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/41040 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41040/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41040/comments | https://api.github.com/repos/huggingface/transformers/issues/41040/events | https://github.com/huggingface/transformers/pull/41040 | 3,439,078,084 | PR_kwDOCUB6oc6pvjSn | 41,040 | Add Keye vl 8b 1.5 | {
"login": "Kwai-Keye",
"id": 216671074,
"node_id": "U_kgDODOojYg",
"avatar_url": "https://avatars.githubusercontent.com/u/216671074?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Kwai-Keye",
"html_url": "https://github.com/Kwai-Keye",
"followers_url": "https://api.github.com/users/Kwai-Keye/followers",
"following_url": "https://api.github.com/users/Kwai-Keye/following{/other_user}",
"gists_url": "https://api.github.com/users/Kwai-Keye/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Kwai-Keye/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Kwai-Keye/subscriptions",
"organizations_url": "https://api.github.com/users/Kwai-Keye/orgs",
"repos_url": "https://api.github.com/users/Kwai-Keye/repos",
"events_url": "https://api.github.com/users/Kwai-Keye/events{/privacy}",
"received_events_url": "https://api.github.com/users/Kwai-Keye/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-22T03:51:07 | 2025-10-21T09:19:46 | null | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41040",
"html_url": "https://github.com/huggingface/transformers/pull/41040",
"diff_url": "https://github.com/huggingface/transformers/pull/41040.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41040.patch",
"merged_at": null
} | # Model Upgrade: Keye-VL-1.5-8B
## Overview
This PR introduces an upgraded version of the visual-language model, transitioning from the previous `keye-preview` to `keye-vl-1.5-8B`. The update includes architectural refinements, documentation improvements, code optimizations, and style enhancements.
## Key Changes
### 1. Model Architecture
- Fine-tuned the model structure for improved performance and efficiency.
- Enhanced visual-language alignment capabilities.
### 2. Documentation
- Updated and optimized relevant documentation for better clarity and usability.
- Added detailed descriptions of new features and modifications.
### 3. Code Optimization
- Streamlined code workflow for enhanced maintainability.
- Removed redundant parameters and code lines to improve efficiency.
- Performed code style adaptations to ensure consistency with project standards.
## Impact
- Improved model performance and accuracy.
- Enhanced code readability and maintainability.
- Reduced computational overhead by eliminating unnecessary parameters.
## Usage
Refer to the updated documentation for detailed instructions on using the new model version.
## Notes
- This upgrade maintains backward compatibility with existing pipelines.
- Users are encouraged to review the updated docs for optimal utilization of new features. | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41040/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/41040/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41039 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41039/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41039/comments | https://api.github.com/repos/huggingface/transformers/issues/41039/events | https://github.com/huggingface/transformers/pull/41039 | 3,438,715,064 | PR_kwDOCUB6oc6puZME | 41,039 | docs: fully offline usage (env vars, local_files_only, caches) | {
"login": "Username46786",
"id": 98800422,
"node_id": "U_kgDOBeOTJg",
"avatar_url": "https://avatars.githubusercontent.com/u/98800422?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Username46786",
"html_url": "https://github.com/Username46786",
"followers_url": "https://api.github.com/users/Username46786/followers",
"following_url": "https://api.github.com/users/Username46786/following{/other_user}",
"gists_url": "https://api.github.com/users/Username46786/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Username46786/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Username46786/subscriptions",
"organizations_url": "https://api.github.com/users/Username46786/orgs",
"repos_url": "https://api.github.com/users/Username46786/repos",
"events_url": "https://api.github.com/users/Username46786/events{/privacy}",
"received_events_url": "https://api.github.com/users/Username46786/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-21T23:32:40 | 2025-09-23T18:09:19 | 2025-09-23T18:09:19 | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41039",
"html_url": "https://github.com/huggingface/transformers/pull/41039",
"diff_url": "https://github.com/huggingface/transformers/pull/41039.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41039.patch",
"merged_at": null
} | Adds a focused guide for running Transformers in air-gapped/firewalled environments:
- Explains `TRANSFORMERS_OFFLINE`, `HF_HUB_OFFLINE`, and `local_files_only`
- Shows how to pre-populate & relocate caches
- Covers pipelines/trainers loading strictly from local files
- Adds entry to `_toctree.yml`
References:
- Offline mode mention in installation docs (historic) and offline flags discussions/issues.
| {
"login": "Username46786",
"id": 98800422,
"node_id": "U_kgDOBeOTJg",
"avatar_url": "https://avatars.githubusercontent.com/u/98800422?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Username46786",
"html_url": "https://github.com/Username46786",
"followers_url": "https://api.github.com/users/Username46786/followers",
"following_url": "https://api.github.com/users/Username46786/following{/other_user}",
"gists_url": "https://api.github.com/users/Username46786/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Username46786/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Username46786/subscriptions",
"organizations_url": "https://api.github.com/users/Username46786/orgs",
"repos_url": "https://api.github.com/users/Username46786/repos",
"events_url": "https://api.github.com/users/Username46786/events{/privacy}",
"received_events_url": "https://api.github.com/users/Username46786/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41039/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/41039/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41038 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41038/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41038/comments | https://api.github.com/repos/huggingface/transformers/issues/41038/events | https://github.com/huggingface/transformers/issues/41038 | 3,438,430,631 | I_kwDOCUB6oc7M8kmn | 41,038 | Method `from_pretrained` does not respect the mapping of tied tensors in `safetensors` checkpoint | {
"login": "daskol",
"id": 9336514,
"node_id": "MDQ6VXNlcjkzMzY1MTQ=",
"avatar_url": "https://avatars.githubusercontent.com/u/9336514?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/daskol",
"html_url": "https://github.com/daskol",
"followers_url": "https://api.github.com/users/daskol/followers",
"following_url": "https://api.github.com/users/daskol/following{/other_user}",
"gists_url": "https://api.github.com/users/daskol/gists{/gist_id}",
"starred_url": "https://api.github.com/users/daskol/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/daskol/subscriptions",
"organizations_url": "https://api.github.com/users/daskol/orgs",
"repos_url": "https://api.github.com/users/daskol/repos",
"events_url": "https://api.github.com/users/daskol/events{/privacy}",
"received_events_url": "https://api.github.com/users/daskol/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 | [] | 2025-09-21T17:20:26 | 2025-09-29T13:10:57 | 2025-09-29T13:10:57 | CONTRIBUTOR | null | null | null | null | ### System Info
The issue exists for `transformers>=4.54.0<=4.56.1` for sure (tested).
```bash
$ transformers env
- `transformers` version: 4.56.1
- Platform: Linux-6.16.7-arch1-1-x86_64-with-glibc2.42
- Python version: 3.13.7
- Huggingface_hub version: 0.35.0
- Safetensors version: 0.5.3
- Accelerate version: 1.10.1
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.8.0 (NA)
- Tensorflow version (GPU?): not installed (NA)
- Flax version (CPU?/GPU?/TPU?): 0.10.5 (cpu)
- Jax version: 0.7.0
- JaxLib version: 0.7.0
- Using distributed or parallel set-up in script?: <fill in>
```
### Who can help?
@CyrilVallez Method `AutoModel.from_pretrained` populates model weights from a checkpoint in `safetensors` format then it ties model weights in a way independent of checkpoint metadata and overwrite loaded weights on the real device with fake weights on `meta`. Here is relevant piece of code.
https://github.com/huggingface/transformers/blob/91393fe4cc3266a05bc0d129e34ff5f761bb46e2/src/transformers/modeling_utils.py#L5164-L5194
### Details
The `transformers` library loads weights first and then ties input embeddings to output embeddings as follows (see [transformers/modeling_utils.py#L3236][1]).
```python
def _tie_or_clone_weights(self, output_embeddings, input_embeddings):
if self.config.torchscript:
...
else:
output_embeddings.weight = input_embeddings.weight
```
However, the choice what keys of state dict to load depends on `transformers` implementation. Specifically, it makes decision based on available keys in state dict loaded from a checkpoint. This can bring to the situation when
1. it loads output embeddings but not input ones;
2. then it ties input embeddings to output ones.
As the result, both the input and output embeddings lives on `meta` device! In other words, no weights are loaded for bot input and output embeddings.
### Reproduction
The issue can be reproduced with a `convert.py` script from `safetensors` repo (see [bindings/python/convert.py#L216][2]). The routine `convert_file` sorts all keys alphabetically and takes the first item in the list. For example, it keeps `model.lm_head.weight` rather than `model.model.decoder.embed_tokens.weight` in case of OPT model.
```python
>>> from convert import convert_file
>>> convert_file('facebook/opt-125m/pytorch_model.bin', 'facebook/opt-125m/model.safetensors', [])
>>> from transformers import AutoModelForCausalLM
>>> model = AutoModelForCausalLM.from_pretrained('facebook/opt-125m')
>>> model.to('cpu')
NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.
```
[1]: https://github.com/huggingface/transformers/blob/v4.56.1/src/transformers/modeling_utils.py#L3236
[2]: https://github.com/huggingface/safetensors/blob/v0.6.2/bindings/python/convert.py#L216
### Information
- [x] 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
The issue can be reproduced with a `convert.py` script from `safetensors` repo (see [bindings/python/convert.py#L216][2]). The routine `convert_file` sorts all keys alphabetically and takes the first item in the list. For example, it keeps `model.lm_head.weight` rather than `model.model.decoder.embed_tokens.weight` in case of OPT model.
```python
>>> from convert import convert_file
>>> convert_file('facebook/opt-125m/pytorch_model.bin', 'facebook/opt-125m/model.safetensors', [])
>>> from transformers import AutoModelForCausalLM
>>> model = AutoModelForCausalLM.from_pretrained('facebook/opt-125m')
>>> model.to('cpu')
NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.
```
[1]: https://github.com/huggingface/transformers/blob/v4.56.1/src/transformers/modeling_utils.py#L3236
[2]: https://github.com/huggingface/safetensors/blob/v0.6.2/bindings/python/convert.py#L216
### Expected behavior
- All weights in state dict are restored from a checkpoint.
- No weights remains on `meta` device. | {
"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/41038/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/41038/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/41037 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41037/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41037/comments | https://api.github.com/repos/huggingface/transformers/issues/41037/events | https://github.com/huggingface/transformers/pull/41037 | 3,438,368,851 | PR_kwDOCUB6oc6ptYrU | 41,037 | Tests: Apertus integration tests | {
"login": "andresnowak",
"id": 35544006,
"node_id": "MDQ6VXNlcjM1NTQ0MDA2",
"avatar_url": "https://avatars.githubusercontent.com/u/35544006?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/andresnowak",
"html_url": "https://github.com/andresnowak",
"followers_url": "https://api.github.com/users/andresnowak/followers",
"following_url": "https://api.github.com/users/andresnowak/following{/other_user}",
"gists_url": "https://api.github.com/users/andresnowak/gists{/gist_id}",
"starred_url": "https://api.github.com/users/andresnowak/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/andresnowak/subscriptions",
"organizations_url": "https://api.github.com/users/andresnowak/orgs",
"repos_url": "https://api.github.com/users/andresnowak/repos",
"events_url": "https://api.github.com/users/andresnowak/events{/privacy}",
"received_events_url": "https://api.github.com/users/andresnowak/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-21T16:00:11 | 2025-10-03T13:40:10 | null | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41037",
"html_url": "https://github.com/huggingface/transformers/pull/41037",
"diff_url": "https://github.com/huggingface/transformers/pull/41037.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41037.patch",
"merged_at": null
} | # What does this PR do?
This PR adds integration tests for the Apertus model, the tests are:
- Check short greedy generation
- Check short instruct model greedy generation
- Check model logits in bfloat16, and compare with XIELU cuda output
- Check model logits in float32, and compare with XIELU cuda output (I think float32 tests are not desired?)
Note: Right now I do an environment variable just to show the difference between experimental XIELU cuda and no XIELU cuda outputs (but if this isn't desirable I'll remove it and just do a test for one of the two versions instead)
Related PR: #39381
@ArthurZucker
| null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41037/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/41037/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41036 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41036/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41036/comments | https://api.github.com/repos/huggingface/transformers/issues/41036/events | https://github.com/huggingface/transformers/pull/41036 | 3,438,266,279 | PR_kwDOCUB6oc6ptFNS | 41,036 | [Qwen3-next] Fix dimension mismatch in torch_chunk_gated_delta_rule and torch_recurrent_gated_delta_rule (#40963) | {
"login": "notkisk",
"id": 107971634,
"node_id": "U_kgDOBm-EMg",
"avatar_url": "https://avatars.githubusercontent.com/u/107971634?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/notkisk",
"html_url": "https://github.com/notkisk",
"followers_url": "https://api.github.com/users/notkisk/followers",
"following_url": "https://api.github.com/users/notkisk/following{/other_user}",
"gists_url": "https://api.github.com/users/notkisk/gists{/gist_id}",
"starred_url": "https://api.github.com/users/notkisk/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/notkisk/subscriptions",
"organizations_url": "https://api.github.com/users/notkisk/orgs",
"repos_url": "https://api.github.com/users/notkisk/repos",
"events_url": "https://api.github.com/users/notkisk/events{/privacy}",
"received_events_url": "https://api.github.com/users/notkisk/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-21T13:48:11 | 2025-09-24T11:19:11 | 2025-09-24T11:18:27 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41036",
"html_url": "https://github.com/huggingface/transformers/pull/41036",
"diff_url": "https://github.com/huggingface/transformers/pull/41036.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41036.patch",
"merged_at": "2025-09-24T11:18:27"
} | ## Fixes #40963
## Issue
When running inference with **Qwen3-next** (`modeling_qwen3_next.py`), a dimension mismatch occurs in the `Qwen3NextGatedDeltaNet` class inside the `torch_chunk_gated_delta_rule` function.
- Input tensors start as `(batch_size, seq_len, num_heads, head_dim)`.
- After `transpose(1, 2)`, the shape becomes `(batch_size, num_heads, seq_len, head_dim)`.
- The code incorrectly unpacks as `(batch_size, sequence_length, num_heads, k_head_dim)`, causing inconsistent calculations and runtime errors.
See [line 461](https://github.com/huggingface/transformers/blob/main/src/transformers/models/qwen3_next/modeling_qwen3_next.py#L461).
## Solution
Updated both `torch_chunk_gated_delta_rule` and `torch_recurrent_gated_delta_rule` in **`modular_qwen3_next.py`**:
- Corrected dimension unpacking:
```python
batch_size, num_heads, sequence_length, k_head_dim = key.shape
##Verification
- Reproduced the bug and confirmed the mismatch.
- Applied the fix and re-ran tests: 96 passed, 134 skipped (all as expected).
- Verified inference works correctly on the official example query:
“Give me a short introduction to large language model.”
##Root Cause
The bug was caused by assuming the original dimension order after transpose(1, 2).
This led to wrong unpacking of key.shape, which then broke padding, chunking, and recurrent state initialization.
The fix ensures consistent handling of dimensions after the transpose operation. | {
"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/41036/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/41036/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41035 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41035/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41035/comments | https://api.github.com/repos/huggingface/transformers/issues/41035/events | https://github.com/huggingface/transformers/pull/41035 | 3,438,183,599 | PR_kwDOCUB6oc6ps1vA | 41,035 | docs: update speech recognition examples to use modern Common Voice d… | {
"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
} | [] | open | false | null | [] | null | [] | 2025-09-21T11:59:16 | 2025-09-25T17:02:37 | null | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41035",
"html_url": "https://github.com/huggingface/transformers/pull/41035",
"diff_url": "https://github.com/huggingface/transformers/pull/41035.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41035.patch",
"merged_at": null
} |
# What does this PR do?
This PR updates the existing `"common_voice" \` dataset with modern `"mozilla-foundation/common_voice_11_0" \`.
