https://api.github.com/users/Karthikg998/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Karthikg998/subscriptions", "organizations_url": "https://api.github.com/users/Karthikg998/orgs", "repos_url": "https://api.github.com/users/Karthikg998/repos", "events_url": "https://api.github.com/users/Karthikg998/events{/privacy}", "received_events_url": "https://api.github.com/users/Karthikg998/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T11:36:23
2025-10-06T13:47:18
2025-10-06T13:44:42
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41365", "html_url": "https://github.com/huggingface/transformers/pull/41365", "diff_url": "https://github.com/huggingface/transformers/pull/41365.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41365.patch", "merged_at": null }
This commit is to let know first timers that reviews are normal # What does this PR do? This PR adds an acknowledgment/feedback loop message to guide first-time contributors. It explicitly tells them: “Your PR may get small suggestions; don’t worry — it’s normal." This helps reduce discouragement for beginners and encourages them to continue contributing confidently. It is important because many first-time contributors abandon their PRs when they receive minor review comments, not realizing that this is a standard part of the process. By adding this guidance, we make the contribution experience more welcoming and supportive. ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41365/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41365/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41364
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41364/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41364/comments
https://api.github.com/repos/huggingface/transformers/issues/41364/events
https://github.com/huggingface/transformers/pull/41364
3,486,735,273
PR_kwDOCUB6oc6sPfDn
41,364
Change RT-Detr docs to reflect fixed 640x640 input size
{ "login": "konstantinos-p", "id": 23555352, "node_id": "MDQ6VXNlcjIzNTU1MzUy", "avatar_url": "https://avatars.githubusercontent.com/u/23555352?v=4", "gravatar_id": "", "url": "https://api.github.com/users/konstantinos-p", "html_url": "https://github.com/konstantinos-p", "followers_url": "https://api.github.com/users/konstantinos-p/followers", "following_url": "https://api.github.com/users/konstantinos-p/following{/other_user}", "gists_url": "https://api.github.com/users/konstantinos-p/gists{/gist_id}", "starred_url": "https://api.github.com/users/konstantinos-p/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/konstantinos-p/subscriptions", "organizations_url": "https://api.github.com/users/konstantinos-p/orgs", "repos_url": "https://api.github.com/users/konstantinos-p/repos", "events_url": "https://api.github.com/users/konstantinos-p/events{/privacy}", "received_events_url": "https://api.github.com/users/konstantinos-p/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T11:05:25
2025-10-09T14:29:59
2025-10-09T14:29:17
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41364", "html_url": "https://github.com/huggingface/transformers/pull/41364", "diff_url": "https://github.com/huggingface/transformers/pull/41364.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41364.patch", "merged_at": "2025-10-09T14:29:16" }
# What does this PR do? The authors of RT-Detr mention that the model was trained on 640x640 images and was meant to be used for inference on 640x640 images. Also, the current implementation has certain quirks that make training/inferring on images of different sizes problematic. For example, the pixel masks used for batching images of varying sizes are discarded. I've added a few lines in the docs to notify the user about these issues. `` proj_feats = [self.encoder_input_proj[level](source) for level, (source, mask) in enumerate(features)] `` Fixes # (issue) #41363 ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. Models: - vision models: @yonigozlan @molbap - documentation: @stevhliu
{ "login": "yonigozlan", "id": 74535834, "node_id": "MDQ6VXNlcjc0NTM1ODM0", "avatar_url": "https://avatars.githubusercontent.com/u/74535834?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yonigozlan", "html_url": "https://github.com/yonigozlan", "followers_url": "https://api.github.com/users/yonigozlan/followers", "following_url": "https://api.github.com/users/yonigozlan/following{/other_user}", "gists_url": "https://api.github.com/users/yonigozlan/gists{/gist_id}", "starred_url": "https://api.github.com/users/yonigozlan/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/yonigozlan/subscriptions", "organizations_url": "https://api.github.com/users/yonigozlan/orgs", "repos_url": "https://api.github.com/users/yonigozlan/repos", "events_url": "https://api.github.com/users/yonigozlan/events{/privacy}", "received_events_url": "https://api.github.com/users/yonigozlan/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41364/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41364/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41363
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41363/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41363/comments
https://api.github.com/repos/huggingface/transformers/issues/41363/events
https://github.com/huggingface/transformers/issues/41363
3,486,732,144
I_kwDOCUB6oc7P009w
41,363
RT-Detr docs should reflect fixed 640x640 input size
{ "login": "konstantinos-p", "id": 23555352, "node_id": "MDQ6VXNlcjIzNTU1MzUy", "avatar_url": "https://avatars.githubusercontent.com/u/23555352?v=4", "gravatar_id": "", "url": "https://api.github.com/users/konstantinos-p", "html_url": "https://github.com/konstantinos-p", "followers_url": "https://api.github.com/users/konstantinos-p/followers", "following_url": "https://api.github.com/users/konstantinos-p/following{/other_user}", "gists_url": "https://api.github.com/users/konstantinos-p/gists{/gist_id}", "starred_url": "https://api.github.com/users/konstantinos-p/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/konstantinos-p/subscriptions", "organizations_url": "https://api.github.com/users/konstantinos-p/orgs", "repos_url": "https://api.github.com/users/konstantinos-p/repos", "events_url": "https://api.github.com/users/konstantinos-p/events{/privacy}", "received_events_url": "https://api.github.com/users/konstantinos-p/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 1834067346, "node_id": "MDU6TGFiZWwxODM0MDY3MzQ2", "url": "https://api.github.com/repos/huggingface/transformers/labels/Documentation", "name": "Documentation", "color": "77cc3b", "default": false, "description": "" } ]
open
false
null
[]
null
[]
2025-10-06T11:04:37
2025-10-08T19:01:15
null
CONTRIBUTOR
null
null
null
null
The authors of RT-Detr mention that the model was trained on 640x640 images and was meant to be used for inference on 640x640 images. Also, the current implementation has certain quirks that make training/inferring on images of different sizes problematic. For example, the pixel masks used for batching images of varying sizes are discarded. https://github.com/huggingface/transformers/blob/0452f28544f3626273d25f07f83c0e5f7da2d47a/src/transformers/models/rt_detr/modeling_rt_detr.py#L1645 The above are not clear in the current docs. I'll open a PR which adds a few lines in the docs to notify users about these issues.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41363/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41363/timeline
null
null
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
false
https://api.github.com/repos/huggingface/transformers/issues/41362
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41362/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41362/comments
https://api.github.com/repos/huggingface/transformers/issues/41362/events
https://github.com/huggingface/transformers/pull/41362
3,486,627,566
PR_kwDOCUB6oc6sPHWm
41,362
Added Hacktoberfest banner image to README.md
{ "login": "Akanksha-Patnaik", "id": 215271069, "node_id": "U_kgDODNTGnQ", "avatar_url": "https://avatars.githubusercontent.com/u/215271069?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Akanksha-Patnaik", "html_url": "https://github.com/Akanksha-Patnaik", "followers_url": "https://api.github.com/users/Akanksha-Patnaik/followers", "following_url": "https://api.github.com/users/Akanksha-Patnaik/following{/other_user}", "gists_url": "https://api.github.com/users/Akanksha-Patnaik/gists{/gist_id}", "starred_url": "https://api.github.com/users/Akanksha-Patnaik/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Akanksha-Patnaik/subscriptions", "organizations_url": "https://api.github.com/users/Akanksha-Patnaik/orgs", "repos_url": "https://api.github.com/users/Akanksha-Patnaik/repos", "events_url": "https://api.github.com/users/Akanksha-Patnaik/events{/privacy}", "received_events_url": "https://api.github.com/users/Akanksha-Patnaik/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-06T10:35:39