Fixes #40977
## 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?
- [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
<!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @
Models:
- text models: @ArthurZucker
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41035/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/41035/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41034 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41034/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41034/comments | https://api.github.com/repos/huggingface/transformers/issues/41034/events | https://github.com/huggingface/transformers/issues/41034 | 3,437,876,949 | I_kwDOCUB6oc7M6dbV | 41,034 | G | {
"login": "ESCAPEEEE111",
"id": 223460714,
"node_id": "U_kgDODVG9ag",
"avatar_url": "https://avatars.githubusercontent.com/u/223460714?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/ESCAPEEEE111",
"html_url": "https://github.com/ESCAPEEEE111",
"followers_url": "https://api.github.com/users/ESCAPEEEE111/followers",
"following_url": "https://api.github.com/users/ESCAPEEEE111/following{/other_user}",
"gists_url": "https://api.github.com/users/ESCAPEEEE111/gists{/gist_id}",
"starred_url": "https://api.github.com/users/ESCAPEEEE111/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/ESCAPEEEE111/subscriptions",
"organizations_url": "https://api.github.com/users/ESCAPEEEE111/orgs",
"repos_url": "https://api.github.com/users/ESCAPEEEE111/repos",
"events_url": "https://api.github.com/users/ESCAPEEEE111/events{/privacy}",
"received_events_url": "https://api.github.com/users/ESCAPEEEE111/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-21T06:37:35 | 2025-09-21T06:37:38 | 2025-09-21T06:37:38 | NONE | null | null | null | null | null | {
"login": "ESCAPEEEE111",
"id": 223460714,
"node_id": "U_kgDODVG9ag",
"avatar_url": "https://avatars.githubusercontent.com/u/223460714?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/ESCAPEEEE111",
"html_url": "https://github.com/ESCAPEEEE111",
"followers_url": "https://api.github.com/users/ESCAPEEEE111/followers",
"following_url": "https://api.github.com/users/ESCAPEEEE111/following{/other_user}",
"gists_url": "https://api.github.com/users/ESCAPEEEE111/gists{/gist_id}",
"starred_url": "https://api.github.com/users/ESCAPEEEE111/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/ESCAPEEEE111/subscriptions",
"organizations_url": "https://api.github.com/users/ESCAPEEEE111/orgs",
"repos_url": "https://api.github.com/users/ESCAPEEEE111/repos",
"events_url": "https://api.github.com/users/ESCAPEEEE111/events{/privacy}",
"received_events_url": "https://api.github.com/users/ESCAPEEEE111/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41034/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/41034/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/41033 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41033/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41033/comments | https://api.github.com/repos/huggingface/transformers/issues/41033/events | https://github.com/huggingface/transformers/pull/41033 | 3,437,873,055 | PR_kwDOCUB6oc6pr2vu | 41,033 | feat: make audio feature extractors torch.export-able | {
"login": "The5cheduler",
"id": 46714886,
"node_id": "MDQ6VXNlcjQ2NzE0ODg2",
"avatar_url": "https://avatars.githubusercontent.com/u/46714886?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/The5cheduler",
"html_url": "https://github.com/The5cheduler",
"followers_url": "https://api.github.com/users/The5cheduler/followers",
"following_url": "https://api.github.com/users/The5cheduler/following{/other_user}",
"gists_url": "https://api.github.com/users/The5cheduler/gists{/gist_id}",
"starred_url": "https://api.github.com/users/The5cheduler/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/The5cheduler/subscriptions",
"organizations_url": "https://api.github.com/users/The5cheduler/orgs",
"repos_url": "https://api.github.com/users/The5cheduler/repos",
"events_url": "https://api.github.com/users/The5cheduler/events{/privacy}",
"received_events_url": "https://api.github.com/users/The5cheduler/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [
{
"id": 6470596964,
"node_id": "LA_kwDOCUB6oc8AAAABga15ZA",
"url": "https://api.github.com/repos/huggingface/transformers/labels/Audio",
"name": "Audio",
"color": "760453",
"default": false,
"description": ""
}
] | open | false | null | [] | null | [] | 2025-09-21T06:30:15 | 2025-09-26T16:49:32 | null | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41033",
"html_url": "https://github.com/huggingface/transformers/pull/41033",
"diff_url": "https://github.com/huggingface/transformers/pull/41033.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41033.patch",
"merged_at": null
} | Fix #40986
Refactors the feature extraction logic for both `GraniteSpeechFeatureExtractor` and `WhisperFeatureExtractor` into separate `nn.Module` subclasses. This encapsulates the feature extraction computation in a way that is compatible with `torch.export`.
A new method, `to_exportable_module()`, was added to both feature extractor classes to return an instance of these new modules.
The original `__call__` and `_torch_extract_fbank_features` methods were updated to use these exportable modules.
Finally, new tests were added to verify that both feature extractors can be successfully exported using `torch.export`.
# 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
- [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?
- [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.
<!-- 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41033/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/41033/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41032 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41032/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41032/comments | https://api.github.com/repos/huggingface/transformers/issues/41032/events | https://github.com/huggingface/transformers/pull/41032 | 3,437,852,840 | PR_kwDOCUB6oc6przFJ | 41,032 | fix(trainer): Avoid moving model with device_map | {
"login": "The5cheduler",
"id": 46714886,
"node_id": "MDQ6VXNlcjQ2NzE0ODg2",
"avatar_url": "https://avatars.githubusercontent.com/u/46714886?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/The5cheduler",
"html_url": "https://github.com/The5cheduler",
"followers_url": "https://api.github.com/users/The5cheduler/followers",
"following_url": "https://api.github.com/users/The5cheduler/following{/other_user}",
"gists_url": "https://api.github.com/users/The5cheduler/gists{/gist_id}",
"starred_url": "https://api.github.com/users/The5cheduler/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/The5cheduler/subscriptions",
"organizations_url": "https://api.github.com/users/The5cheduler/orgs",
"repos_url": "https://api.github.com/users/The5cheduler/repos",
"events_url": "https://api.github.com/users/The5cheduler/events{/privacy}",
"received_events_url": "https://api.github.com/users/The5cheduler/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-21T05:55:58 | 2025-09-29T14:32:09 | 2025-09-29T14:31:42 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41032",
"html_url": "https://github.com/huggingface/transformers/pull/41032",
"diff_url": "https://github.com/huggingface/transformers/pull/41032.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41032.patch",
"merged_at": "2025-09-29T14:31:42"
} | Fix #41013
When a model is loaded with `device_map="auto"` and is too large to fit on a single GPU, `accelerate` will offload some layers to the CPU or disk. The `Trainer` would previously attempt to move the entire model to the specified device, causing a `RuntimeError` because a model dispatched with `accelerate` hooks cannot be moved.
This commit fixes the issue by adding a check in `_move_model_to_device` to see if the model has an `hf_device_map` attribute. If it does, the device placement is assumed to be handled by `accelerate`, and the `model.to(device)` call is skipped.
A regression test is added to ensure the `Trainer` can be initialized with a model that has a `hf_device_map` that simulates offloading without raising an error.
# 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
- [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?
- [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.
<!-- 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| {
"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/41032/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/41032/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41031 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41031/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41031/comments | https://api.github.com/repos/huggingface/transformers/issues/41031/events | https://github.com/huggingface/transformers/pull/41031 | 3,437,772,920 | PR_kwDOCUB6oc6prkWV | 41,031 | Fix typos in English/Chinese documentation | {
"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 | [] | 2025-09-21T03:37:22 | 2025-09-22T11:33:46 | 2025-09-22T11:31:46 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41031",
"html_url": "https://github.com/huggingface/transformers/pull/41031",
"diff_url": "https://github.com/huggingface/transformers/pull/41031.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41031.patch",
"merged_at": "2025-09-22T11:31:46"
} | # What does this PR do?
Fix typos and fix other formatting errors.
## Before submitting
- [X] 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @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/41031/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/41031/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41030 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41030/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41030/comments | https://api.github.com/repos/huggingface/transformers/issues/41030/events | https://github.com/huggingface/transformers/pull/41030 | 3,437,732,326 | PR_kwDOCUB6oc6prbz0 | 41,030 | Remove optax | {
"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 | [] | 2025-09-21T02:41:12 | 2025-09-22T11:33:53 | 2025-09-22T11:30:39 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41030",
"html_url": "https://github.com/huggingface/transformers/pull/41030",
"diff_url": "https://github.com/huggingface/transformers/pull/41030.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41030.patch",
"merged_at": "2025-09-22T11:30:39"
} | # What does this PR do?
`optax` is dependency of `flax`.
| {
"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/41030/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/41030/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41029 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41029/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41029/comments | https://api.github.com/repos/huggingface/transformers/issues/41029/events | https://github.com/huggingface/transformers/pull/41029 | 3,437,721,485 | PR_kwDOCUB6oc6prZlf | 41,029 | Remove doc of tf and flax | {
"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 | [] | 2025-09-21T02:22:24 | 2025-09-22T12:52:57 | 2025-09-22T12:42:26 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41029",
"html_url": "https://github.com/huggingface/transformers/pull/41029",
"diff_url": "https://github.com/huggingface/transformers/pull/41029.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41029.patch",
"merged_at": "2025-09-22T12:42:26"
} | # What does this PR do?
Remove remaining text of tensor flow and flax inside `<tf> </tf>` and `<flax> </flax>` from documentation.
## 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.
- [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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| {
"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/41029/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/41029/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41028 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41028/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41028/comments | https://api.github.com/repos/huggingface/transformers/issues/41028/events | https://github.com/huggingface/transformers/pull/41028 | 3,437,699,407 | PR_kwDOCUB6oc6prVU1 | 41,028 | Fix typing of tuples | {
"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 | [] | 2025-09-21T01:42:35 | 2025-09-22T11:33:59 | 2025-09-22T11:29:08 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41028",
"html_url": "https://github.com/huggingface/transformers/pull/41028",
"diff_url": "https://github.com/huggingface/transformers/pull/41028.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41028.patch",
"merged_at": "2025-09-22T11:29:07"
} | # What does this PR do?
Use `tuple[T,...]` to indicate tuples of the same element type `T`.
| {
"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/41028/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/41028/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41027 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41027/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41027/comments | https://api.github.com/repos/huggingface/transformers/issues/41027/events | https://github.com/huggingface/transformers/pull/41027 | 3,437,577,034 | PR_kwDOCUB6oc6pq_Hq | 41,027 | docs: update speech recognition examples | {
"login": "tayo4christ",
"id": 58476114,
"node_id": "MDQ6VXNlcjU4NDc2MTE0",
"avatar_url": "https://avatars.githubusercontent.com/u/58476114?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/tayo4christ",
"html_url": "https://github.com/tayo4christ",
"followers_url": "https://api.github.com/users/tayo4christ/followers",
"following_url": "https://api.github.com/users/tayo4christ/following{/other_user}",
"gists_url": "https://api.github.com/users/tayo4christ/gists{/gist_id}",
"starred_url": "https://api.github.com/users/tayo4christ/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/tayo4christ/subscriptions",
"organizations_url": "https://api.github.com/users/tayo4christ/orgs",
"repos_url": "https://api.github.com/users/tayo4christ/repos",
"events_url": "https://api.github.com/users/tayo4christ/events{/privacy}",
"received_events_url": "https://api.github.com/users/tayo4christ/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-20T22:10:36 | 2025-09-30T08:38:31 | 2025-09-30T08:38:31 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41027",
"html_url": "https://github.com/huggingface/transformers/pull/41027",
"diff_url": "https://github.com/huggingface/transformers/pull/41027.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41027.patch",
"merged_at": "2025-09-30T08:38:31"
} | This PR modernizes the ASR examples and improves cross-platform onboarding.
## What’s changed
- **Pin CTC example commands** to Hub datasets (e.g., `mozilla-foundation/common_voice_17_0`) instead of `common_voice` local scripts.
- **Add a Whisper section tip** clarifying that `--dataset_name` should point to a Hub dataset to avoid “dataset scripts are no longer supported” errors.
- **Add a short Windows helper note** (PowerShell venv activation; formatter/linter commands without `make`).
- Minor copyedits for clarity.
## Why
Recent `datasets` versions deprecated local dataset scripts. Using Hub IDs is the supported path and:
- Prevents setup failures across environments (especially Windows),
- Reduces support overhead,
- Keeps docs future-proof and consistent.
## Files touched
- `examples/pytorch/speech-recognition/README.md`
- CTC commands now use `mozilla-foundation/common_voice_17_0` (retained existing `--dataset_config_name` values).
- Whisper section includes a one-liner tip about Hub datasets.
- Added a short Windows (PowerShell) note near the existing OMP note.
## How I tested
- Verified the updated commands on Windows (PowerShell, Python 3.12) in a clean venv.
- Ran Black/Ruff on the changed paths.
## Related
Fixes #40977
## Checklist
- [x] Docs/examples updated
- [x] Format/lint pass on changed files (Black/Ruff)
## Reviewers
@stevhliu (docs) @eustlb (speech) | {
"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/41027/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/41027/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41026 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41026/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41026/comments | https://api.github.com/repos/huggingface/transformers/issues/41026/events | https://github.com/huggingface/transformers/issues/41026 | 3,437,371,868 | I_kwDOCUB6oc7M4iHc | 41,026 | Attention sinks are not applied correctly in `integrations.flex_attention` | {
"login": "jonny-so",
"id": 3903806,
"node_id": "MDQ6VXNlcjM5MDM4MDY=",
"avatar_url": "https://avatars.githubusercontent.com/u/3903806?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/jonny-so",
"html_url": "https://github.com/jonny-so",
"followers_url": "https://api.github.com/users/jonny-so/followers",
"following_url": "https://api.github.com/users/jonny-so/following{/other_user}",
"gists_url": "https://api.github.com/users/jonny-so/gists{/gist_id}",
"starred_url": "https://api.github.com/users/jonny-so/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/jonny-so/subscriptions",
"organizations_url": "https://api.github.com/users/jonny-so/orgs",
"repos_url": "https://api.github.com/users/jonny-so/repos",
"events_url": "https://api.github.com/users/jonny-so/events{/privacy}",
"received_events_url": "https://api.github.com/users/jonny-so/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-20T18:03:27 | 2025-09-29T14:33:05 | 2025-09-29T14:33:05 | NONE | null | null | null | null | The `score_mod` function passed to `flex_attention` should operate on the pre-softmax attention scores, but the snippet below appears to be applying the attention biases (`s_aux`) and computing the **_post_**-softmax scores.
https://github.com/huggingface/transformers/blob/67097bf34055c55b886dc92014fd628c9a70e168/src/transformers/integrations/flex_attention.py#L275-L280
I don't think it is possible to apply (gpt-oss-style) attention sinks using the `score_mod` alone, but you can do it by passing `return_lse=True` to `flex_attention` and renormalising using the extra return value. If someone can point me to where unit tests for this code should live I'm happy to PR a fix. | {
"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/41026/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/41026/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/41025 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41025/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41025/comments | https://api.github.com/repos/huggingface/transformers/issues/41025/events | https://github.com/huggingface/transformers/pull/41025 | 3,437,282,088 | PR_kwDOCUB6oc6pqDlM | 41,025 | Adding support for Qwen3Omni | {
"login": "BakerBunker",
"id": 17872844,
"node_id": "MDQ6VXNlcjE3ODcyODQ0",
"avatar_url": "https://avatars.githubusercontent.com/u/17872844?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/BakerBunker",
"html_url": "https://github.com/BakerBunker",
"followers_url": "https://api.github.com/users/BakerBunker/followers",
"following_url": "https://api.github.com/users/BakerBunker/following{/other_user}",
"gists_url": "https://api.github.com/users/BakerBunker/gists{/gist_id}",
"starred_url": "https://api.github.com/users/BakerBunker/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/BakerBunker/subscriptions",
"organizations_url": "https://api.github.com/users/BakerBunker/orgs",
"repos_url": "https://api.github.com/users/BakerBunker/repos",
"events_url": "https://api.github.com/users/BakerBunker/events{/privacy}",
"received_events_url": "https://api.github.com/users/BakerBunker/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [
{
"id": 1843244711,
"node_id": "MDU6TGFiZWwxODQzMjQ0NzEx",
"url": "https://api.github.com/repos/huggingface/transformers/labels/New%20model",
"name": "New model",
"color": "fbca04",
"default": false,
"description": ""
}
] | closed | false | null | [] | null | [] | 2025-09-20T16:16:58 | 2025-09-22T08:46:41 | 2025-09-21T21:46:27 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41025",
"html_url": "https://github.com/huggingface/transformers/pull/41025",
"diff_url": "https://github.com/huggingface/transformers/pull/41025.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41025.patch",
"merged_at": "2025-09-21T21:46:27"
} | ## Qwen3-Omni here!
This PR introduces support for the upcoming **Qwen3-Omni** models, including **Instruct and Thinking versions.**
As the next generation of the Qwen-Omni family, Qwen3-Omni brings new architecture, multilingual and reasoning ability to omni model, achieving superior performance across complex multimodal tasks.
Special thanks to @ArthurZucker, and @zucchini-nlp for their valuable feedback and thorough reviews! 🙏 | {
"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/41025/reactions",
"total_count": 75,
"+1": 31,
"-1": 0,
"laugh": 0,
"hooray": 13,
"confused": 0,
"heart": 2,
"rocket": 29,
"eyes": 0
} | https://api.github.com/repos/huggingface/transformers/issues/41025/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41024 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41024/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41024/comments | https://api.github.com/repos/huggingface/transformers/issues/41024/events | https://github.com/huggingface/transformers/pull/41024 | 3,437,044,956 | PR_kwDOCUB6oc6ppYGq | 41,024 | Deprecate `max_size` in ConditionalDetrImageProcessor with warning | {
"login": "vaishnavigavi",
"id": 69896977,
"node_id": "MDQ6VXNlcjY5ODk2OTc3",
"avatar_url": "https://avatars.githubusercontent.com/u/69896977?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/vaishnavigavi",
"html_url": "https://github.com/vaishnavigavi",
"followers_url": "https://api.github.com/users/vaishnavigavi/followers",
"following_url": "https://api.github.com/users/vaishnavigavi/following{/other_user}",
"gists_url": "https://api.github.com/users/vaishnavigavi/gists{/gist_id}",
"starred_url": "https://api.github.com/users/vaishnavigavi/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/vaishnavigavi/subscriptions",
"organizations_url": "https://api.github.com/users/vaishnavigavi/orgs",
"repos_url": "https://api.github.com/users/vaishnavigavi/repos",
"events_url": "https://api.github.com/users/vaishnavigavi/events{/privacy}",
"received_events_url": "https://api.github.com/users/vaishnavigavi/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-20T11:25:27 | 2025-09-25T09:48:07 | null | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41024",
"html_url": "https://github.com/huggingface/transformers/pull/41024",
"diff_url": "https://github.com/huggingface/transformers/pull/41024.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41024.patch",
"merged_at": null
} | # What does this PR do?
Deprecates the `max_size` parameter in `ConditionalDetrImageProcessor` with a proper warning message and guides users to use the new `size={'longest_edge': <int>}` format instead.
## Why this change?
This aligns the ConditionalDetrImageProcessor with other processors (like DETR) and provides a clearer migration path for users. The current `max_size` parameter is deprecated and should be replaced with the more explicit `size` parameter format.
## Changes Made
- Added deprecation warning in `ConditionalDetrImageProcessor.__init__()` that warns users about the deprecated `max_size` parameter
- Updated warning message to guide users to use `size={'longest_edge': <int>}` instead
- Updated docstring in `from_dict` method to document the deprecation
- Simplified parameter handling by ignoring `max_size` instead of trying to use it
- Added test case to verify the deprecation warning works and processor still functions correctly
## 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?
This affects vision models and image processing, so tagging: @amyeroberts @qubvel | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41024/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/41024/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41023 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41023/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41023/comments | https://api.github.com/repos/huggingface/transformers/issues/41023/events | https://github.com/huggingface/transformers/issues/41023 | 3,437,015,054 | I_kwDOCUB6oc7M3LAO | 41,023 | modular_qwen2_5_vl.py冗余代码，影响代码整洁和引入歧义 | {
"login": "xiangqian19831224",
"id": 23026083,
"node_id": "MDQ6VXNlcjIzMDI2MDgz",
"avatar_url": "https://avatars.githubusercontent.com/u/23026083?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/xiangqian19831224",
"html_url": "https://github.com/xiangqian19831224",
"followers_url": "https://api.github.com/users/xiangqian19831224/followers",
"following_url": "https://api.github.com/users/xiangqian19831224/following{/other_user}",
"gists_url": "https://api.github.com/users/xiangqian19831224/gists{/gist_id}",
"starred_url": "https://api.github.com/users/xiangqian19831224/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/xiangqian19831224/subscriptions",
"organizations_url": "https://api.github.com/users/xiangqian19831224/orgs",
"repos_url": "https://api.github.com/users/xiangqian19831224/repos",
"events_url": "https://api.github.com/users/xiangqian19831224/events{/privacy}",
"received_events_url": "https://api.github.com/users/xiangqian19831224/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 | [] | 2025-09-20T10:39:08 | 2025-10-29T08:03:10 | 2025-10-29T08:03:10 | NONE | null | null | null | null | ### System Info
建议： 应该删除
理由：
1.请看我注释中的TODO部分，modular_qwen2_5_vl.py的功能基本上在modeling_qwen2_5_vl.py有最新实现
2.qwen2.5-vl的开源代码，没有用到modular_qwen2_5_vl.py
3.transformer中只是旧代码，可能是qwen2.5开发过程中的中间历史文件
4.该文件影响代码质量，引入疑惑
5.如果是这样，请确保代码的整洁与高质量
# coding=utf-8
# Copyright 2025 The Qwen Team and The HuggingFace Inc. team. All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""PyTorch Qwen2.5-VL model."""
from dataclasses import dataclass
from typing import List, Optional, Tuple, Union
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint
from torch.nn import CrossEntropyLoss
from transformers.models.qwen2_vl.configuration_qwen2_vl import Qwen2VLConfig
from transformers.models.qwen2_vl.modeling_qwen2_vl import (
PatchEmbed,
PatchMerger,
Qwen2RMSNorm,
Qwen2VLCausalLMOutputWithPast,
Qwen2VLForConditionalGeneration,
Qwen2VLModel,
Qwen2VLPreTrainedModel,
VisionAttention,
VisionRotaryEmbedding,
VisionSdpaAttention,
)
from transformers.models.qwen2_vl.processing_qwen2_vl import Qwen2VLImagesKwargs, Qwen2VLProcessor
from ...activations import ACT2FN
from ...configuration_utils import PretrainedConfig
from ...feature_extraction_utils import BatchFeature
from ...image_utils import ImageInput, VideoInput
from ...modeling_flash_attention_utils import is_flash_attn_available
from ...processing_utils import ProcessingKwargs, Unpack, VideosKwargs
from ...tokenization_utils_base import PreTokenizedInput, TextInput
from ...utils import logging
if is_flash_attn_available():
from ...modeling_flash_attention_utils import apply_rotary_emb, flash_attn_varlen_func
logger = logging.get_logger(__name__)
# TODO modeling_qwen2_5_vl.py实现过了 一样
def apply_rotary_pos_emb_flashatt(
q: torch.Tensor, k: torch.Tensor, cos: torch.Tensor, sin: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor]:
"""
输入：
q: query 向量，形状通常是 [batch, seq_len, num_heads, head_dim]
k: key 向量，形状同上
cos, sin: 预计算好的余弦、正弦位置编码张量，形状通常是 [seq_len, 1, head_dim] 或 [batch, seq_len, head_dim]
返回值：
返回旋转位置编码后的 query 和 key
"""
# cos.chunk(2, dim=-1) 表示在最后一维把 cos 切成两份，只取前一份
# RoPE 通常把 偶数维度和奇数维度配对 起来进行旋转，而这里只取一半，
# 是因为 apply_rotary_emb 内部会自动处理这对 (cos, sin) 与向量分解
# .contiguous() 保证内存布局是连续的，方便后面计算
cos = cos.chunk(2, dim=-1)[0].contiguous()
sin = sin.chunk(2, dim=-1)[0].contiguous()
# apply_rotary_emb 负责真正执行旋转编码
# .float()：防止半精度/混合精度下三角函数运算精度不够
# type_as(q)：再把结果转回和 q 一样的数据类型（比如 fp16/bf16
q_embed = apply_rotary_emb(q.float(), cos.float(), sin.float()).type_as(q)
k_embed = apply_rotary_emb(k.float(), cos.float(), sin.float()).type_as(k)
return q_embed, k_embed
class Qwen2_5_VLVisionConfig(PretrainedConfig):
"""
它主要用于保存模型结构超参数（hyperparameters），在初始化视觉模块时提供配置
构造参数
depth=32 Transformer block 的层数，即视觉编码器的深度
hidden_size=3584 隐藏层维度，每个 patch embedding/token 的表示维度
hidden_act="silu" 激活函数，SiLU（Swish）常用于 Transformer
intermediate_size=3420 FFN（前馈网络）中间层的维度，通常比 hidden_size 大
num_heads=16 多头注意力机制的头数
in_channels=3 输入图像通道数，RGB 图像为 3
patch_size=14 将图像切分为 patch 的大小，每个 patch 是 14x14
spatial_merge_size=2 空间合并参数，用于逐层减少 token 数（类似 pooling，提升计算效率）
temporal_patch_size=2 处理视频时的时间维度 patch 大小，例如把 2 帧合成 1 个 token
tokens_per_second=4 视频采样率控制，每秒生成多少个视觉 token
window_size=112 局部注意力的窗口大小，用于计算时的限制范围
举例
如果 patch_size=14，一张输入图像假设大小为 224 × 224
每个 patch = 14×14
总共 patch 数 = (224/14)² = 16 × 16 = 256 tokens
现在 window_size=112，意思是窗口覆盖 112 × 112 的原始像素区域：
每个窗口包含 (112/14)² = 8 × 8 = 64 tokens
注意力只在这 64 个 token 内计算，而不是 256 个 token 全部互算
这样复杂度从 256² 降到 64²，显著减少计算量
out_hidden_size=3584 输出 embedding 维度，通常与 hidden_size 相同，但也可能不同，用于和语言模型对齐
fullatt_block_indexes=[7, 15, 23, 31]
指定哪些层使用 全局注意力（full attention），其他层可能使用局部窗口注意力（windowed attention）以节省计算量
"""
model_type = "qwen2_5_vl"
base_config_key = "vision_config"
def __init__(
self,
depth=32,
hidden_size=3584,
hidden_act="silu",
intermediate_size=3420,
num_heads=16,
in_channels=3,
patch_size=14,
spatial_merge_size=2,
temporal_patch_size=2,
tokens_per_second=4,
window_size=112,
out_hidden_size=3584,
fullatt_block_indexes=[7, 15, 23, 31],
**kwargs,
):
super().__init__(**kwargs)
self.depth = depth
self.hidden_size = hidden_size
self.hidden_act = hidden_act
self.intermediate_size = intermediate_size
self.num_heads = num_heads
self.in_channels = in_channels
self.patch_size = patch_size
self.spatial_merge_size = spatial_merge_size
self.temporal_patch_size = temporal_patch_size
self.tokens_per_second = tokens_per_second
self.window_size = window_size
self.fullatt_block_indexes = fullatt_block_indexes
self.out_hidden_size = out_hidden_size
class Qwen2_5_VLConfig(Qwen2VLConfig):
model_type = "qwen2_5_vl"
sub_configs = {"vision_config": Qwen2_5_VLVisionConfig}
class Qwen2_5_VLMLP(nn.Module):
def __init__(self, config, bias: bool = False):
super().__init__()
self.hidden_size = config.hidden_size
self.intermediate_size = config.intermediate_size
self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=bias)
self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=bias)
self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=bias)
self.act_fn = ACT2FN[config.hidden_act]
def forward(self, hidden_state):
return self.down_proj(self.act_fn(self.gate_proj(hidden_state)) * self.up_proj(hidden_state))
# TODO PatchEmbed 依赖qwen2-vl的实现
# modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VisionPatchEmbed(PatchEmbed):
pass
# TODO PatchEmbed 依赖qwen2-vl的实现
# modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VisionRotaryEmbedding(VisionRotaryEmbedding):
pass
# TODO PatchEmbed 依赖qwen2-vl的实现
# modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VLPatchMerger(PatchMerger):
def __init__(self, dim: int, context_dim: int, spatial_merge_size: int = 2) -> None:
super().__init__(dim, context_dim, spatial_merge_size)
self.ln_q = Qwen2RMSNorm(context_dim, eps=1e-6)
class Qwen2_5_VLVisionFlashAttention2(nn.Module):
def __init__(self, dim: int, num_heads: int = 16) -> None:
super().__init__()
self.num_heads = num_heads
self.qkv = nn.Linear(dim, dim * 3, bias=True)
self.proj = nn.Linear(dim, dim)
def forward(
self,
hidden_states: torch.Tensor,
cu_seqlens: torch.Tensor,
rotary_pos_emb: Optional[torch.Tensor] = None,
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
) -> torch.Tensor:
"""
输入参数：
hidden_states: 输入的 token/patch 表示
注意这里代码使用的是把 batch 内所有样本的 token 拼接在一起的变长表示（见 cu_seqlens）
因此 hidden_states 的第 0 维是所有 tokens 的总和，形状通常是 (total_seq_len, dim)
cu_seqlens: cumulative sequence lengths，长度为 batch_size + 1
第一个元素通常是 0，最后一个是 total_seq_len。用于变长（varlen）FlashAttention
rotary_pos_emb：旧接口的 RoPE 值（可能是 theta/角度值），可选
position_embeddings：新接口，期望是 (cos, sin) 的 Tuple（预计算好的 cos 与 sin 矩阵）；二者互斥/替代
返回值：
output： (total_tokens, num_heads * head_dim)
当前实现下，每个句子在 q/k 上的 RoPE 不是从 0 开始，而是沿拼接序列连续旋转
"""
seq_length = hidden_states.shape[0]
# self.qkv(hidden_states)：对 (seq_length, dim) 应用线性层，输出形状 (seq_length, 3*dim)
# .reshape(seq_length, 3, self.num_heads, -1)：将最后一维拆成 3 × num_heads × head_dim，
# 其中 head_dim = dim // num_heads（最后用 -1 自动推断）
# 结果形状 (seq_length, 3, num_heads, head_dim)
# 注意：此处要求 dim 能被 num_heads 整除，否则 head_dim 不是整数，会报错或结果不符合预期
# .permute(1, 0, 2, 3)：把维度顺序改为 (3, seq_length, num_heads, head_dim)，即把 q/k/v 的轴放到最前面，便于拆分
# .unbind(0)：沿第 0 维（长度为 3 的那一维）拆出三个张量，分别对应 q、k、v
# 拆出的每个张量形状为 (seq_length, num_heads, head_dim)
# 最后通过解包赋给 q, k, v。
# 形状总结：q.shape == k.shape == v.shape == (seq_length, num_heads, head_dim)
q, k, v = self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
# 判断是否有 position_embeddings（外部预计算好的 cos/sin 对）
# 如果没有，走兼容旧接口的代码路径（使用 rotary_pos_emb）
if position_embeddings is None:
logger.warning_once(
"The attention layers in this model are transitioning from computing the RoPE embeddings internally "
"through `rotary_pos_emb` (2D tensor of RoPE theta values), to using externally computed "
"`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.54 `rotary_pos_emb` will be "
"removed and `position_embeddings` will be mandatory."