2025-10-06T10:35:39
null
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41362", "html_url": "https://github.com/huggingface/transformers/pull/41362", "diff_url": "https://github.com/huggingface/transformers/pull/41362.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41362.patch", "merged_at": null }
Added the Hacktoberfest 2025 banner image to the README.md file. This showcases the project’s participation in Hacktoberfest and encourages contributors to join. Fixes #41244
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41362/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41362/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41361
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41361/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41361/comments
https://api.github.com/repos/huggingface/transformers/issues/41361/events
https://github.com/huggingface/transformers/pull/41361
3,486,508,760
PR_kwDOCUB6oc6sOtfd
41,361
[`ModularChecker`] QOL for the modular checker
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T10:03:05
2025-10-06T10:52:11
2025-10-06T10:52:10
COLLABORATOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41361", "html_url": "https://github.com/huggingface/transformers/pull/41361", "diff_url": "https://github.com/huggingface/transformers/pull/41361.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41361.patch", "merged_at": "2025-10-06T10:52:10" }
# What does this PR do? - adds pretty print - adds simple model-name as file - adds release date parsing - adds highest overall match - optional Jaccard - snapshot download <img width="1724" height="1348" alt="image" src="https://github.com/user-attachments/assets/be01c840-cd27-47a7-ae41-d5f4bbb375c3" />
{ "login": "ArthurZucker", "id": 48595927, "node_id": "MDQ6VXNlcjQ4NTk1OTI3", "avatar_url": "https://avatars.githubusercontent.com/u/48595927?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ArthurZucker", "html_url": "https://github.com/ArthurZucker", "followers_url": "https://api.github.com/users/ArthurZucker/followers", "following_url": "https://api.github.com/users/ArthurZucker/following{/other_user}", "gists_url": "https://api.github.com/users/ArthurZucker/gists{/gist_id}", "starred_url": "https://api.github.com/users/ArthurZucker/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/ArthurZucker/subscriptions", "organizations_url": "https://api.github.com/users/ArthurZucker/orgs", "repos_url": "https://api.github.com/users/ArthurZucker/repos", "events_url": "https://api.github.com/users/ArthurZucker/events{/privacy}", "received_events_url": "https://api.github.com/users/ArthurZucker/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41361/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41361/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41360
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41360/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41360/comments
https://api.github.com/repos/huggingface/transformers/issues/41360/events
https://github.com/huggingface/transformers/pull/41360
3,486,437,056
PR_kwDOCUB6oc6sOeKl
41,360
Add granit speech doc
{ "login": "Deep-unlearning", "id": 58599908, "node_id": "MDQ6VXNlcjU4NTk5OTA4", "avatar_url": "https://avatars.githubusercontent.com/u/58599908?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Deep-unlearning", "html_url": "https://github.com/Deep-unlearning", "followers_url": "https://api.github.com/users/Deep-unlearning/followers", "following_url": "https://api.github.com/users/Deep-unlearning/following{/other_user}", "gists_url": "https://api.github.com/users/Deep-unlearning/gists{/gist_id}", "starred_url": "https://api.github.com/users/Deep-unlearning/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Deep-unlearning/subscriptions", "organizations_url": "https://api.github.com/users/Deep-unlearning/orgs", "repos_url": "https://api.github.com/users/Deep-unlearning/repos", "events_url": "https://api.github.com/users/Deep-unlearning/events{/privacy}", "received_events_url": "https://api.github.com/users/Deep-unlearning/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-06T09:43:27
2025-10-09T14:01:44
null
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41360", "html_url": "https://github.com/huggingface/transformers/pull/41360", "diff_url": "https://github.com/huggingface/transformers/pull/41360.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41360.patch", "merged_at": null }
# What does this PR do? Add usage example for the Granite Speech models @eustlb
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41360/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41360/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41359
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41359/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41359/comments
https://api.github.com/repos/huggingface/transformers/issues/41359/events
https://github.com/huggingface/transformers/pull/41359
3,486,357,027
PR_kwDOCUB6oc6sONBG
41,359
Fix trainer for py3.9
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 8103865784, "node_id": "LA_kwDOCUB6oc8AAAAB4wctuA", "url": "https://api.github.com/repos/huggingface/transformers/labels/for%20patch", "name": "for patch", "color": "D93F0B", "default": false, "description": "Tag issues / labels that should be included in the next patch" } ]
closed
false
null
[]
null
[]
2025-10-06T09:18:44
2025-10-23T13:42:14
2025-10-06T09:36:05
MEMBER
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41359", "html_url": "https://github.com/huggingface/transformers/pull/41359", "diff_url": "https://github.com/huggingface/transformers/pull/41359.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41359.patch", "merged_at": "2025-10-06T09:36:05" }
# What does this PR do? This PR fixes trainer compatibility with py3.9. To be added in a patch for the latest release. Asked by optimum folks cc @IlyasMoutawwakil
{ "login": "SunMarc", "id": 57196510, "node_id": "MDQ6VXNlcjU3MTk2NTEw", "avatar_url": "https://avatars.githubusercontent.com/u/57196510?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SunMarc", "html_url": "https://github.com/SunMarc", "followers_url": "https://api.github.com/users/SunMarc/followers", "following_url": "https://api.github.com/users/SunMarc/following{/other_user}", "gists_url": "https://api.github.com/users/SunMarc/gists{/gist_id}", "starred_url": "https://api.github.com/users/SunMarc/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/SunMarc/subscriptions", "organizations_url": "https://api.github.com/users/SunMarc/orgs", "repos_url": "https://api.github.com/users/SunMarc/repos", "events_url": "https://api.github.com/users/SunMarc/events{/privacy}", "received_events_url": "https://api.github.com/users/SunMarc/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41359/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41359/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41358
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41358/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41358/comments
https://api.github.com/repos/huggingface/transformers/issues/41358/events
https://github.com/huggingface/transformers/pull/41358
3,485,784,475
PR_kwDOCUB6oc6sMSVp
41,358
Remove unused function patameters
{ "login": "cyyever", "id": 17618148, "node_id": "MDQ6VXNlcjE3NjE4MTQ4", "avatar_url": "https://avatars.githubusercontent.com/u/17618148?v=4", "gravatar_id": "", "url": "https://api.github.com/users/cyyever", "html_url": "https://github.com/cyyever", "followers_url": "https://api.github.com/users/cyyever/followers", "following_url": "https://api.github.com/users/cyyever/following{/other_user}", "gists_url": "https://api.github.com/users/cyyever/gists{/gist_id}", "starred_url": "https://api.github.com/users/cyyever/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/cyyever/subscriptions", "organizations_url": "https://api.github.com/users/cyyever/orgs", "repos_url": "https://api.github.com/users/cyyever/repos", "events_url": "https://api.github.com/users/cyyever/events{/privacy}", "received_events_url": "https://api.github.com/users/cyyever/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T05:57:34
2025-10-06T23:43:41
2025-10-06T15:38:18
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41358", "html_url": "https://github.com/huggingface/transformers/pull/41358", "diff_url": "https://github.com/huggingface/transformers/pull/41358.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41358.patch", "merged_at": "2025-10-06T15:38:18" }
# What does this PR do? Remove unused arguments detected by pylint.
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41358/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41358/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41357
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41357/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41357/comments
https://api.github.com/repos/huggingface/transformers/issues/41357/events
https://github.com/huggingface/transformers/pull/41357
3,485,707,375
PR_kwDOCUB6oc6sMB0q
41,357
chore(docs): small README note (contrib test)
{ "login": "Shank312", "id": 219105003, "node_id": "U_kgDODQ9G6w", "avatar_url": "https://avatars.githubusercontent.com/u/219105003?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Shank312", "html_url": "https://github.com/Shank312", "followers_url": "https://api.github.com/users/Shank312/followers", "following_url": "https://api.github.com/users/Shank312/following{/other_user}", "gists_url": "https://api.github.com/users/Shank312/gists{/gist_id}", "starred_url": "https://api.github.com/users/Shank312/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Shank312/subscriptions", "organizations_url": "https://api.github.com/users/Shank312/orgs", "repos_url": "https://api.github.com/users/Shank312/repos", "events_url": "https://api.github.com/users/Shank312/events{/privacy}", "received_events_url": "https://api.github.com/users/Shank312/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-06T05:11:30
2025-10-06T13:17:38
2025-10-06T13:17:38
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41357", "html_url": "https://github.com/huggingface/transformers/pull/41357", "diff_url": "https://github.com/huggingface/transformers/pull/41357.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41357.patch", "merged_at": null }
# What does this PR do? Tiny docs improvement / test contribution by Shankar. This PR is a small test to verify contribution workflow. <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41357/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41357/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41356
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41356/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41356/comments
https://api.github.com/repos/huggingface/transformers/issues/41356/events
https://github.com/huggingface/transformers/pull/41356
3,485,189,552
PR_kwDOCUB6oc6sKWHJ
41,356
Add DEIMv2 model, image processor, and basic tests
{ "login": "Vinayaktoor", "id": 145313915, "node_id": "U_kgDOCKlQew", "avatar_url": "https://avatars.githubusercontent.com/u/145313915?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Vinayaktoor", "html_url": "https://github.com/Vinayaktoor", "followers_url": "https://api.github.com/users/Vinayaktoor/followers", "following_url": "https://api.github.com/users/Vinayaktoor/following{/other_user}", "gists_url": "https://api.github.com/users/Vinayaktoor/gists{/gist_id}", "starred_url": "https://api.github.com/users/Vinayaktoor/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Vinayaktoor/subscriptions", "organizations_url": "https://api.github.com/users/Vinayaktoor/orgs", "repos_url": "https://api.github.com/users/Vinayaktoor/repos", "events_url": "https://api.github.com/users/Vinayaktoor/events{/privacy}", "received_events_url": "https://api.github.com/users/Vinayaktoor/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-05T21:35:01
2025-10-07T22:03:19
null
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41356", "html_url": "https://github.com/huggingface/transformers/pull/41356", "diff_url": "https://github.com/huggingface/transformers/pull/41356.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41356.patch", "merged_at": null }
# What does this PR do? This PR introduces the DEIMv2 model along with its configuration, image processor, and unit tests. The changes include: Model Implementation: Added Deimv2ForObjectDetection supporting object detection tasks. Configuration: Added Deimv2Config for model hyperparameters. Image Processing: Added Deimv2ImageProcessor to handle preprocessing and postprocessing of images. Unit Tests: Added tests for configuration, image processing, and model forward pass (tests/models/deimv2/). This PR builds upon and makes changes to PR by @kekwlboy12469, PR #41291. Fixes #41211 Fixes missing or incomplete DEIMv2 model integration. Fixes absence of unit tests for DEIMv2 model and image processor. Fixes compatibility issues with Transformers pipelines for object detection. ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [x] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [x] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [x] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41356/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41356/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41355
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41355/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41355/comments
https://api.github.com/repos/huggingface/transformers/issues/41355/events
https://github.com/huggingface/transformers/pull/41355
3,485,057,727
PR_kwDOCUB6oc6sJ8aV
41,355
Better typehints for `apply_chat_template`
{ "login": "Samoed", "id": 36135455, "node_id": "MDQ6VXNlcjM2MTM1NDU1", "avatar_url": "https://avatars.githubusercontent.com/u/36135455?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Samoed", "html_url": "https://github.com/Samoed", "followers_url": "https://api.github.com/users/Samoed/followers", "following_url": "https://api.github.com/users/Samoed/following{/other_user}", "gists_url": "https://api.github.com/users/Samoed/gists{/gist_id}", "starred_url": "https://api.github.com/users/Samoed/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Samoed/subscriptions", "organizations_url": "https://api.github.com/users/Samoed/orgs", "repos_url": "https://api.github.com/users/Samoed/repos", "events_url": "https://api.github.com/users/Samoed/events{/privacy}", "received_events_url": "https://api.github.com/users/Samoed/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T18:42:12
2025-10-06T13:14:46
2025-10-06T13:14:03
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41355", "html_url": "https://github.com/huggingface/transformers/pull/41355", "diff_url": "https://github.com/huggingface/transformers/pull/41355.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41355.patch", "merged_at": "2025-10-06T13:14:03" }
# What does this PR do? Make better typehints for `apply_chat_template`. Now IDEs and type checkers will correctly infer types for this code. ```py from transformers import AutoTokenizer tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-0.6B") # Returns raw string when tokenize=False raw_prompt: str = tokenizer.apply_chat_template( [{"role": "user", "content": "Hello!"}], tokenize=False, ) # Returns token ids when tokenize=True (default) and return_dict=False token_ids: list[int] = tokenizer.apply_chat_template( [{"role": "user", "content": "Hello!"}], ) # Returns BatchEncoding when return_dict=True batch_encoding: BatchEncoding = tokenizer.apply_chat_template( [{"role": "user", "content": "Hello!"}], return_dict=True, ) ``` Similarly, can be done for [`ProcessorMixin.apply_chat_template`](https://github.com/huggingface/transformers/blob/e11a00a16f925b7d3b52f5007bdce3464edb361f/src/transformers/processing_utils.py#L1559), but I'm not sure how to add to it, because it's using kwargs instead of arguments ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [X] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? @ArthurZucker and @itazap
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41355/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41355/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41354
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41354/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41354/comments
https://api.github.com/repos/huggingface/transformers/issues/41354/events
https://github.com/huggingface/transformers/pull/41354
3,484,931,586
PR_kwDOCUB6oc6sJkB6
41,354
Fixed tiny incorrect import in `gemma3`
{ "login": "Sai-Suraj-27", "id": 87087741, "node_id": "MDQ6VXNlcjg3MDg3NzQx", "avatar_url": "https://avatars.githubusercontent.com/u/87087741?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Sai-Suraj-27", "html_url": "https://github.com/Sai-Suraj-27", "followers_url": "https://api.github.com/users/Sai-Suraj-27/followers", "following_url": "https://api.github.com/users/Sai-Suraj-27/following{/other_user}", "gists_url": "https://api.github.com/users/Sai-Suraj-27/gists{/gist_id}", "starred_url": "https://api.github.com/users/Sai-Suraj-27/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Sai-Suraj-27/subscriptions", "organizations_url": "https://api.github.com/users/Sai-Suraj-27/orgs", "repos_url": "https://api.github.com/users/Sai-Suraj-27/repos", "events_url": "https://api.github.com/users/Sai-Suraj-27/events{/privacy}", "received_events_url": "https://api.github.com/users/Sai-Suraj-27/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T16:06:15