)
# 旧接口下把 rotary_pos_emb 拼接自身一次（在最后一个维度上）
# 解释与原因：通常 rotary_pos_emb 可能只包含 half-dim 的角度/频率值（或一种表示）
# 为方便得到与 head_dim 对齐的 cos/sin，此处通过复制来扩展维度（使其与 q/k 的最后维度匹配）
emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
cos = emb.cos()
sin = emb.sin()
else:
cos, sin = position_embeddings
# 1.q.unsqueeze(0):
# 把 q 从 (seq_length, num_heads, head_dim) 变为 (1, seq_length, num_heads, head_dim)，
# 即在最前面加一个 batch 维（因为 apply_rotary_pos_emb_flashatt 的实现可能期望有 batch 维）
# 2.k.unsqueeze(0): 同 k
# 3.apply_rotary_pos_emb_flashatt: TODO需要研究过去看看
# 它预计会使用提供的 cos 和 sin 对 q,k 做 RoPE（交错旋转/复数映射等）
# 它返回变换后的 (q,k)，形状仍为 (1, seq_length, num_heads, head_dim)
q, k = apply_rotary_pos_emb_flashatt(q.unsqueeze(0), k.unsqueeze(0), cos, sin)
q = q.squeeze(0)
k = k.squeeze(0)
max_seqlen = (cu_seqlens[1:] - cu_seqlens[:-1]).max().item()
# output： (total_tokens, num_heads * head_dim)
attn_output = flash_attn_varlen_func(q, k, v, cu_seqlens, cu_seqlens, max_seqlen, max_seqlen).reshape(
seq_length, -1
)
attn_output = self.proj(attn_output)
return attn_output
# TODO PatchEmbed 依赖qwen2-vl的实现
# modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VLVisionAttention(VisionAttention):
pass
# TODO PatchEmbed 依赖qwen2-vl的实现
# modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VLVisionSdpaAttention(VisionSdpaAttention):
pass
# eager: 即时模式注意力 直接按照普通 PyTorch 操作计算注意力，没有特殊优化；每一步操作立即执行
# flash_attention_2: 高效实现的注意力（利用 GPU 内存优化和并行计算），适合处理长序列或变长序列
# sdpa: 标准 Transformer 中的 Scaled Dot-Product Attention，按 Q·Kᵀ / √d → softmax → V 的流程计算
QWEN2_5_VL_VISION_ATTENTION_CLASSES = {
"eager": Qwen2_5_VLVisionAttention,
"flash_attention_2": Qwen2_5_VLVisionFlashAttention2,
"sdpa": Qwen2_5_VLVisionSdpaAttention,
}
# TODO 重复实现： modeling_qwen2_5_vl.py实现过了，一摸一样
class Qwen2_5_VLVisionBlock(nn.Module):
def __init__(self, config, attn_implementation: str = "sdpa") -> None:
super().__init__()
self.norm1 = Qwen2RMSNorm(config.hidden_size, eps=1e-6)
self.norm2 = Qwen2RMSNorm(config.hidden_size, eps=1e-6)
self.attn = QWEN2_5_VL_VISION_ATTENTION_CLASSES[attn_implementation](
config.hidden_size, num_heads=config.num_heads
)
self.mlp = Qwen2_5_VLMLP(config, bias=True)
def forward(
self,
hidden_states: torch.Tensor,
cu_seqlens: torch.Tensor,
rotary_pos_emb: Optional[torch.Tensor] = None,
position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
) -> torch.Tensor:
hidden_states = hidden_states + self.attn(
self.norm1(hidden_states),
cu_seqlens=cu_seqlens,
rotary_pos_emb=rotary_pos_emb,
position_embeddings=position_embeddings,
)
hidden_states = hidden_states + self.mlp(self.norm2(hidden_states))
return hidden_states
# TODO PatchEmbed 依赖qwen2-vl的实现
# modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VLPreTrainedModel(Qwen2VLPreTrainedModel):
pass
# TODO modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VisionTransformerPretrainedModel(Qwen2_5_VLPreTrainedModel):
config_class = Qwen2_5_VLVisionConfig
_no_split_modules = ["Qwen2_5_VLVisionBlock"]
def __init__(self, config, *inputs, **kwargs) -> None:
super().__init__(config, *inputs, **kwargs)
self.spatial_merge_size = config.spatial_merge_size
self.patch_size = config.patch_size
self.fullatt_block_indexes = config.fullatt_block_indexes
self.window_size = config.window_size
self.spatial_merge_unit = self.spatial_merge_size * self.spatial_merge_size
self.patch_embed = Qwen2_5_VisionPatchEmbed(
patch_size=config.patch_size,
temporal_patch_size=config.temporal_patch_size,
in_channels=config.in_channels,
embed_dim=config.hidden_size,
)
head_dim = config.hidden_size // config.num_heads
self.rotary_pos_emb = Qwen2_5_VisionRotaryEmbedding(head_dim // 2)
self.blocks = nn.ModuleList(
[Qwen2_5_VLVisionBlock(config, config._attn_implementation) for _ in range(config.depth)]
)
self.merger = Qwen2_5_VLPatchMerger(
dim=config.out_hidden_size,
context_dim=config.hidden_size,
spatial_merge_size=config.spatial_merge_size,
)
self.gradient_checkpointing = False
def rot_pos_emb(self, grid_thw):
pos_ids = []
for t, h, w in grid_thw:
hpos_ids = torch.arange(h).unsqueeze(1).expand(-1, w)
hpos_ids = hpos_ids.reshape(
h // self.spatial_merge_size,
self.spatial_merge_size,
w // self.spatial_merge_size,
self.spatial_merge_size,
)
hpos_ids = hpos_ids.permute(0, 2, 1, 3)
hpos_ids = hpos_ids.flatten()
wpos_ids = torch.arange(w).unsqueeze(0).expand(h, -1)
wpos_ids = wpos_ids.reshape(
h // self.spatial_merge_size,
self.spatial_merge_size,
w // self.spatial_merge_size,
self.spatial_merge_size,
)
wpos_ids = wpos_ids.permute(0, 2, 1, 3)
wpos_ids = wpos_ids.flatten()
pos_ids.append(torch.stack([hpos_ids, wpos_ids], dim=-1).repeat(t, 1))
pos_ids = torch.cat(pos_ids, dim=0)
max_grid_size = grid_thw[:, 1:].max()
rotary_pos_emb_full = self.rotary_pos_emb(max_grid_size)
rotary_pos_emb = rotary_pos_emb_full[pos_ids].flatten(1)
return rotary_pos_emb
def get_window_index(self, grid_thw):
window_index: list = []
cu_window_seqlens: list = [0]
window_index_id = 0
vit_merger_window_size = self.window_size // self.spatial_merge_size // self.patch_size
for grid_t, grid_h, grid_w in grid_thw:
llm_grid_h, llm_grid_w = (
grid_h // self.spatial_merge_size,
grid_w // self.spatial_merge_size,
)
index = torch.arange(grid_t * llm_grid_h * llm_grid_w).reshape(grid_t, llm_grid_h, llm_grid_w)
pad_h = vit_merger_window_size - llm_grid_h % vit_merger_window_size
pad_w = vit_merger_window_size - llm_grid_w % vit_merger_window_size
num_windows_h = (llm_grid_h + pad_h) // vit_merger_window_size
num_windows_w = (llm_grid_w + pad_w) // vit_merger_window_size
index_padded = F.pad(index, (0, pad_w, 0, pad_h), "constant", -100)
index_padded = index_padded.reshape(
grid_t,
num_windows_h,
vit_merger_window_size,
num_windows_w,
vit_merger_window_size,
)
index_padded = index_padded.permute(0, 1, 3, 2, 4).reshape(
grid_t,
num_windows_h * num_windows_w,
vit_merger_window_size,
vit_merger_window_size,
)
seqlens = (index_padded != -100).sum([2, 3]).reshape(-1)
index_padded = index_padded.reshape(-1)
index_new = index_padded[index_padded != -100]
window_index.append(index_new + window_index_id)
cu_seqlens_tmp = seqlens.cumsum(0) * self.spatial_merge_unit + cu_window_seqlens[-1]
cu_window_seqlens.extend(cu_seqlens_tmp.tolist())
window_index_id += (grid_t * llm_grid_h * llm_grid_w).item()
window_index = torch.cat(window_index, dim=0)
return window_index, cu_window_seqlens
def forward(self, hidden_states: torch.Tensor, grid_thw: torch.Tensor) -> torch.Tensor:
"""
Args:
hidden_states (`torch.Tensor` of shape `(seq_len, hidden_size)`):
The final hidden states of the model.
grid_thw (`torch.Tensor` of shape `(num_images_or_videos, 3)`):
The temporal, height and width of feature shape of each image in LLM.
Returns:
`torch.Tensor`: hidden_states.
"""
hidden_states = self.patch_embed(hidden_states)
rotary_pos_emb = self.rot_pos_emb(grid_thw)
window_index, cu_window_seqlens = self.get_window_index(grid_thw)
cu_window_seqlens = torch.tensor(
cu_window_seqlens,
device=hidden_states.device,
dtype=grid_thw.dtype if torch.jit.is_tracing() else torch.int32,
)
cu_window_seqlens = torch.unique_consecutive(cu_window_seqlens)
seq_len, _ = hidden_states.size()
hidden_states = hidden_states.reshape(seq_len // self.spatial_merge_unit, self.spatial_merge_unit, -1)
hidden_states = hidden_states[window_index, :, :]
hidden_states = hidden_states.reshape(seq_len, -1)
rotary_pos_emb = rotary_pos_emb.reshape(seq_len // self.spatial_merge_unit, self.spatial_merge_unit, -1)
rotary_pos_emb = rotary_pos_emb[window_index, :, :]
rotary_pos_emb = rotary_pos_emb.reshape(seq_len, -1)
emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
position_embeddings = (emb.cos(), emb.sin())
cu_seqlens = torch.repeat_interleave(grid_thw[:, 1] * grid_thw[:, 2], grid_thw[:, 0]).cumsum(
dim=0,
# Select dtype based on the following factors:
# - FA2 requires that cu_seqlens_q must have dtype int32
# - torch.onnx.export requires that cu_seqlens_q must have same dtype as grid_thw
# See https://github.com/huggingface/transformers/pull/34852 for more information
dtype=grid_thw.dtype if torch.jit.is_tracing() else torch.int32,
)
cu_seqlens = F.pad(cu_seqlens, (1, 0), value=0)
for layer_num, blk in enumerate(self.blocks):
if layer_num in self.fullatt_block_indexes:
cu_seqlens_now = cu_seqlens
else:
cu_seqlens_now = cu_window_seqlens
if self.gradient_checkpointing and self.training:
hidden_states = self._gradient_checkpointing_func(
blk.__call__, hidden_states, cu_seqlens_now, None, position_embeddings
)
else:
hidden_states = blk(hidden_states, cu_seqlens=cu_seqlens_now, position_embeddings=position_embeddings)
hidden_states = self.merger(hidden_states)
reverse_indices = torch.argsort(window_index)
hidden_states = hidden_states[reverse_indices, :]
return hidden_states
# TODO modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VLModel(Qwen2VLModel):
pass
# TODO modeling_qwen2_5_vl.py实现过了 但是不一样
@dataclass
class Qwen2_5_VLCausalLMOutputWithPast(Qwen2VLCausalLMOutputWithPast):
pass
# TODO modeling_qwen2_5_vl.py实现过了 但是不一样
class Qwen2_5_VLForConditionalGeneration(Qwen2VLForConditionalGeneration):
config_class = Qwen2_5_VLConfig
_no_split_modules = ["Qwen2_5_VLDecoderLayer", "Qwen2_5_VLVisionBlock"]
def __init__(self, config):
super().__init__(config)
self.visual = Qwen2_5_VisionTransformerPretrainedModel._from_config(config.vision_config)
def get_rope_index(
self,
input_ids: Optional[torch.LongTensor] = None,
image_grid_thw: Optional[torch.LongTensor] = None,
video_grid_thw: Optional[torch.LongTensor] = None,
second_per_grid_ts: Optional[torch.Tensor] = None,
attention_mask: Optional[torch.Tensor] = None,
) -> Tuple[torch.Tensor, torch.Tensor]:
"""
Calculate the 3D rope index based on image and video's temporal, height and width in LLM.
Explanation:
Each embedding sequence contains vision embedding and text embedding or just contains text embedding.
For pure text embedding sequence, the rotary position embedding has no difference with modern LLMs.
Examples:
input_ids: [T T T T T], here T is for text.
temporal position_ids: [0, 1, 2, 3, 4]
height position_ids: [0, 1, 2, 3, 4]
width position_ids: [0, 1, 2, 3, 4]
For vision and text embedding sequence, we calculate 3D rotary position embedding for vision part
and 1D rotary position embedding for text part.
Examples:
Temporal (Time): 3 patches, representing different segments of the video in time.
Height: 2 patches, dividing each frame vertically.
Width: 2 patches, dividing each frame horizontally.
We also have some important parameters:
fps (Frames Per Second): The video's frame rate, set to 1. This means one frame is processed each second.
tokens_per_second: This is a crucial parameter. It dictates how many "time-steps" or "temporal tokens" are conceptually packed into a one-second interval of the video. In this case, we have 25 tokens per second. So each second of the video will be represented with 25 separate time points. It essentially defines the temporal granularity.
temporal_patch_size: The number of frames that compose one temporal patch. Here, it's 2 frames.
interval: The step size for the temporal position IDs, calculated as tokens_per_second * temporal_patch_size / fps. In this case, 25 * 2 / 1 = 50. This means that each temporal patch will be have a difference of 50 in the temporal position IDs.
input_ids: [V V V V V V V V V V V V T T T T T], here V is for vision.
vision temporal position_ids: [0, 0, 0, 0, 50, 50, 50, 50, 100, 100, 100, 100]
vision height position_ids: [0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1]
vision width position_ids: [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1]
text temporal position_ids: [101, 102, 103, 104, 105]
text height position_ids: [101, 102, 103, 104, 105]
text width position_ids: [101, 102, 103, 104, 105]
Here we calculate the text start position_ids as the max vision position_ids plus 1.