2025-10-09T15:32:00
2025-10-06T08:55:43
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41354", "html_url": "https://github.com/huggingface/transformers/pull/41354", "diff_url": "https://github.com/huggingface/transformers/pull/41354.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41354.patch", "merged_at": "2025-10-06T08:55:43" }
# What does this PR do? Fixes tiny incorrect import in `gemma3`. This is the correct class name https://github.com/huggingface/transformers/blob/e11a00a16f925b7d3b52f5007bdce3464edb361f/src/transformers/models/paligemma/modeling_paligemma.py#L67 Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? @gante @zucchini-nlp
{ "login": "zucchini-nlp", "id": 100715397, "node_id": "U_kgDOBgDLhQ", "avatar_url": "https://avatars.githubusercontent.com/u/100715397?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zucchini-nlp", "html_url": "https://github.com/zucchini-nlp", "followers_url": "https://api.github.com/users/zucchini-nlp/followers", "following_url": "https://api.github.com/users/zucchini-nlp/following{/other_user}", "gists_url": "https://api.github.com/users/zucchini-nlp/gists{/gist_id}", "starred_url": "https://api.github.com/users/zucchini-nlp/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/zucchini-nlp/subscriptions", "organizations_url": "https://api.github.com/users/zucchini-nlp/orgs", "repos_url": "https://api.github.com/users/zucchini-nlp/repos", "events_url": "https://api.github.com/users/zucchini-nlp/events{/privacy}", "received_events_url": "https://api.github.com/users/zucchini-nlp/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41354/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41354/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41353
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41353/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41353/comments
https://api.github.com/repos/huggingface/transformers/issues/41353/events
https://github.com/huggingface/transformers/pull/41353
3,484,926,425
PR_kwDOCUB6oc6sJjBK
41,353
serve: add non-streaming mode to /v1/responses; stream event parity; remove placeholder logprobs
{ "login": "antznette1", "id": 106275232, "node_id": "U_kgDOBlWhoA", "avatar_url": "https://avatars.githubusercontent.com/u/106275232?v=4", "gravatar_id": "", "url": "https://api.github.com/users/antznette1", "html_url": "https://github.com/antznette1", "followers_url": "https://api.github.com/users/antznette1/followers", "following_url": "https://api.github.com/users/antznette1/following{/other_user}", "gists_url": "https://api.github.com/users/antznette1/gists{/gist_id}", "starred_url": "https://api.github.com/users/antznette1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/antznette1/subscriptions", "organizations_url": "https://api.github.com/users/antznette1/orgs", "repos_url": "https://api.github.com/users/antznette1/repos", "events_url": "https://api.github.com/users/antznette1/events{/privacy}", "received_events_url": "https://api.github.com/users/antznette1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T16:00:25
2025-10-07T17:59:41
2025-10-06T16:04:17
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41353", "html_url": "https://github.com/huggingface/transformers/pull/41353", "diff_url": "https://github.com/huggingface/transformers/pull/41353.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41353.patch", "merged_at": "2025-10-06T16:04:17" }
- Add `stream=false` support in `POST /v1/responses` returning a single OpenAI `Response`. - Emit streaming deltas consistently; avoid placeholder logprobs by using empty lists. - Stabilize tests: robust event checks; deterministic full request. **Files:** - src/transformers/commands/serving.py - tests/commands/test_serving.py _**Notes**_: Minimal diff, aligns with AGENTS.md. Improves OpenAI parity.
{ "login": "gante", "id": 12240844, "node_id": "MDQ6VXNlcjEyMjQwODQ0", "avatar_url": "https://avatars.githubusercontent.com/u/12240844?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gante", "html_url": "https://github.com/gante", "followers_url": "https://api.github.com/users/gante/followers", "following_url": "https://api.github.com/users/gante/following{/other_user}", "gists_url": "https://api.github.com/users/gante/gists{/gist_id}", "starred_url": "https://api.github.com/users/gante/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/gante/subscriptions", "organizations_url": "https://api.github.com/users/gante/orgs", "repos_url": "https://api.github.com/users/gante/repos", "events_url": "https://api.github.com/users/gante/events{/privacy}", "received_events_url": "https://api.github.com/users/gante/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41353/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41353/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41352
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41352/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41352/comments
https://api.github.com/repos/huggingface/transformers/issues/41352/events
https://github.com/huggingface/transformers/pull/41352
3,484,920,884
PR_kwDOCUB6oc6sJh7W
41,352
Fix typo in model proposal template
{ "login": "Ombucha", "id": 79257339, "node_id": "MDQ6VXNlcjc5MjU3MzM5", "avatar_url": "https://avatars.githubusercontent.com/u/79257339?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Ombucha", "html_url": "https://github.com/Ombucha", "followers_url": "https://api.github.com/users/Ombucha/followers", "following_url": "https://api.github.com/users/Ombucha/following{/other_user}", "gists_url": "https://api.github.com/users/Ombucha/gists{/gist_id}", "starred_url": "https://api.github.com/users/Ombucha/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Ombucha/subscriptions", "organizations_url": "https://api.github.com/users/Ombucha/orgs", "repos_url": "https://api.github.com/users/Ombucha/repos", "events_url": "https://api.github.com/users/Ombucha/events{/privacy}", "received_events_url": "https://api.github.com/users/Ombucha/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T15:52:37
2025-10-06T13:07:34
2025-10-06T13:06:51
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41352", "html_url": "https://github.com/huggingface/transformers/pull/41352", "diff_url": "https://github.com/huggingface/transformers/pull/41352.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41352.patch", "merged_at": "2025-10-06T13:06:51" }
# What does this PR do? This PR fixes a grammatical error. ## Before submitting - [x] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [x] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [x] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. @stevhiku
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41352/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41352/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41351
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41351/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41351/comments
https://api.github.com/repos/huggingface/transformers/issues/41351/events
https://github.com/huggingface/transformers/pull/41351
3,484,906,038
PR_kwDOCUB6oc6sJfDx
41,351
serve: add non-streaming mode to /v1/responses; stream event parity; remove placeholder logprobs
{ "login": "antznette1", "id": 106275232, "node_id": "U_kgDOBlWhoA", "avatar_url": "https://avatars.githubusercontent.com/u/106275232?v=4", "gravatar_id": "", "url": "https://api.github.com/users/antznette1", "html_url": "https://github.com/antznette1", "followers_url": "https://api.github.com/users/antznette1/followers", "following_url": "https://api.github.com/users/antznette1/following{/other_user}", "gists_url": "https://api.github.com/users/antznette1/gists{/gist_id}", "starred_url": "https://api.github.com/users/antznette1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/antznette1/subscriptions", "organizations_url": "https://api.github.com/users/antznette1/orgs", "repos_url": "https://api.github.com/users/antznette1/repos", "events_url": "https://api.github.com/users/antznette1/events{/privacy}", "received_events_url": "https://api.github.com/users/antznette1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T15:32:59
2025-10-05T15:34:12
2025-10-05T15:33:10
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41351", "html_url": "https://github.com/huggingface/transformers/pull/41351", "diff_url": "https://github.com/huggingface/transformers/pull/41351.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41351.patch", "merged_at": null }
- Add `stream=false` support in `POST /v1/responses` returning a single OpenAI `Response`. - Emit streaming deltas consistently; avoid placeholder logprobs by using empty lists. - Stabilize tests: robust event checks; deterministic full request. **Files:** - src/transformers/commands/serving.py - tests/commands/test_serving.py _**Notes**_: Minimal diff, aligns with AGENTS.md. Improves OpenAI parity.
{ "login": "antznette1", "id": 106275232, "node_id": "U_kgDOBlWhoA", "avatar_url": "https://avatars.githubusercontent.com/u/106275232?v=4", "gravatar_id": "", "url": "https://api.github.com/users/antznette1", "html_url": "https://github.com/antznette1", "followers_url": "https://api.github.com/users/antznette1/followers", "following_url": "https://api.github.com/users/antznette1/following{/other_user}", "gists_url": "https://api.github.com/users/antznette1/gists{/gist_id}", "starred_url": "https://api.github.com/users/antznette1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/antznette1/subscriptions", "organizations_url": "https://api.github.com/users/antznette1/orgs", "repos_url": "https://api.github.com/users/antznette1/repos", "events_url": "https://api.github.com/users/antznette1/events{/privacy}", "received_events_url": "https://api.github.com/users/antznette1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41351/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41351/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41350
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41350/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41350/comments
https://api.github.com/repos/huggingface/transformers/issues/41350/events
https://github.com/huggingface/transformers/pull/41350
3,484,775,789
PR_kwDOCUB6oc6sJFlW
41,350
Resolve issue 41296
{ "login": "DivyanshRajSoni", "id": 110761086, "node_id": "U_kgDOBpoUfg", "avatar_url": "https://avatars.githubusercontent.com/u/110761086?v=4", "gravatar_id": "", "url": "https://api.github.com/users/DivyanshRajSoni", "html_url": "https://github.com/DivyanshRajSoni", "followers_url": "https://api.github.com/users/DivyanshRajSoni/followers", "following_url": "https://api.github.com/users/DivyanshRajSoni/following{/other_user}", "gists_url": "https://api.github.com/users/DivyanshRajSoni/gists{/gist_id}", "starred_url": "https://api.github.com/users/DivyanshRajSoni/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/DivyanshRajSoni/subscriptions", "organizations_url": "https://api.github.com/users/DivyanshRajSoni/orgs", "repos_url": "https://api.github.com/users/DivyanshRajSoni/repos", "events_url": "https://api.github.com/users/DivyanshRajSoni/events{/privacy}", "received_events_url": "https://api.github.com/users/DivyanshRajSoni/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T12:45:36
2025-10-06T12:58:51
2025-10-06T12:46:01
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41350", "html_url": "https://github.com/huggingface/transformers/pull/41350", "diff_url": "https://github.com/huggingface/transformers/pull/41350.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41350.patch", "merged_at": null }
# Issues solved 41296. What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [x] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [x] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41350/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41350/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41349
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41349/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41349/comments
https://api.github.com/repos/huggingface/transformers/issues/41349/events
https://github.com/huggingface/transformers/pull/41349
3,484,740,970
PR_kwDOCUB6oc6sI-5P
41,349
Create (3d_parrallel_v2.py) - Add 3D parallelism training example script
{ "login": "Haziq2006", "id": 142998784, "node_id": "U_kgDOCIX9AA", "avatar_url": "https://avatars.githubusercontent.com/u/142998784?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Haziq2006", "html_url": "https://github.com/Haziq2006", "followers_url": "https://api.github.com/users/Haziq2006/followers", "following_url": "https://api.github.com/users/Haziq2006/following{/other_user}", "gists_url": "https://api.github.com/users/Haziq2006/gists{/gist_id}", "starred_url": "https://api.github.com/users/Haziq2006/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Haziq2006/subscriptions", "organizations_url": "https://api.github.com/users/Haziq2006/orgs", "repos_url": "https://api.github.com/users/Haziq2006/repos", "events_url": "https://api.github.com/users/Haziq2006/events{/privacy}", "received_events_url": "https://api.github.com/users/Haziq2006/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
open
false
null
[]
null
[]
2025-10-05T11:59:22
2025-10-05T11:59:22
null
NONE
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41349", "html_url": "https://github.com/huggingface/transformers/pull/41349", "diff_url": "https://github.com/huggingface/transformers/pull/41349.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41349.patch", "merged_at": null }
This PR adds a new training example under examples/3D_parallel.py that demonstrates a simple training loop with distributed (DDP/TP/PP) support. Changes included: -Fixed invalid docstring (""" instead of """:). -Added fallback for local_rank in non-distributed setups. -Wrapped dist.get_rank() and dist.destroy_process_group() with if dist.is_initialized(). -Lowered num_proc in datasets.map to avoid crashes on small machines. -Integrated optional Weights & Biases tracking (--with_tracking). -This script is intended to help users quickly test 3D parallelism setups with Hugging Face models. # What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ If you know how to use git blame, that is the easiest way, otherwise, here is a rough guide of **who to tag**. Please tag fewer than 3 people. Models: - text models: @ArthurZucker @Cyrilvallez - vision models: @yonigozlan @molbap - audio models: @eustlb @ebezzam @vasqu - multimodal models: @zucchini-nlp - graph models: @clefourrier Library: - generate: @zucchini-nlp (visual-language models) or @gante (all others) - continuous batching: @remi-or @ArthurZucker @McPatate - pipelines: @Rocketknight1 - tokenizers: @ArthurZucker and @itazap - trainer: @zach-huggingface @SunMarc - attention: @vasqu @ArthurZucker @CyrilVallez - model loading (from pretrained, etc): @CyrilVallez - distributed: @3outeille @ArthurZucker @S1ro1 - CIs: @ydshieh Integrations: - deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface - ray/raytune: @richardliaw, @amogkam - Big Model Inference: @SunMarc - quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber - kernels: @MekkCyber @drbh - peft: @BenjaminBossan @githubnemo Devices/Backends: - AMD ROCm: @ivarflakstad - Intel XPU: @IlyasMoutawwakil - Ascend NPU: @ivarflakstad Documentation: @stevhliu Research projects are not maintained and should be taken as is. -->
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41349/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41349/timeline
null
null
null
null
true
false
https://api.github.com/repos/huggingface/transformers/issues/41348
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41348/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41348/comments
https://api.github.com/repos/huggingface/transformers/issues/41348/events
https://github.com/huggingface/transformers/pull/41348
3,484,655,204
PR_kwDOCUB6oc6sIuf4
41,348
fix python39 trainer compatibility
{ "login": "Samoed", "id": 36135455, "node_id": "MDQ6VXNlcjM2MTM1NDU1", "avatar_url": "https://avatars.githubusercontent.com/u/36135455?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Samoed", "html_url": "https://github.com/Samoed", "followers_url": "https://api.github.com/users/Samoed/followers", "following_url": "https://api.github.com/users/Samoed/following{/other_user}", "gists_url": "https://api.github.com/users/Samoed/gists{/gist_id}", "starred_url": "https://api.github.com/users/Samoed/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Samoed/subscriptions", "organizations_url": "https://api.github.com/users/Samoed/orgs", "repos_url": "https://api.github.com/users/Samoed/repos", "events_url": "https://api.github.com/users/Samoed/events{/privacy}", "received_events_url": "https://api.github.com/users/Samoed/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-05T10:21:48
2025-10-06T09:32:39
2025-10-06T09:32:39
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41348", "html_url": "https://github.com/huggingface/transformers/pull/41348", "diff_url": "https://github.com/huggingface/transformers/pull/41348.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41348.patch", "merged_at": null }