Args:
input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
it.
image_grid_thw (`torch.LongTensor` of shape `(num_images, 3)`, *optional*):
The temporal, height and width of feature shape of each image in LLM.
video_grid_thw (`torch.LongTensor` of shape `(num_videos, 3)`, *optional*):
The temporal, height and width of feature shape of each video in LLM.
second_per_grid_ts (`torch.Tensor` of shape `(num_videos)`, *optional*):
The time interval (in seconds) for each grid along the temporal dimension in the 3D position IDs.
attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
- 1 for tokens that are **not masked**,
- 0 for tokens that are **masked**.
Returns:
position_ids (`torch.LongTensor` of shape `(3, batch_size, sequence_length)`)
mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`)
"""
spatial_merge_size = self.config.vision_config.spatial_merge_size
image_token_id = self.config.image_token_id
video_token_id = self.config.video_token_id
vision_start_token_id = self.config.vision_start_token_id
mrope_position_deltas = []
if input_ids is not None and (image_grid_thw is not None or video_grid_thw is not None):
total_input_ids = input_ids
if attention_mask is None:
attention_mask = torch.ones_like(total_input_ids)
position_ids = torch.ones(
3,
input_ids.shape[0],
input_ids.shape[1],
dtype=input_ids.dtype,
device=input_ids.device,
)
image_index, video_index = 0, 0
attention_mask = attention_mask.to(total_input_ids.device)
for i, input_ids in enumerate(total_input_ids):
input_ids = input_ids[attention_mask[i] == 1]
image_nums, video_nums = 0, 0
vision_start_indices = torch.argwhere(input_ids == vision_start_token_id).squeeze(1)
vision_tokens = input_ids[vision_start_indices + 1]
image_nums = (vision_tokens == image_token_id).sum()
video_nums = (vision_tokens == video_token_id).sum()
input_tokens = input_ids.tolist()
llm_pos_ids_list: list = []
st = 0
remain_images, remain_videos = image_nums, video_nums
for _ in range(image_nums + video_nums):
if image_token_id in input_tokens and remain_images > 0:
ed_image = input_tokens.index(image_token_id, st)
else:
ed_image = len(input_tokens) + 1
if video_token_id in input_tokens and remain_videos > 0:
ed_video = input_tokens.index(video_token_id, st)
else:
ed_video = len(input_tokens) + 1
if ed_image < ed_video:
t, h, w = (
image_grid_thw[image_index][0],
image_grid_thw[image_index][1],
image_grid_thw[image_index][2],
)
second_per_grid_t = 0
image_index += 1
remain_images -= 1
ed = ed_image
else:
t, h, w = (
video_grid_thw[video_index][0],
video_grid_thw[video_index][1],
video_grid_thw[video_index][2],
)
if second_per_grid_ts is not None:
second_per_grid_t = second_per_grid_ts[video_index]
else:
second_per_grid_t = 1.0
video_index += 1
remain_videos -= 1
ed = ed_video
llm_grid_t, llm_grid_h, llm_grid_w = (
t.item(),
h.item() // spatial_merge_size,
w.item() // spatial_merge_size,
)
text_len = ed - st
st_idx = llm_pos_ids_list[-1].max() + 1 if len(llm_pos_ids_list) > 0 else 0
llm_pos_ids_list.append(torch.arange(text_len).view(1, -1).expand(3, -1) + st_idx)
range_tensor = torch.arange(llm_grid_t).view(-1, 1)
expanded_range = range_tensor.expand(-1, llm_grid_h * llm_grid_w)
time_tensor = expanded_range * second_per_grid_t * self.config.vision_config.tokens_per_second
time_tensor_long = time_tensor.long()
t_index = time_tensor_long.flatten()
h_index = torch.arange(llm_grid_h).view(1, -1, 1).expand(llm_grid_t, -1, llm_grid_w).flatten()
w_index = torch.arange(llm_grid_w).view(1, 1, -1).expand(llm_grid_t, llm_grid_h, -1).flatten()
llm_pos_ids_list.append(torch.stack([t_index, h_index, w_index]) + text_len + st_idx)
st = ed + llm_grid_t * llm_grid_h * llm_grid_w
if st < len(input_tokens):
st_idx = llm_pos_ids_list[-1].max() + 1 if len(llm_pos_ids_list) > 0 else 0
text_len = len(input_tokens) - st
llm_pos_ids_list.append(torch.arange(text_len).view(1, -1).expand(3, -1) + st_idx)
llm_positions = torch.cat(llm_pos_ids_list, dim=1).reshape(3, -1)
position_ids[..., i, attention_mask[i] == 1] = llm_positions.to(position_ids.device)
mrope_position_deltas.append(llm_positions.max() + 1 - len(total_input_ids[i]))
mrope_position_deltas = torch.tensor(mrope_position_deltas, device=input_ids.device).unsqueeze(1)
return position_ids, mrope_position_deltas
else:
if attention_mask is not None:
position_ids = attention_mask.long().cumsum(-1) - 1
position_ids.masked_fill_(attention_mask == 0, 1)
position_ids = position_ids.unsqueeze(0).expand(3, -1, -1).to(attention_mask.device)
max_position_ids = position_ids.max(0, keepdim=False)[0].max(-1, keepdim=True)[0]
mrope_position_deltas = max_position_ids + 1 - attention_mask.shape[-1]
else:
position_ids = (
torch.arange(input_ids.shape[1], device=input_ids.device)
.view(1, 1, -1)
.expand(3, input_ids.shape[0], -1)
)
mrope_position_deltas = torch.zeros(
[input_ids.shape[0], 1],
device=input_ids.device,
dtype=input_ids.dtype,
)
return position_ids, mrope_position_deltas
def forward(
self,
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
past_key_values: Optional[List[torch.FloatTensor]] = None,
inputs_embeds: Optional[torch.FloatTensor] = None,
labels: Optional[torch.LongTensor] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
pixel_values: Optional[torch.Tensor] = None,
pixel_values_videos: Optional[torch.FloatTensor] = None,
image_grid_thw: Optional[torch.LongTensor] = None,
video_grid_thw: Optional[torch.LongTensor] = None,
rope_deltas: Optional[torch.LongTensor] = None,
cache_position: Optional[torch.LongTensor] = None,
second_per_grid_ts: Optional[torch.Tensor] = None,
) -> Union[Tuple, Qwen2_5_VLCausalLMOutputWithPast]:
r"""
labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
(masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
Returns:
Example:
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, Qwen2_5_VLForConditionalGeneration
>>> model = Qwen2_5_VLForConditionalGeneration.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")
>>> processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")
>>> messages = [
{
"role": "user",
"content": [
{"type": "image"},
{"type": "text", "text": "What is shown in this image?"},
],
},
]
>>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
>>> inputs = processor(text=[text], images=[image], vision_infos=[vision_infos])
>>> # Generate
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
"The image shows a street scene with a red stop sign in the foreground. In the background, there is a large red gate with Chinese characters ..."
```"""
output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
output_hidden_states = (
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
)
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
if inputs_embeds is None:
inputs_embeds = self.model.embed_tokens(input_ids)
if pixel_values is not None:
pixel_values = pixel_values.type(self.visual.dtype)
image_embeds = self.visual(pixel_values, grid_thw=image_grid_thw)
n_image_tokens = (input_ids == self.config.image_token_id).sum().item()
n_image_features = image_embeds.shape[0]
if n_image_tokens != n_image_features:
raise ValueError(
f"Image features and image tokens do not match: tokens: {n_image_tokens}, features {n_image_features}"
)
mask = input_ids == self.config.image_token_id
mask_unsqueezed = mask.unsqueeze(-1)
mask_expanded = mask_unsqueezed.expand_as(inputs_embeds)
image_mask = mask_expanded.to(inputs_embeds.device)
image_embeds = image_embeds.to(inputs_embeds.device, inputs_embeds.dtype)
inputs_embeds = inputs_embeds.masked_scatter(image_mask, image_embeds)
if pixel_values_videos is not None:
pixel_values_videos = pixel_values_videos.type(self.visual.dtype)
video_embeds = self.visual(pixel_values_videos, grid_thw=video_grid_thw)
n_video_tokens = (input_ids == self.config.video_token_id).sum().item()
n_video_features = video_embeds.shape[0]
if n_video_tokens != n_video_features:
raise ValueError(
f"Video features and video tokens do not match: tokens: {n_video_tokens}, features {n_video_features}"
)
mask = input_ids == self.config.video_token_id
mask_unsqueezed = mask.unsqueeze(-1)
mask_expanded = mask_unsqueezed.expand_as(inputs_embeds)
video_mask = mask_expanded.to(inputs_embeds.device)
video_embeds = video_embeds.to(inputs_embeds.device, inputs_embeds.dtype)
inputs_embeds = inputs_embeds.masked_scatter(video_mask, video_embeds)
if attention_mask is not None:
attention_mask = attention_mask.to(inputs_embeds.device)
# if we get 4D attention mask we cannot calculate rope deltas anymore. TODO @raushan fixme
if position_ids is None and (attention_mask is None or attention_mask.ndim == 2):
# calculate RoPE index once per generation in the pre-fill stage only
if (
(cache_position is not None and cache_position[0] == 0)
or self.rope_deltas is None
or (past_key_values is None or past_key_values.get_seq_length() == 0)
):
position_ids, rope_deltas = self.get_rope_index(
input_ids,
image_grid_thw,
video_grid_thw,
second_per_grid_ts,
attention_mask,
)
self.rope_deltas = rope_deltas
# then use the prev pre-calculated rope-deltas to get the correct position ids
else:
batch_size, seq_length, _ = inputs_embeds.shape
delta = (
(cache_position[0] + self.rope_deltas).to(inputs_embeds.device)
if cache_position is not None
else 0
)
position_ids = torch.arange(seq_length, device=inputs_embeds.device)
position_ids = position_ids.view(1, -1).expand(batch_size, -1)
if cache_position is not None: # otherwise `deltas` is an int `0`
delta = delta.repeat_interleave(batch_size // delta.shape[0], dim=0)
position_ids = position_ids.add(delta)
position_ids = position_ids.unsqueeze(0).expand(3, -1, -1)
outputs = self.model(
input_ids=None,
position_ids=position_ids,
attention_mask=attention_mask,
past_key_values=past_key_values,
inputs_embeds=inputs_embeds,
use_cache=use_cache,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
cache_position=cache_position,
)
hidden_states = outputs[0]
logits = self.lm_head(hidden_states)
loss = None
if labels is not None:
# Upcast to float if we need to compute the loss to avoid potential precision issues
logits = logits.float()
# Shift so that tokens < n predict n
shift_logits = logits[..., :-1, :].contiguous()
shift_labels = labels[..., 1:].contiguous()
# Flatten the tokens
loss_fct = CrossEntropyLoss()
shift_logits = shift_logits.view(-1, self.config.vocab_size)
shift_labels = shift_labels.view(-1)
# Enable model parallelism
shift_labels = shift_labels.to(shift_logits.device)
loss = loss_fct(shift_logits, shift_labels)
if not return_dict:
output = (logits,) + outputs[1:]
return (loss,) + output if loss is not None else output
return Qwen2_5_VLCausalLMOutputWithPast(
loss=loss,
logits=logits,
past_key_values=outputs.past_key_values,
hidden_states=outputs.hidden_states,
attentions=outputs.attentions,
rope_deltas=self.rope_deltas,
)
def prepare_inputs_for_generation(
self,
input_ids,
past_key_values=None,
attention_mask=None,
inputs_embeds=None,
cache_position=None,
position_ids=None,
use_cache=True,
pixel_values=None,
pixel_values_videos=None,
image_grid_thw=None,
video_grid_thw=None,
second_per_grid_ts=None,
**kwargs,
):
# Overwritten -- in specific circumstances we don't want to forward image inputs to the model
model_inputs = super().prepare_inputs_for_generation(
input_ids,
past_key_values=past_key_values,
attention_mask=attention_mask,
inputs_embeds=inputs_embeds,
cache_position=cache_position,
position_ids=position_ids,
pixel_values=pixel_values,
pixel_values_videos=pixel_values_videos,
image_grid_thw=image_grid_thw,
video_grid_thw=video_grid_thw,
second_per_grid_ts=second_per_grid_ts,
use_cache=use_cache,
**kwargs,
)
# Qwen2-5-VL position_ids are prepareed with rope_deltas in forward
model_inputs["position_ids"] = None
if cache_position[0] != 0:
model_inputs["pixel_values"] = None
model_inputs["pixel_values_videos"] = None
return model_inputs
class Qwen2_5_VLVideosProcessorKwargs(VideosKwargs, total=False):
# fps 字段表示视频的帧率
fps: Union[List[float], float]
class Qwen2_5_VLImagesKwargs(Qwen2VLImagesKwargs):
pass
class Qwen2_5_VLProcessorKwargs(ProcessingKwargs, total=False):
images_kwargs: Qwen2_5_VLImagesKwargs
videos_kwargs: Qwen2_5_VLVideosProcessorKwargs
_defaults = {
"text_kwargs": {
"padding": False,
},
"videos_kwargs": {"fps": 2.0},
}
class Qwen2_5_VLProcessor(Qwen2VLProcessor):
r"""
Constructs a Qwen2.5-VL processor which wraps a Qwen2.5-VL image processor and a Qwen2 tokenizer into a single processor.
[`Qwen2_5_VLProcessor`] offers all the functionalities of [`Qwen2VLImageProcessor`] and [`Qwen2TokenizerFast`]. See the
[`~Qwen2_5_VLProcessor.__call__`] and [`~Qwen2_5_VLProcessor.decode`] for more information.
Args:
image_processor ([`Qwen2VLImageProcessor`], *optional*):
The image processor is a required input.
tokenizer ([`Qwen2TokenizerFast`], *optional*):
The tokenizer is a required input.
chat_template (`str`, *optional*): A Jinja template which will be used to convert lists of messages
in a chat into a tokenizable string.
"""
image_processor_class = "AutoImageProcessor"
@property
def model_input_names(self):
tokenizer_input_names = self.tokenizer.model_input_names
image_processor_input_names = self.image_processor.model_input_names
names_from_processor = list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
return names_from_processor + ["second_per_grid_ts"]
def __call__(
self,
images: ImageInput = None,
text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None,
videos: VideoInput = None,
**kwargs: Unpack[Qwen2_5_VLProcessorKwargs],
) -> BatchFeature:
"""
Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
and `kwargs` arguments to Qwen2TokenizerFast's [`~Qwen2TokenizerFast.__call__`] if `text` is not `None` to encode
the text. To prepare the vision inputs, this method forwards the `vision_infos` and `kwrags` arguments to
Qwen2VLImageProcessor's [`~Qwen2VLImageProcessor.__call__`] if `vision_infos` is not `None`.
Args:
images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
tensor. Both channels-first and channels-last formats are supported.
text (`str`, `List[str]`, `List[List[str]]`):
The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
(pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
`is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
videos (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`):
The image or batch of videos to be prepared. Each video can be a 4D NumPy array or PyTorch
tensor, or a nested list of 3D frames. Both channels-first and channels-last formats are supported.
return_tensors (`str` or [`~utils.TensorType`], *optional*):
If set, will return tensors of a particular framework. Acceptable values are:
- `'tf'`: Return TensorFlow `tf.constant` objects.
- `'pt'`: Return PyTorch `torch.Tensor` objects.
- `'np'`: Return NumPy `np.ndarray` objects.
- `'jax'`: Return JAX `jnp.ndarray` objects.
Returns:
[`BatchFeature`]: A [`BatchFeature`] with the following fields:
- **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
- **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
`return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not
`None`).