# What does this PR do? Fix typehint in trainer for python3.9 compatibility Fixes https://github.com/huggingface/transformers/issues/41339 ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? @zach-huggingface @SunMarc
{ "login": "Samoed", "id": 36135455, "node_id": "MDQ6VXNlcjM2MTM1NDU1", "avatar_url": "https://avatars.githubusercontent.com/u/36135455?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Samoed", "html_url": "https://github.com/Samoed", "followers_url": "https://api.github.com/users/Samoed/followers", "following_url": "https://api.github.com/users/Samoed/following{/other_user}", "gists_url": "https://api.github.com/users/Samoed/gists{/gist_id}", "starred_url": "https://api.github.com/users/Samoed/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Samoed/subscriptions", "organizations_url": "https://api.github.com/users/Samoed/orgs", "repos_url": "https://api.github.com/users/Samoed/repos", "events_url": "https://api.github.com/users/Samoed/events{/privacy}", "received_events_url": "https://api.github.com/users/Samoed/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41348/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41348/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41347
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41347/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41347/comments
https://api.github.com/repos/huggingface/transformers/issues/41347/events
https://github.com/huggingface/transformers/pull/41347
3,484,228,012
PR_kwDOCUB6oc6sHapv
41,347
Fix flash_attention.py: wrong argument passing for attn_implementation
{ "login": "TKONIY", "id": 47519522, "node_id": "MDQ6VXNlcjQ3NTE5NTIy", "avatar_url": "https://avatars.githubusercontent.com/u/47519522?v=4", "gravatar_id": "", "url": "https://api.github.com/users/TKONIY", "html_url": "https://github.com/TKONIY", "followers_url": "https://api.github.com/users/TKONIY/followers", "following_url": "https://api.github.com/users/TKONIY/following{/other_user}", "gists_url": "https://api.github.com/users/TKONIY/gists{/gist_id}", "starred_url": "https://api.github.com/users/TKONIY/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/TKONIY/subscriptions", "organizations_url": "https://api.github.com/users/TKONIY/orgs", "repos_url": "https://api.github.com/users/TKONIY/repos", "events_url": "https://api.github.com/users/TKONIY/events{/privacy}", "received_events_url": "https://api.github.com/users/TKONIY/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-04T23:51:05
2025-10-07T17:57:16
2025-10-06T13:36:40
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41347", "html_url": "https://github.com/huggingface/transformers/pull/41347", "diff_url": "https://github.com/huggingface/transformers/pull/41347.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41347.patch", "merged_at": "2025-10-06T13:36:40" }
# What does this PR do? The name of the attn type argument for `_flash_attention_forward()` should be `implementation`, instead of `attn_implementation` which currently uses in the function call. This would result in wrong type specification. <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) ## Before submitting - [x] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## CC @vasqu @ArthurZucker @CyrilVallez
{ "login": "Cyrilvallez", "id": 71554963, "node_id": "MDQ6VXNlcjcxNTU0OTYz", "avatar_url": "https://avatars.githubusercontent.com/u/71554963?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Cyrilvallez", "html_url": "https://github.com/Cyrilvallez", "followers_url": "https://api.github.com/users/Cyrilvallez/followers", "following_url": "https://api.github.com/users/Cyrilvallez/following{/other_user}", "gists_url": "https://api.github.com/users/Cyrilvallez/gists{/gist_id}", "starred_url": "https://api.github.com/users/Cyrilvallez/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Cyrilvallez/subscriptions", "organizations_url": "https://api.github.com/users/Cyrilvallez/orgs", "repos_url": "https://api.github.com/users/Cyrilvallez/repos", "events_url": "https://api.github.com/users/Cyrilvallez/events{/privacy}", "received_events_url": "https://api.github.com/users/Cyrilvallez/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41347/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41347/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41346
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41346/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41346/comments
https://api.github.com/repos/huggingface/transformers/issues/41346/events
https://github.com/huggingface/transformers/pull/41346
3,483,981,116
PR_kwDOCUB6oc6sGrm4
41,346
Prefer raising `TypeError` exception for invalid type
{ "login": "Sai-Suraj-27", "id": 87087741, "node_id": "MDQ6VXNlcjg3MDg3NzQx", "avatar_url": "https://avatars.githubusercontent.com/u/87087741?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Sai-Suraj-27", "html_url": "https://github.com/Sai-Suraj-27", "followers_url": "https://api.github.com/users/Sai-Suraj-27/followers", "following_url": "https://api.github.com/users/Sai-Suraj-27/following{/other_user}", "gists_url": "https://api.github.com/users/Sai-Suraj-27/gists{/gist_id}", "starred_url": "https://api.github.com/users/Sai-Suraj-27/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Sai-Suraj-27/subscriptions", "organizations_url": "https://api.github.com/users/Sai-Suraj-27/orgs", "repos_url": "https://api.github.com/users/Sai-Suraj-27/repos", "events_url": "https://api.github.com/users/Sai-Suraj-27/events{/privacy}", "received_events_url": "https://api.github.com/users/Sai-Suraj-27/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-04T17:24:42
2025-10-07T13:12:09
2025-10-07T13:11:43
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41346", "html_url": "https://github.com/huggingface/transformers/pull/41346", "diff_url": "https://github.com/huggingface/transformers/pull/41346.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41346.patch", "merged_at": "2025-10-07T13:11:43" }
# What does this PR do? Fixes raising of `TypeError` exception instead of `ValueError` for invalid type ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? @Rocketknight1 @ArthurZucker
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41346/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41346/timeline
null
null
null
null
true
true
https://api.github.com/repos/huggingface/transformers/issues/41345
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41345/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41345/comments
https://api.github.com/repos/huggingface/transformers/issues/41345/events
https://github.com/huggingface/transformers/issues/41345
3,483,828,480
I_kwDOCUB6oc7PpwEA
41,345
Add standardized multimodal capability metadata
{ "login": "adarshxs", "id": 114558126, "node_id": "U_kgDOBtQErg", "avatar_url": "https://avatars.githubusercontent.com/u/114558126?v=4", "gravatar_id": "", "url": "https://api.github.com/users/adarshxs", "html_url": "https://github.com/adarshxs", "followers_url": "https://api.github.com/users/adarshxs/followers", "following_url": "https://api.github.com/users/adarshxs/following{/other_user}", "gists_url": "https://api.github.com/users/adarshxs/gists{/gist_id}", "starred_url": "https://api.github.com/users/adarshxs/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/adarshxs/subscriptions", "organizations_url": "https://api.github.com/users/adarshxs/orgs", "repos_url": "https://api.github.com/users/adarshxs/repos", "events_url": "https://api.github.com/users/adarshxs/events{/privacy}", "received_events_url": "https://api.github.com/users/adarshxs/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 2648621985, "node_id": "MDU6TGFiZWwyNjQ4NjIxOTg1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Feature%20request", "name": "Feature request", "color": "FBCA04", "default": false, "description": "Request for a new feature" } ]
closed
false
null
[]
null
[]
2025-10-04T14:57:52
2025-10-06T11:03:42
2025-10-06T11:03:10
NONE
null
null
null
null
### Feature request Downstream systems like SGLang/vLLM rely on registry/plugin systems or model-specific integration and metadata to decide whether a model is multimodal (i.e. `has_processor` in MultiModalRegistry). There is no consistent, library-level field in PretrainedConfig that captures modality support. Having a standardized config field would help. rely on a standardized flag rather than custom heuristics or registry hacks. anyway we can pass tags like the ones on the huggingface model tags to model configs without breaking/maintaining backstream support? ``` multimodal_tags = { "multimodal", "image-to-text", "image-text-to-text", "video-to-text", "video-text-to-text", "audio-to-text", "audio-text-to-text", "any-to-any", } ``` ### Motivation ``` supported_modalities: Optional[List[str]] = None is_multimodal: bool = False modality_token_limits: Optional[Dict[str, int]] = None pipeline_tag ... ``` fields like these would be helpful ### Your contribution would like to discuss if this idea makes sense and is possible to implement.
{ "login": "adarshxs", "id": 114558126, "node_id": "U_kgDOBtQErg", "avatar_url": "https://avatars.githubusercontent.com/u/114558126?v=4", "gravatar_id": "", "url": "https://api.github.com/users/adarshxs", "html_url": "https://github.com/adarshxs", "followers_url": "https://api.github.com/users/adarshxs/followers", "following_url": "https://api.github.com/users/adarshxs/following{/other_user}", "gists_url": "https://api.github.com/users/adarshxs/gists{/gist_id}", "starred_url": "https://api.github.com/users/adarshxs/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/adarshxs/subscriptions", "organizations_url": "https://api.github.com/users/adarshxs/orgs", "repos_url": "https://api.github.com/users/adarshxs/repos", "events_url": "https://api.github.com/users/adarshxs/events{/privacy}", "received_events_url": "https://api.github.com/users/adarshxs/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41345/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 1 }
https://api.github.com/repos/huggingface/transformers/issues/41345/timeline
null
completed
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
true
https://api.github.com/repos/huggingface/transformers/issues/41344
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41344/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41344/comments
https://api.github.com/repos/huggingface/transformers/issues/41344/events
https://github.com/huggingface/transformers/issues/41344
3,483,799,879
I_kwDOCUB6oc7PppFH
41,344
No need to gather all parameters to save model when training with DeepSpeed Zero3 and LoRA
{ "login": "Taise228", "id": 95762401, "node_id": "U_kgDOBbU34Q", "avatar_url": "https://avatars.githubusercontent.com/u/95762401?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Taise228", "html_url": "https://github.com/Taise228", "followers_url": "https://api.github.com/users/Taise228/followers", "following_url": "https://api.github.com/users/Taise228/following{/other_user}", "gists_url": "https://api.github.com/users/Taise228/gists{/gist_id}", "starred_url": "https://api.github.com/users/Taise228/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Taise228/subscriptions", "organizations_url": "https://api.github.com/users/Taise228/orgs", "repos_url": "https://api.github.com/users/Taise228/repos", "events_url": "https://api.github.com/users/Taise228/events{/privacy}", "received_events_url": "https://api.github.com/users/Taise228/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[ { "id": 2648621985, "node_id": "MDU6TGFiZWwyNjQ4NjIxOTg1", "url": "https://api.github.com/repos/huggingface/transformers/labels/Feature%20request", "name": "Feature request", "color": "FBCA04", "default": false, "description": "Request for a new feature" } ]
open
false
null
[]
null
[]
2025-10-04T14:31:55
2025-10-06T13:56:32
null
NONE
null
null
null
null
### Feature request Gather only adapter weights to save model weights when training with LoRA and DeepSpeed Zero3. ### Motivation When we use `Trainer` and train models with LoRA, trainer only saves adapter weights. However, when we use DeepSpeed Zero3, which distributes model weights on multi GPUs, and set `stage3_gather_16bit_weights_on_model_save: true`, current implementation tries to gather the whole model weights to rank0 in `self.accelerator.get_state_dict(self.deepspeed)`. It is not optimal and also it can lead to OOM when training large models even if trained with LoRA. ### Your contribution I think I can fix it by adding a few lines to `Trainer.save_model` as below: before ``` elif self.is_deepspeed_enabled: try: state_dict = self.accelerator.get_state_dict(self.deepspeed) ``` after ``` elif self.is_deepspeed_enabled: try: accept_exclude_frozen_parameters = "exclude_frozen_parameters" in set( inspect.signature(self.model_wrapped.save_checkpoint).parameters.keys() ) zero3_sharding = self.deepspeed.config.get("zero_optimization", {}).get("stage", None) == 3 if accept_exclude_frozen_parameters and _is_peft_model(self.model) and zero3_sharding: # When using PEFT with DeepSpeed ZeRO Stage 3, # we do not need to load the frozen parameters state_dict = self.deepspeed._zero3_consolidated_16bit_state_dict(exclude_frozen_parameters=True) else: state_dict = self.accelerator.get_state_dict(self.deepspeed) ``` If it is okay, I can take responsibility to file PR.