- **pixel_values** -- Pixel values to be fed to a model. Returned when `images` is not `None`.
- **pixel_values_videos** -- Pixel values of videos to be fed to a model. Returned when `videos` is not `None`.
- **image_grid_thw** -- List of image 3D grid in LLM. Returned when `images` is not `None`.
- **video_grid_thw** -- List of video 3D grid in LLM. Returned when `videos` is not `None`.
- **second_per_grid_ts** -- List of video seconds per time grid. Returned when `videos` is not `None`.
"""
output_kwargs = self._merge_kwargs(
Qwen2_5_VLProcessorKwargs,
tokenizer_init_kwargs=self.tokenizer.init_kwargs,
**kwargs,
)
if images is not None:
image_inputs = self.image_processor(images=images, videos=None, **output_kwargs["images_kwargs"])
image_grid_thw = image_inputs["image_grid_thw"]
else:
image_inputs = {}
image_grid_thw = None
if videos is not None:
videos_inputs = self.image_processor(images=None, videos=videos, **output_kwargs["images_kwargs"])
video_grid_thw = videos_inputs["video_grid_thw"]
fps = output_kwargs["videos_kwargs"].pop("fps", 2.0)
if isinstance(fps, (int, float)):
second_per_grid_ts = [self.image_processor.temporal_patch_size / fps] * len(video_grid_thw)
elif hasattr(fps, "__len__") and len(fps) == len(video_grid_thw):
second_per_grid_ts = [self.image_processor.temporal_patch_size / tmp for tmp in fps]
else:
raise ValueError(
f"The length of fps ({len(fps) if hasattr(fps, '__len__') else fps}) must be equal to the length of video_grid_thw ({len(video_grid_thw)}) or fps should be a single number."
)
videos_inputs.update({"second_per_grid_ts": second_per_grid_ts})
else:
videos_inputs = {}
video_grid_thw = None
if not isinstance(text, list):
text = [text]
if image_grid_thw is not None:
merge_length = self.image_processor.merge_size ** 2
index = 0
for i in range(len(text)):
while self.image_token in text[i]:
text[i] = text[i].replace(
self.image_token,
"<|placeholder|>" * (image_grid_thw[index].prod() // merge_length),
1,
)
index += 1
text[i] = text[i].replace("<|placeholder|>", self.image_token)
if video_grid_thw is not None:
merge_length = self.image_processor.merge_size ** 2
index = 0
for i in range(len(text)):
while self.video_token in text[i]:
text[i] = text[i].replace(
self.video_token,
"<|placeholder|>" * (video_grid_thw[index].prod() // merge_length),
1,
)
index += 1
text[i] = text[i].replace("<|placeholder|>", self.video_token)
text_inputs = self.tokenizer(text, **output_kwargs["text_kwargs"])
return BatchFeature(data={**text_inputs, **image_inputs, **videos_inputs})
__all__ = [
"Qwen2_5_VLConfig",
"Qwen2_5_VLForConditionalGeneration",
"Qwen2_5_VLModel",
"Qwen2_5_VLPreTrainedModel",
"Qwen2_5_VLProcessor",
]
### Who can help?
@vanpelt
### 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
删除无用代码
### Expected behavior
删除无用代码 | {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41023/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/41023/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/41022 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41022/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41022/comments | https://api.github.com/repos/huggingface/transformers/issues/41022/events | https://github.com/huggingface/transformers/pull/41022 | 3,436,906,853 | PR_kwDOCUB6oc6po9od | 41,022 | 🌐 [i18n-KO] Translated `backbones.md` to Korean | {
"login": "yijun-lee",
"id": 119404328,
"node_id": "U_kgDOBx33KA",
"avatar_url": "https://avatars.githubusercontent.com/u/119404328?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/yijun-lee",
"html_url": "https://github.com/yijun-lee",
"followers_url": "https://api.github.com/users/yijun-lee/followers",
"following_url": "https://api.github.com/users/yijun-lee/following{/other_user}",
"gists_url": "https://api.github.com/users/yijun-lee/gists{/gist_id}",
"starred_url": "https://api.github.com/users/yijun-lee/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/yijun-lee/subscriptions",
"organizations_url": "https://api.github.com/users/yijun-lee/orgs",
"repos_url": "https://api.github.com/users/yijun-lee/repos",
"events_url": "https://api.github.com/users/yijun-lee/events{/privacy}",
"received_events_url": "https://api.github.com/users/yijun-lee/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-20T08:24:27 | 2025-09-22T02:59:56 | null | CONTRIBUTOR | null | null | true | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41022",
"html_url": "https://github.com/huggingface/transformers/pull/41022",
"diff_url": "https://github.com/huggingface/transformers/pull/41022.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41022.patch",
"merged_at": null
} | # What does this PR do?
Translated the `backbones.md` file of the documentation to Korean.
Thank you in advance for your review.
Part of https://github.com/huggingface/transformers/issues/20179
## Before reviewing
- [x] Check for missing / redundant translations (번역 누락/중복 검사)
- [x] Grammar Check (맞춤법 검사)
- [x] Review or Add new terms to glossary (용어 확인 및 추가)
- [x] Check Inline TOC (e.g. `[[lowercased-header]]`)
- [x] Check live-preview for gotchas (live-preview로 정상작동 확인)
## Who can review? (Initial)
<!-- 1. 위 체크가 모두 완료된 뒤에만 KREW 팀원들에게 리뷰를 요청하는 아래 주석을 노출해주세요!-->
May you please review this PR?
<!-- @jungnerd, @luckyvickyricky, @chelsseeey, @skwh54, @maximizemaxwell, @D15M4S -->
<!-- @harheem, @nsbg, @Youngdong2, @xhaktm00, @judy-choi -->
<!-- @4N3MONE, @Kim-Ju-won, @ahnjj, @FacerAin, @ssum21, @TaskerJang, @HyunZ118 -->
<!-- @yijun-lee, @chhaewxn, @AhnJoonSung, @seopp, @pyapyapya -->
@AhnJoonSung, @nsbg, @Kim-Ju-won, @FacerAin, @judy-choi, @HyunZ118, @jungnerd
## 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#start-contributing-pull-requests),
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? (Final)
<!-- 2. KREW 팀원들의 리뷰가 끝난 후에 아래 주석을 노출해주세요! -->
<!-- @stevhliu May you please review this PR? -->
| null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41022/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/41022/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41021 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41021/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41021/comments | https://api.github.com/repos/huggingface/transformers/issues/41021/events | https://github.com/huggingface/transformers/pull/41021 | 3,436,905,475 | PR_kwDOCUB6oc6po9WJ | 41,021 | 🌐 [i18n-KO] Translated `video_processors.md` to Korean | {
"login": "AhnJoonSung",
"id": 53860803,
"node_id": "MDQ6VXNlcjUzODYwODAz",
"avatar_url": "https://avatars.githubusercontent.com/u/53860803?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/AhnJoonSung",
"html_url": "https://github.com/AhnJoonSung",
"followers_url": "https://api.github.com/users/AhnJoonSung/followers",
"following_url": "https://api.github.com/users/AhnJoonSung/following{/other_user}",
"gists_url": "https://api.github.com/users/AhnJoonSung/gists{/gist_id}",
"starred_url": "https://api.github.com/users/AhnJoonSung/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/AhnJoonSung/subscriptions",
"organizations_url": "https://api.github.com/users/AhnJoonSung/orgs",
"repos_url": "https://api.github.com/users/AhnJoonSung/repos",
"events_url": "https://api.github.com/users/AhnJoonSung/events{/privacy}",
"received_events_url": "https://api.github.com/users/AhnJoonSung/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-20T08:23:26 | 2025-10-18T05:42:43 | null | CONTRIBUTOR | null | null | true | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41021",
"html_url": "https://github.com/huggingface/transformers/pull/41021",
"diff_url": "https://github.com/huggingface/transformers/pull/41021.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41021.patch",
"merged_at": null
} | # What does this PR do?
Translated the `video_processors.md` file of the documentation to Korean.
Thank you in advance for your review.
Part of https://github.com/huggingface/transformers/issues/20179
## Before reviewing
- [x] Check for missing / redundant translations (번역 누락/중복 검사)
- [x] Grammar Check (맞춤법 검사)
- [x] Review or Add new terms to glossary (용어 확인 및 추가)
- [x] Check Inline TOC (e.g. `[[lowercased-header]]`)
- [x] Check live-preview for gotchas (live-preview로 정상작동 확인)
## Who can review? (Initial)
May you please review this PR?
@jungnerd, @yijun-lee, @Kim-Ju-won, @FacerAin, @judy-choi, @maximizemaxwell, @nsbg, @HyunZ118
## 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#start-contributing-pull-requests),
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? (Final)
<!-- 2. KREW 팀원들의 리뷰가 끝난 후에 아래 주석을 노출해주세요! -->
@stevhliu May you please review this PR? | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41021/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/41021/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41020 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41020/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41020/comments | https://api.github.com/repos/huggingface/transformers/issues/41020/events | https://github.com/huggingface/transformers/issues/41020 | 3,436,871,790 | I_kwDOCUB6oc7M2oBu | 41,020 | Config rope_scaling and text_config.rope_scaling might be the same or different dict objects | {
"login": "albertvillanova",
"id": 8515462,
"node_id": "MDQ6VXNlcjg1MTU0NjI=",
"avatar_url": "https://avatars.githubusercontent.com/u/8515462?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/albertvillanova",
"html_url": "https://github.com/albertvillanova",
"followers_url": "https://api.github.com/users/albertvillanova/followers",
"following_url": "https://api.github.com/users/albertvillanova/following{/other_user}",
"gists_url": "https://api.github.com/users/albertvillanova/gists{/gist_id}",
"starred_url": "https://api.github.com/users/albertvillanova/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/albertvillanova/subscriptions",
"organizations_url": "https://api.github.com/users/albertvillanova/orgs",
"repos_url": "https://api.github.com/users/albertvillanova/repos",
"events_url": "https://api.github.com/users/albertvillanova/events{/privacy}",
"received_events_url": "https://api.github.com/users/albertvillanova/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 | [] | 2025-09-20T07:46:23 | 2025-09-30T17:23:45 | 2025-09-30T17:23:45 | MEMBER | null | null | null | null | ### System Info
- `transformers` version: 4.57.0.dev0
- Platform: Linux-5.15.0-153-generic-x86_64-with-glibc2.31
- Python version: 3.12.9
- Huggingface_hub version: 0.34.4
- Safetensors version: 0.6.2
- Accelerate version: 1.4.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.8.0+cu128 (NA)
- 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?: no
### 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
After investigation of an issue in `trl`, I found a weird behavior of `transformers` config `rope_scaling`: the `config.rope_scaling` (at the root config level) and `config.text_config.rope_scaling` (under `text_config`) might be the same or different dict objects depending on whether we pass `text_config` param to `AutoConfig.from_pretrained`
- if we don't pass `text_config` param, the 2 `rope_scaling` point to the same dict object
- if we pass `text_config` param, the 2 `rope_scaling` are different dict objects
```python
In [1]: from transformers import AutoConfig
In [2]: model_id = "Qwen/Qwen2.5-VL-3B-Instruct"
In [3]: config1 = AutoConfig.from_pretrained(model_id)
In [4]: config1.text_config.rope_scaling
Out[4]: {'type': 'default', 'mrope_section': [16, 24, 24], 'rope_type': 'default'}
In [5]: config1.rope_scaling
Out[5]: {'type': 'default', 'mrope_section': [16, 24, 24], 'rope_type': 'default'}
In [6]: id(config1.text_config.rope_scaling)
Out[6]: 140211029392000
In [7]: id(config1.rope_scaling)
Out[7]: 140211029392000
# Both are the same dict object
In [8]: config2 = AutoConfig.from_pretrained(model_id, text_config={})
In [9]:config2.text_config.rope_scaling
Out[9]: {'type': 'default', 'mrope_section': [16, 24, 24], 'rope_type': 'default'}
In [10]: config2.rope_scaling
Out[10]: {'type': 'default', 'mrope_section': [16, 24, 24], 'rope_type': 'default'}
In [11]: id(config2.text_config.rope_scaling)
Out[11]: 140210801100608
In [12]: id(config2.rope_scaling)
Out[12]: 140211029786688
# Both are different dict objects
```
Is this expected?
We discovered this while investigating why changing (after initialization) the `config.text_config.rope_scaling` will or will not change the `config.rope_scaling` as well. See related comment in `trl` PR:
- https://github.com/huggingface/trl/pull/4101#issuecomment-3305590742
### Expected behavior
- Either they should be the same dict object in any case
- Or they should be different dict objects in any case | {
"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/41020/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/41020/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/41019 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41019/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41019/comments | https://api.github.com/repos/huggingface/transformers/issues/41019/events | https://github.com/huggingface/transformers/pull/41019 | 3,436,869,382 | PR_kwDOCUB6oc6po3U4 | 41,019 | docs(tokenization): Add clarification for train_new_from_iterator | {
"login": "lilin-1",
"id": 177207022,
"node_id": "U_kgDOCo_27g",
"avatar_url": "https://avatars.githubusercontent.com/u/177207022?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/lilin-1",
"html_url": "https://github.com/lilin-1",
"followers_url": "https://api.github.com/users/lilin-1/followers",
"following_url": "https://api.github.com/users/lilin-1/following{/other_user}",
"gists_url": "https://api.github.com/users/lilin-1/gists{/gist_id}",
"starred_url": "https://api.github.com/users/lilin-1/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/lilin-1/subscriptions",
"organizations_url": "https://api.github.com/users/lilin-1/orgs",
"repos_url": "https://api.github.com/users/lilin-1/repos",
"events_url": "https://api.github.com/users/lilin-1/events{/privacy}",
"received_events_url": "https://api.github.com/users/lilin-1/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-20T07:42:39 | 2025-09-22T10:16:37 | 2025-09-22T10:16:37 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41019",
"html_url": "https://github.com/huggingface/transformers/pull/41019",
"diff_url": "https://github.com/huggingface/transformers/pull/41019.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41019.patch",
"merged_at": null
} |
# Add a note to the documentation clarifying that the train_new_from_iterator method in the 🤗 Tokenizers library uses the BPE algorithm to construct the vocabulary, not a true WordPiece implementation.
## What does this PR do?
This PR improves the documentation of the `train_new_from_iterator` method in `tokenization_utils_fast.py` by adding an explicit note to clarify its underlying vocabulary construction logic.
Specifically, a line is added to the method's docstring:
`Note that the train_new_from_iterator() method in the 🤗 Tokenizers library uses BPE to construct the vocabulary, not a true WordPiece implementation.`
### Motivation & Context
The change addresses two key pain points observed in practical use and community contexts:
1. **End-user misunderstanding**: When relying solely on existing documentation, users (including the contributor's team) may expect the method to generate a WordPiece vocabulary (e.g., when working with WordPiece-based tokenizers like `BertTokenizerFast`), but instead receive a BPE-based vocabulary—leading to unintended discrepancies in workflow.
2. **Community confusion**: Multiple research papers reference "using WordPiece via the 🤗 Tokenizers library" without distinguishing that `train_new_from_iterator` relies on BPE under the hood. This note helps align documentation with real-world usage and reduces academic/developmental ambiguity.
The clarification is consistent with details provided in the [Hugging Face LLM Course](https://huggingface.co/course/chapter6/5), which explicitly states that the 🤗 Tokenizers library uses BPE for training (rather than WordPiece) due to incomplete clarity on WordPiece's internal mechanisms.
Fixes # (N/A – this is a proactive documentation improvement, not tied to an existing 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.
*Note: This is a proactive documentation update based on observed user/community confusion, not pre-discussed in an issue/forum.*
- [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?
*Note: No new tests are required, as this change only modifies documentation (no code logic updates).*
## Who can review?
Tag relevant reviewers below (fewer than 3 people):
- @ArthurZucker (relevant to tokenizer-related changes, per template guidance)
- @stevhliu (relevant to documentation improvements, per template guidance) | {
"login": "lilin-1",
"id": 177207022,
"node_id": "U_kgDOCo_27g",
"avatar_url": "https://avatars.githubusercontent.com/u/177207022?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/lilin-1",
"html_url": "https://github.com/lilin-1",
"followers_url": "https://api.github.com/users/lilin-1/followers",
"following_url": "https://api.github.com/users/lilin-1/following{/other_user}",
"gists_url": "https://api.github.com/users/lilin-1/gists{/gist_id}",
"starred_url": "https://api.github.com/users/lilin-1/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/lilin-1/subscriptions",
"organizations_url": "https://api.github.com/users/lilin-1/orgs",
"repos_url": "https://api.github.com/users/lilin-1/repos",
"events_url": "https://api.github.com/users/lilin-1/events{/privacy}",
"received_events_url": "https://api.github.com/users/lilin-1/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41019/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/41019/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41018 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41018/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41018/comments | https://api.github.com/repos/huggingface/transformers/issues/41018/events | https://github.com/huggingface/transformers/pull/41018 | 3,436,760,409 | PR_kwDOCUB6oc6pofik | 41,018 | [testing] Fix `qwen2_audio` | {
"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 | [] | 2025-09-20T06:31:07 | 2025-09-22T10:45:32 | 2025-09-22T10:45:31 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41018",
"html_url": "https://github.com/huggingface/transformers/pull/41018",
"diff_url": "https://github.com/huggingface/transformers/pull/41018.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41018.patch",
"merged_at": "2025-09-22T10:45:31"
} | # What does this PR do?
complete broken qwen2_audio ... | {
"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/41018/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/41018/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41017 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41017/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41017/comments | https://api.github.com/repos/huggingface/transformers/issues/41017/events | https://github.com/huggingface/transformers/pull/41017 | 3,436,408,813 | PR_kwDOCUB6oc6pnTXy | 41,017 | 🚨 [V5] Remove deprecated training arguments | {
"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 | [] | 2025-09-20T01:50:23 | 2025-09-24T10:10:35 | 2025-09-24T10:01:28 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41017",
"html_url": "https://github.com/huggingface/transformers/pull/41017",
"diff_url": "https://github.com/huggingface/transformers/pull/41017.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41017.patch",
"merged_at": "2025-09-24T10:01:28"
} | # What does this PR do?
Remove deprecated training arguments 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
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41017/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/41017/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41016 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41016/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41016/comments | https://api.github.com/repos/huggingface/transformers/issues/41016/events | https://github.com/huggingface/transformers/pull/41016 | 3,435,797,676 | PR_kwDOCUB6oc6plLmF | 41,016 | Add efficient logits_to_keep masking to ZambaForCausalLM for constrained generation | {
"login": "Aniketsy",
"id": 148300120,
"node_id": "U_kgDOCNbhWA",
"avatar_url": "https://avatars.githubusercontent.com/u/148300120?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Aniketsy",
"html_url": "https://github.com/Aniketsy",
"followers_url": "https://api.github.com/users/Aniketsy/followers",
"following_url": "https://api.github.com/users/Aniketsy/following{/other_user}",
"gists_url": "https://api.github.com/users/Aniketsy/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Aniketsy/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Aniketsy/subscriptions",
"organizations_url": "https://api.github.com/users/Aniketsy/orgs",
"repos_url": "https://api.github.com/users/Aniketsy/repos",
"events_url": "https://api.github.com/users/Aniketsy/events{/privacy}",
"received_events_url": "https://api.github.com/users/Aniketsy/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-19T20:24:38 | 2025-10-01T12:08:36 | null | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41016",
"html_url": "https://github.com/huggingface/transformers/pull/41016",
"diff_url": "https://github.com/huggingface/transformers/pull/41016.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41016.patch",
"merged_at": null
} | #40984
This PR updates the `ZambaForCausalLM` forward method to efficiently mask logits using the `logits_to_keep` argument. It supports both top-k and index-based selection, setting all other logits to `-inf` after projection.
Please let me know if my approach or fix needs any improvements . I’m open to feedback and happy to make changes based on suggestions.
Thankyou ! | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41016/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/41016/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41015 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41015/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41015/comments | https://api.github.com/repos/huggingface/transformers/issues/41015/events | https://github.com/huggingface/transformers/pull/41015 | 3,435,672,874 | PR_kwDOCUB6oc6pkwWt | 41,015 | Embed interactive timeline in docs | {
"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 | [] | 2025-09-19T19:35:41 | 2025-09-30T01:36:08 | 2025-09-30T01:36:08 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41015",
"html_url": "https://github.com/huggingface/transformers/pull/41015",
"diff_url": "https://github.com/huggingface/transformers/pull/41015.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41015.patch",
"merged_at": "2025-09-30T01:36:08"
} | # What does this PR do?
As the title says | {
"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/41015/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/41015/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41014 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41014/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41014/comments | https://api.github.com/repos/huggingface/transformers/issues/41014/events | https://github.com/huggingface/transformers/pull/41014 | 3,435,589,815 | PR_kwDOCUB6oc6pkddq | 41,014 | Update after #41007 | {
"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 | [] | 2025-09-19T19:09:38 | 2025-09-19T19:55:48 | 2025-09-19T19:55:46 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41014",
"html_url": "https://github.com/huggingface/transformers/pull/41014",
"diff_url": "https://github.com/huggingface/transformers/pull/41014.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41014.patch",
"merged_at": "2025-09-19T19:55:46"
} | # What does this PR do?
Kudos to @vasqu to point out it's a missing torch.no_grad()
https://github.com/huggingface/transformers/pull/41007#discussion_r2363556135
https://github.com/huggingface/transformers/pull/41007#discussion_r2364096801 | {
"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/41014/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/41014/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41013 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41013/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41013/comments | https://api.github.com/repos/huggingface/transformers/issues/41013/events | https://github.com/huggingface/transformers/issues/41013 | 3,435,486,915 | I_kwDOCUB6oc7MxV7D | 41,013 | Runtime errors when trying to call Trainer() on a model that exceeds GPU vRAM | {
"login": "ag-TJNII",
"id": 69820911,
"node_id": "MDQ6VXNlcjY5ODIwOTEx",
"avatar_url": "https://avatars.githubusercontent.com/u/69820911?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/ag-TJNII",
"html_url": "https://github.com/ag-TJNII",
"followers_url": "https://api.github.com/users/ag-TJNII/followers",
"following_url": "https://api.github.com/users/ag-TJNII/following{/other_user}",
"gists_url": "https://api.github.com/users/ag-TJNII/gists{/gist_id}",
"starred_url": "https://api.github.com/users/ag-TJNII/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/ag-TJNII/subscriptions",
"organizations_url": "https://api.github.com/users/ag-TJNII/orgs",
"repos_url": "https://api.github.com/users/ag-TJNII/repos",
"events_url": "https://api.github.com/users/ag-TJNII/events{/privacy}",
"received_events_url": "https://api.github.com/users/ag-TJNII/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 | [] | 2025-09-19T18:40:51 | 2025-09-29T14:31:43 | 2025-09-29T14:31:43 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.56.1
- Platform: Linux-6.12.43+deb13-amd64-x86_64-with-glibc2.41
- Python version: 3.13.5
- Huggingface_hub version: 0.35.0
- Safetensors version: 0.6.2
- Accelerate version: 1.10.1
- Accelerate config:
- compute_environment: LOCAL_MACHINE
- distributed_type: NO
- mixed_precision: no
- use_cpu: False
- debug: False
- num_processes: 1
- machine_rank: 0
- num_machines: 1
- gpu_ids: all
- rdzv_backend: static
- same_network: True
- main_training_function: main
- enable_cpu_affinity: False
- downcast_bf16: no
- tpu_use_cluster: False
- tpu_use_sudo: False
- tpu_env: []
- DeepSpeed version: not installed
- 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?: <fill in>
- Using GPU in script?: Yes
- GPU type: NVIDIA GeForce RTX 5060 Ti
### Who can help?
@zach-huggingface @SunMarc
### 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
This is based on the quickstart instructions, but is not an officially published script.
Models referenced are `git clone`s of the following models:
- https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
- https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-14B
```
#!/usr/bin/env python3
# Simplification of the example at https://huggingface.co/docs/transformers/quicktour
import os
import argparse
from pathlib import Path
from transformers import AutoModelForCausalLM, TrainingArguments, Trainer
def main():
parser = argparse.ArgumentParser(description="Fine-tune a LLM on HTML files using LoRA.")
parser.add_argument("--model", type=Path, required=True, help="Directory containing pretrained HuggingFace model")
args = parser.parse_args()
model = AutoModelForCausalLM.from_pretrained(args.model, dtype="auto", device_map="auto")
training_args = TrainingArguments(
output_dir="/tmp/spool",
per_device_train_batch_size=1,
num_train_epochs=1,
)
Trainer(
model=model,
args=training_args,
)
print("Success")
if __name__ == "__main__":
main()
```
```
$ ./trainer_example.py --model /host/models/DeepSeek-R1-Distill-Qwen-1.5B
Success
$ ./trainer_example.py --model /host/models/DeepSeek-R1-Distill-Qwen-14B
Loading checkpoint shards: <Snip TUI status bar>
Some parameters are on the meta device because they were offloaded to the cpu.
You shouldn't move a model that is dispatched using accelerate hooks.
Traceback (most recent call last):
File "/host/trainer/build_files/app/./trainer_example.py", line 30, in <module>
main()
~~~~^^
File "/host/trainer/build_files/app/./trainer_example.py", line 22, in main
Trainer(
~~~~~~~^
model=model,
^^^^^^^^^^^^
args=training_args,
^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.13/dist-packages/transformers/utils/deprecation.py", line 172, in wrapped_func
return func(*args, **kwargs)
File "/usr/local/lib/python3.13/dist-packages/transformers/trainer.py", line 620, in __init__
self._move_model_to_device(model, args.device)
~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.13/dist-packages/transformers/trainer.py", line 913, in _move_model_to_device
model = model.to(device)
File "/usr/local/lib/python3.13/dist-packages/accelerate/big_modeling.py", line 462, in wrapper
raise RuntimeError("You can't move a model that has some modules offloaded to cpu or disk.")
RuntimeError: You can't move a model that has some modules offloaded to cpu or disk.
```
Calling via `accelerate launch` did not impact the error.
### Expected behavior
The exception is about an internal action being taken by the library. As a user of the library I'm not sure what action I need to take to resolve this or what config I need to change. The exception should ideally be caught by the code trying to do the move and mitigate it, or bubble it up in a way the user can identify which inputs are causing the error.
From reading the docs offloading should allow this to work. There's still a high probability of user error here, but I can't suss it out from the exceptions I'm getting back, and that's a issue. If this is user error then I think the bug is how the error is presented to the top level code.
Thank you for your work here. I was able to train a small model without needing a deep understanding of what was going on. That's really cool! | {
"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/41013/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/41013/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/41012 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41012/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41012/comments | https://api.github.com/repos/huggingface/transformers/issues/41012/events | https://github.com/huggingface/transformers/pull/41012 | 3,435,366,111 | PR_kwDOCUB6oc6pjwuJ | 41,012 | Fix benchmark runner argument name | {
"login": "ahadnagy",
"id": 21314428,
"node_id": "MDQ6VXNlcjIxMzE0NDI4",
"avatar_url": "https://avatars.githubusercontent.com/u/21314428?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/ahadnagy",
"html_url": "https://github.com/ahadnagy",
"followers_url": "https://api.github.com/users/ahadnagy/followers",
"following_url": "https://api.github.com/users/ahadnagy/following{/other_user}",
"gists_url": "https://api.github.com/users/ahadnagy/gists{/gist_id}",
"starred_url": "https://api.github.com/users/ahadnagy/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/ahadnagy/subscriptions",
"organizations_url": "https://api.github.com/users/ahadnagy/orgs",
"repos_url": "https://api.github.com/users/ahadnagy/repos",
"events_url": "https://api.github.com/users/ahadnagy/events{/privacy}",
"received_events_url": "https://api.github.com/users/ahadnagy/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-19T18:07:54 | 2025-09-20T08:53:57 | 2025-09-20T08:53:56 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41012",
"html_url": "https://github.com/huggingface/transformers/pull/41012",
"diff_url": "https://github.com/huggingface/transformers/pull/41012.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41012.patch",
"merged_at": "2025-09-20T08:53:56"
} | # What does this PR do?
Fix a small mistake in the invocation of the benchmark runner script.
## 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| {
"login": "ahadnagy",
"id": 21314428,
"node_id": "MDQ6VXNlcjIxMzE0NDI4",
"avatar_url": "https://avatars.githubusercontent.com/u/21314428?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/ahadnagy",
"html_url": "https://github.com/ahadnagy",
"followers_url": "https://api.github.com/users/ahadnagy/followers",
"following_url": "https://api.github.com/users/ahadnagy/following{/other_user}",
"gists_url": "https://api.github.com/users/ahadnagy/gists{/gist_id}",
"starred_url": "https://api.github.com/users/ahadnagy/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/ahadnagy/subscriptions",
"organizations_url": "https://api.github.com/users/ahadnagy/orgs",
"repos_url": "https://api.github.com/users/ahadnagy/repos",
"events_url": "https://api.github.com/users/ahadnagy/events{/privacy}",
"received_events_url": "https://api.github.com/users/ahadnagy/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41012/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/41012/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41011 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41011/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41011/comments | https://api.github.com/repos/huggingface/transformers/issues/41011/events | https://github.com/huggingface/transformers/pull/41011 | 3,435,091,978 | PR_kwDOCUB6oc6pi5r7 | 41,011 | Fix Glm4v test | {
"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 | [] | 2025-09-19T16:51:17 | 2025-09-23T10:08:51 | 2025-09-19T16:54:26 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41011",
"html_url": "https://github.com/huggingface/transformers/pull/41011",
"diff_url": "https://github.com/huggingface/transformers/pull/41011.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41011.patch",
"merged_at": "2025-09-19T16:54:26"
} | # What does this PR do?
| {
"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/41011/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/41011/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41010 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41010/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41010/comments | https://api.github.com/repos/huggingface/transformers/issues/41010/events | https://github.com/huggingface/transformers/pull/41010 | 3,434,869,506 | PR_kwDOCUB6oc6piIOu | 41,010 | [`Kernels Attention`] Change fallback logic to error out on explicit kernels request and include FA3 | {
"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 | [] | 2025-09-19T15:40:36 | 2025-09-29T15:11:04 | 2025-09-29T15:10:59 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41010",
"html_url": "https://github.com/huggingface/transformers/pull/41010",
"diff_url": "https://github.com/huggingface/transformers/pull/41010.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41010.patch",
"merged_at": "2025-09-29T15:10:59"
} | As per title. Atm, we just fallback to sdpa/eager even when a kernel was explicitly requested and it caused some hidden states when I ran tests for the bert refactor (not good!).
Imo, just as any other attention type, we should error out when the explicit request cannot be met. Fallbacks are just more confusing in that case.