null
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41344/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41344/timeline
null
null
{ "total": 0, "completed": 0, "percent_completed": 0 }
{ "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }
false
false
https://api.github.com/repos/huggingface/transformers/issues/41343
https://api.github.com/repos/huggingface/transformers
https://api.github.com/repos/huggingface/transformers/issues/41343/labels{/name}
https://api.github.com/repos/huggingface/transformers/issues/41343/comments
https://api.github.com/repos/huggingface/transformers/issues/41343/events
https://github.com/huggingface/transformers/pull/41343
3,483,737,753
PR_kwDOCUB6oc6sF7ye
41,343
Fix test for model with dotted name and relative imports
{ "login": "st81", "id": 58893365, "node_id": "MDQ6VXNlcjU4ODkzMzY1", "avatar_url": "https://avatars.githubusercontent.com/u/58893365?v=4", "gravatar_id": "", "url": "https://api.github.com/users/st81", "html_url": "https://github.com/st81", "followers_url": "https://api.github.com/users/st81/followers", "following_url": "https://api.github.com/users/st81/following{/other_user}", "gists_url": "https://api.github.com/users/st81/gists{/gist_id}", "starred_url": "https://api.github.com/users/st81/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/st81/subscriptions", "organizations_url": "https://api.github.com/users/st81/orgs", "repos_url": "https://api.github.com/users/st81/repos", "events_url": "https://api.github.com/users/st81/events{/privacy}", "received_events_url": "https://api.github.com/users/st81/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
[]
closed
false
null
[]
null
[]
2025-10-04T13:36:49
2025-10-07T12:55:54
2025-10-07T12:55:54
CONTRIBUTOR
null
null
false
{ "url": "https://api.github.com/repos/huggingface/transformers/pulls/41343", "html_url": "https://github.com/huggingface/transformers/pull/41343", "diff_url": "https://github.com/huggingface/transformers/pull/41343.diff", "patch_url": "https://github.com/huggingface/transformers/pull/41343.patch", "merged_at": "2025-10-07T12:55:54" }
# What does this PR do? <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> The test `test_model_with_dotted_name_and_relative_imports` added in #40745 is not working as expected because `model_id = "hf-internal-testing/remote_code_model_with_dots"` does not include `.` in the name. If this PR change looks good, could you create this new remote model? (I didn't create it myself because relying on an external user's resource is risky, as you mentioned in https://github.com/huggingface/transformers/pull/40745#pullrequestreview-3206353873). Below is a detailed explanation of why `test_model_with_dotted_name_and_relative_imports` is not working. The explanation is a bit long, so feel free to skip it if you understand the problem from the above explanation. --- If we remove the `_sanitize_module_name` process simply by changing the method like below: ```py def _sanitize_module_name(name: str) -> str: r""" Tries to sanitize a module name so that it can be used as a Python module. The following transformations are applied: 1. Replace `.` in module names with `_dot_`. 2. Replace `-` in module names with `_hyphen_`. 3. If the module name starts with a digit, prepend it with `_`. 4. Warn if the sanitized name is a Python reserved keyword or not a valid identifier. If the input name is already a valid identifier, it is returned unchanged. """ # Not replacing `\W` characters with `_` to avoid collisions, because `_` is very common return name ``` The test should raise an error, but it actually passes: ```sh pytest tests/models/auto/test_modeling_auto.py::AutoModelTest::test_model_with_dotted_name_and_relative_imports ================================================================================== test session starts =================================================================================== platform linux -- Python 3.12.11, pytest-8.4.1, pluggy-1.6.0 rootdir: /home/shutotakahashi/projects/transformers-uv/transformers configfile: pyproject.toml plugins: anyio-4.11.0, xdist-3.8.0 collected 1 item tests/models/auto/test_modeling_auto.py::AutoModelTest::test_model_with_dotted_name_and_relative_imports PASSED [100%] ``` The reason the test passes is that the current `model_id = "hf-internal-testing/remote_code_model_with_dots"` does not include `.` in the name, as reported in the original issue https://github.com/huggingface/transformers/issues/40496. If we replace `model_id` with[ `August4293/test-model_v1.0`](https://huggingface.co/August4293/test-model_v1.0/tree/main), which includes `.` in the name (this model was uploaded by the original PR contributor in https://github.com/huggingface/transformers/pull/40745#issuecomment-3270735455), the test raises the expected error: ```py def test_model_with_dotted_name_and_relative_imports(self): """ Test for issue #40496: AutoModel.from_pretrained() doesn't work for models with '.' in their name when there's a relative import. Without the fix, this raises: ModuleNotFoundError: No module named 'transformers_modules.August4293.test-model_v1' """ model_id = "August4293/test-model_v1.0" model = AutoModel.from_pretrained(model_id, trust_remote_code=True) self.assertIsNotNone(model) ``` ```sh pytest tests/models/auto/test_modeling_auto.py::AutoModelTest::test_model_with_dotted_name_and_relative_imports ================================================================================== test session starts =================================================================================== platform linux -- Python 3.12.11, pytest-8.4.1, pluggy-1.6.0 rootdir: /home/shutotakahashi/projects/transformers-uv/transformers configfile: pyproject.toml plugins: anyio-4.11.0, xdist-3.8.0 collected 1 item tests/models/auto/test_modeling_auto.py::AutoModelTest::test_model_with_dotted_name_and_relative_imports FAILED [100%] ======================================================================================== FAILURES ======================================================================================== _____________________________________________________________ AutoModelTest.test_model_with_dotted_name_and_relative_imports _____________________________________________________________ self = <tests.models.auto.test_modeling_auto.AutoModelTest testMethod=test_model_with_dotted_name_and_relative_imports> def test_model_with_dotted_name_and_relative_imports(self): """ Test for issue #40496: AutoModel.from_pretrained() doesn't work for models with '.' in their name when there's a relative import. Without the fix, this raises: ModuleNotFoundError: No module named 'transformers_modules.August4293.test-model_v1' """ model_id = "August4293/test-model_v1.0" > model = AutoModel.from_pretrained(model_id, trust_remote_code=True) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ tests/models/auto/test_modeling_auto.py:587: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ src/transformers/models/auto/auto_factory.py:367: in from_pretrained model_class = get_class_from_dynamic_module( src/transformers/dynamic_module_utils.py:606: in get_class_from_dynamic_module return get_class_in_module(class_name, final_module, force_reload=force_download) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ src/transformers/dynamic_module_utils.py:312: in get_class_in_module module_spec.loader.exec_module(module) <frozen importlib._bootstrap_external>:999: in exec_module ??? <frozen importlib._bootstrap>:488: in _call_with_frames_removed ??? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ """ Custom model with relative import to demonstrate the bug. """ from transformers import PreTrainedModel from transformers.modeling_outputs import BaseModelOutput import torch import torch.nn as nn # This relative import should cause the bug when the folder has a dot in the name > from .another_module import custom_function E ModuleNotFoundError: No module named 'transformers_modules.August4293.test-model_v1' ``` Fixes #40496 ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [x] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section? - [x] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? ## Who can review? @Rocketknight1
{ "login": "Rocketknight1", "id": 12866554, "node_id": "MDQ6VXNlcjEyODY2NTU0", "avatar_url": "https://avatars.githubusercontent.com/u/12866554?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Rocketknight1", "html_url": "https://github.com/Rocketknight1", "followers_url": "https://api.github.com/users/Rocketknight1/followers", "following_url": "https://api.github.com/users/Rocketknight1/following{/other_user}", "gists_url": "https://api.github.com/users/Rocketknight1/gists{/gist_id}", "starred_url": "https://api.github.com/users/Rocketknight1/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Rocketknight1/subscriptions", "organizations_url": "https://api.github.com/users/Rocketknight1/orgs", "repos_url": "https://api.github.com/users/Rocketknight1/repos", "events_url": "https://api.github.com/users/Rocketknight1/events{/privacy}", "received_events_url": "https://api.github.com/users/Rocketknight1/received_events", "type": "User", "user_view_type": "public", "site_admin": false }
{ "url": "https://api.github.com/repos/huggingface/transformers/issues/41343/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }
https://api.github.com/repos/huggingface/transformers/issues/41343/timeline
null
null
null
null
true
true