Additional change:
- Flash attention 3 was originally not considered in this fallback option. Let's do this here as well (?)
Test via `tests/utils/test_modeling_utils.py -k TestAttentionImplementation` (requires different deps, e.g. kernels + no flash, no kernels + no flash, ...)
cc @ArthurZucker @Cyrilvallez | {
"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/41010/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/41010/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41009 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41009/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41009/comments | https://api.github.com/repos/huggingface/transformers/issues/41009/events | https://github.com/huggingface/transformers/pull/41009 | 3,434,856,230 | PR_kwDOCUB6oc6piFWT | 41,009 | Add Lexa-Delta model support | {
"login": "hovhannisians",
"id": 51156797,
"node_id": "MDQ6VXNlcjUxMTU2Nzk3",
"avatar_url": "https://avatars.githubusercontent.com/u/51156797?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/hovhannisians",
"html_url": "https://github.com/hovhannisians",
"followers_url": "https://api.github.com/users/hovhannisians/followers",
"following_url": "https://api.github.com/users/hovhannisians/following{/other_user}",
"gists_url": "https://api.github.com/users/hovhannisians/gists{/gist_id}",
"starred_url": "https://api.github.com/users/hovhannisians/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/hovhannisians/subscriptions",
"organizations_url": "https://api.github.com/users/hovhannisians/orgs",
"repos_url": "https://api.github.com/users/hovhannisians/repos",
"events_url": "https://api.github.com/users/hovhannisians/events{/privacy}",
"received_events_url": "https://api.github.com/users/hovhannisians/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-19T15:36:42 | 2025-09-22T12:03:16 | null | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41009",
"html_url": "https://github.com/huggingface/transformers/pull/41009",
"diff_url": "https://github.com/huggingface/transformers/pull/41009.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41009.patch",
"merged_at": null
} | This PR introduces **Lexa-Delta**, a new large language model developed by **Robi Labs**, as a first-class model in the 🤗 Transformers library.
### What’s included
* **Configuration**: `LexaDeltaConfig` (`model_type="lexa_delta"`)
* **Model**: `LexaDeltaForCausalLM`, inheriting from `GptOssForCausalLM`
* **Tokenizer**: `LexaDeltaTokenizerFast` (inherits from `PreTrainedTokenizerFast`)
* **Docs**: `docs/source/en/model_doc/lexa_delta.md` with usage examples and architecture details
* **Auto-registry**: Added `lexa_delta` to `configuration_auto.py`, `modeling_auto.py`, and `tokenization_auto.py`
### Motivation
Lexa-Delta is designed as an independent model type under Robi Labs, while reusing GPT-OSS architecture internally. This allows developers to:
* Load models with `AutoModelForCausalLM.from_pretrained("RobiLabs/Lexa-Delta")`
* Use `lexa_delta` config/tokenizer classes directly
* Extend future work on the Lexa family of models without relying on GPT-OSS naming
### Example Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("RobiLabs/Lexa-Delta")
tokenizer = AutoTokenizer.from_pretrained("RobiLabs/Lexa-Delta")
inputs = tokenizer("Hello, Lexa!", return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0]))
```
### Next Steps
* Add unit tests under `tests/models/lexa_delta/`
* Community review for merge | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41009/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/41009/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41008 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41008/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41008/comments | https://api.github.com/repos/huggingface/transformers/issues/41008/events | https://github.com/huggingface/transformers/pull/41008 | 3,434,622,924 | PR_kwDOCUB6oc6phRcP | 41,008 | Patch more `unittest.case.TestCase.assertXXX` methods | {
"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 | [] | 2025-09-19T14:31:54 | 2025-09-19T14:41:05 | 2025-09-19T14:38:12 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41008",
"html_url": "https://github.com/huggingface/transformers/pull/41008",
"diff_url": "https://github.com/huggingface/transformers/pull/41008.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41008.patch",
"merged_at": "2025-09-19T14:38:12"
} | # What does this PR do?
Extend #40727
will merge directly | {
"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/41008/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/41008/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41007 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41007/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41007/comments | https://api.github.com/repos/huggingface/transformers/issues/41007/events | https://github.com/huggingface/transformers/pull/41007 | 3,434,567,750 | PR_kwDOCUB6oc6phFP0 | 41,007 | Fix `PhimoeIntegrationTest` | {
"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 | [] | 2025-09-19T14:18:52 | 2025-09-19T18:57:28 | 2025-09-19T16:43:47 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41007",
"html_url": "https://github.com/huggingface/transformers/pull/41007",
"diff_url": "https://github.com/huggingface/transformers/pull/41007.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41007.patch",
"merged_at": "2025-09-19T16:43:47"
} | # What does this PR do?
This test just get the pytest process being killed due to CPU OOM.
```
tests/models/phimoe/test_modeling_phimoe.py::PhimoeIntegrationTest::test_model_phimoe_instruct_logits
-------------------------------- live log call ---------------------------------
WARNING transformers.modeling_rope_utils:modeling_rope_utils.py:418 Unrecognized keys in `rope_scaling` for 'rope_type'='longrope': {'short_mscale', 'long_mscale'}
WARNING transformers.modeling_rope_utils:logging.py:328 This model has set a `original_max_position_embeddings` field, to be used together with `max_position_embeddings` to determine a scaling factor. Please set the `factor` field of `rope_scaling`with this ratio instead -- we recommend the use of this field over `original_max_position_embeddings`, as it is compatible with most model architectures.
Killed
```
Change to
> dtype="auto", device_map="auto"
helps, but `test_phimoe_instruct_with_static_cache` still get GPU OOM even with super short inputs.
So I also change that test to use a dummy model. There is nothing we can do. | {
"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/41007/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/41007/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41006 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41006/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41006/comments | https://api.github.com/repos/huggingface/transformers/issues/41006/events | https://github.com/huggingface/transformers/pull/41006 | 3,434,525,599 | PR_kwDOCUB6oc6pg732 | 41,006 | [ESM] add accepts_loss_kwargs=False to EsmPreTrainedModel | {
"login": "pstjohn",
"id": 2576846,
"node_id": "MDQ6VXNlcjI1NzY4NDY=",
"avatar_url": "https://avatars.githubusercontent.com/u/2576846?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/pstjohn",
"html_url": "https://github.com/pstjohn",
"followers_url": "https://api.github.com/users/pstjohn/followers",
"following_url": "https://api.github.com/users/pstjohn/following{/other_user}",
"gists_url": "https://api.github.com/users/pstjohn/gists{/gist_id}",
"starred_url": "https://api.github.com/users/pstjohn/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/pstjohn/subscriptions",
"organizations_url": "https://api.github.com/users/pstjohn/orgs",
"repos_url": "https://api.github.com/users/pstjohn/repos",
"events_url": "https://api.github.com/users/pstjohn/events{/privacy}",
"received_events_url": "https://api.github.com/users/pstjohn/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-19T14:07:48 | 2025-09-30T12:07:27 | 2025-09-30T12:06:47 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41006",
"html_url": "https://github.com/huggingface/transformers/pull/41006",
"diff_url": "https://github.com/huggingface/transformers/pull/41006.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41006.patch",
"merged_at": "2025-09-30T12:06:47"
} | #40370 added a `**kwargs` to the ESM-2 model's forward pass, which means this model now triggers [these lines](https://github.com/huggingface/transformers/blob/6d2bb1e04db6c8d193549d4b0c99d2182837c0ad/src/transformers/trainer.py#L4127-L4132) in the HF trainer, making losses inconsistent between single and multi-device training.
```python
if (
self.args.average_tokens_across_devices
and (self.model_accepts_loss_kwargs or self.compute_loss_func)
and num_items_in_batch is not None
):
loss *= self.accelerator.num_processes
```
We can avoid this by setting the `accepts_loss_kwargs` flag to False in the model definition.
CC @zucchini-nlp, @Rocketknight1
| {
"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/41006/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/41006/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41005 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41005/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41005/comments | https://api.github.com/repos/huggingface/transformers/issues/41005/events | https://github.com/huggingface/transformers/issues/41005 | 3,434,495,370 | I_kwDOCUB6oc7Mtj2K | 41,005 | Are we have Qwen3VL Official Model Published by Alibaba | {
"login": "Dineshkumar-Anandan-ZS0367",
"id": 105219055,
"node_id": "U_kgDOBkWD7w",
"avatar_url": "https://avatars.githubusercontent.com/u/105219055?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367",
"html_url": "https://github.com/Dineshkumar-Anandan-ZS0367",
"followers_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/followers",
"following_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/following{/other_user}",
"gists_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/subscriptions",
"organizations_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/orgs",
"repos_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/repos",
"events_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/events{/privacy}",
"received_events_url": "https://api.github.com/users/Dineshkumar-Anandan-ZS0367/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [
{
"id": 1843244711,
"node_id": "MDU6TGFiZWwxODQzMjQ0NzEx",
"url": "https://api.github.com/repos/huggingface/transformers/labels/New%20model",
"name": "New model",
"color": "fbca04",
"default": false,
"description": ""
}
] | closed | false | null | [] | null | [] | 2025-09-19T13:59:34 | 2025-09-20T10:00:04 | 2025-09-20T10:00:04 | NONE | null | null | null | null | ### Model description
Reference - https://huggingface.co/docs/transformers/main/en/model_doc/qwen3_vl#transformers.Qwen3VLForConditionalGeneration
If not when can we expect any guess? | {
"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/41005/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/41005/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/41004 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41004/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41004/comments | https://api.github.com/repos/huggingface/transformers/issues/41004/events | https://github.com/huggingface/transformers/pull/41004 | 3,434,442,781 | PR_kwDOCUB6oc6pgpwa | 41,004 | docs: improved RoPE function Docstrings | {
"login": "RyanMullins",
"id": 868555,
"node_id": "MDQ6VXNlcjg2ODU1NQ==",
"avatar_url": "https://avatars.githubusercontent.com/u/868555?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/RyanMullins",
"html_url": "https://github.com/RyanMullins",
"followers_url": "https://api.github.com/users/RyanMullins/followers",
"following_url": "https://api.github.com/users/RyanMullins/following{/other_user}",
"gists_url": "https://api.github.com/users/RyanMullins/gists{/gist_id}",
"starred_url": "https://api.github.com/users/RyanMullins/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/RyanMullins/subscriptions",
"organizations_url": "https://api.github.com/users/RyanMullins/orgs",
"repos_url": "https://api.github.com/users/RyanMullins/repos",
"events_url": "https://api.github.com/users/RyanMullins/events{/privacy}",
"received_events_url": "https://api.github.com/users/RyanMullins/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-19T13:45:30 | 2025-09-22T12:21:16 | 2025-09-22T12:21:16 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41004",
"html_url": "https://github.com/huggingface/transformers/pull/41004",
"diff_url": "https://github.com/huggingface/transformers/pull/41004.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41004.patch",
"merged_at": "2025-09-22T12:21:16"
} | # What does this PR do?
expands the details of the RoPE function docstrings to better convey which fields are possible, which are required, and what they are used for in the functions.
## Before submitting
- [x] 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.
@gante @Cyrilvallez
| {
"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/41004/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/41004/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41003 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41003/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41003/comments | https://api.github.com/repos/huggingface/transformers/issues/41003/events | https://github.com/huggingface/transformers/pull/41003 | 3,433,972,177 | PR_kwDOCUB6oc6pfDSc | 41,003 | [tests] gpt2 + `CausalLMModelTester` | {
"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 | [] | 2025-09-19T11:16:28 | 2025-09-23T17:07:13 | 2025-09-23T17:07:06 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41003",
"html_url": "https://github.com/huggingface/transformers/pull/41003",
"diff_url": "https://github.com/huggingface/transformers/pull/41003.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41003.patch",
"merged_at": "2025-09-23T17:07:06"
} | # What does this PR do?
This PR is an example of a non-super-standard decoder-only model with `CausalLMModelTester`: with a few extra definitions, we can reuse most of the improved testing logic.
👉 If this pattern is approved, I'll open a separate PR for other decoder-only models 🤗 | {
"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/41003/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/41003/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41002 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41002/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41002/comments | https://api.github.com/repos/huggingface/transformers/issues/41002/events | https://github.com/huggingface/transformers/pull/41002 | 3,433,934,408 | PR_kwDOCUB6oc6pe6_b | 41,002 | fix dict like init for ModelOutput | {
"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 | [] | 2025-09-19T11:03:23 | 2025-09-19T14:14:45 | 2025-09-19T14:14:44 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41002",
"html_url": "https://github.com/huggingface/transformers/pull/41002",
"diff_url": "https://github.com/huggingface/transformers/pull/41002.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41002.patch",
"merged_at": "2025-09-19T14:14:44"
} | # What does this PR do?
This PR fixes `ModelOutput` dict-like init as this doesn't work if you don't specify the first attribute. In our case the loss value wasn't cleared.
### Reproducer
```python
from transformers.modeling_outputs import CausalLMOutputWithPast
import torch
lm_output = CausalLMOutputWithPast({"logits":torch.ones(2, 3, 4)})
```
before:
```
CausalLMOutputWithPast(loss={'logits': tensor([[[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]],
[[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]]])}, logits=tensor([[[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]],
[[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]]]), past_key_values=None, hidden_states=None, attentions=None)
```
after
```
CausalLMOutputWithPast(loss=None, logits=tensor([[[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]],
[[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]]]), past_key_values=None, hidden_states=None, attentions=None)
````
| {
"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/41002/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/41002/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/41001 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41001/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41001/comments | https://api.github.com/repos/huggingface/transformers/issues/41001/events | https://github.com/huggingface/transformers/pull/41001 | 3,433,890,418 | PR_kwDOCUB6oc6pexnU | 41,001 | Refactor `tokenization` in `transformers` | {
"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
} | [
{
"id": 9105758243,
"node_id": "LA_kwDOCUB6oc8AAAACHr7YIw",
"url": "https://api.github.com/repos/huggingface/transformers/labels/for_v5?",
"name": "for_v5?",
"color": "35BC94",
"default": false,
"description": ""
}
] | open | false | null | [] | null | [] | 2025-09-19T10:46:46 | 2025-10-13T02:12:57 | null | COLLABORATOR | null | null | true | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41001",
"html_url": "https://github.com/huggingface/transformers/pull/41001",
"diff_url": "https://github.com/huggingface/transformers/pull/41001.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41001.patch",
"merged_at": null
} | # What does this PR do?
The goal of this PR is to get rid of the old and biased "slow vs fast" definition, simplify the code, unbloat the API and have something simple and stable.
We don't want to break stuff, so will try not to.
Related to #40938 | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/41001/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/41001/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/41000 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/41000/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/41000/comments | https://api.github.com/repos/huggingface/transformers/issues/41000/events | https://github.com/huggingface/transformers/pull/41000 | 3,433,867,933 | PR_kwDOCUB6oc6pesyo | 41,000 | Make `EfficientLoFTRModelTest` faster | {
"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 | [] | 2025-09-19T10:40:30 | 2025-09-19T12:51:06 | 2025-09-19T12:51:05 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/41000",
"html_url": "https://github.com/huggingface/transformers/pull/41000",
"diff_url": "https://github.com/huggingface/transformers/pull/41000.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/41000.patch",
"merged_at": "2025-09-19T12:51:05"
} | # What does this PR do?
`4m` -> `6.69s`
Mostly, from some `test_eager_matches_sdpa_inference`, we get `20s` --> `1s` | {
"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/41000/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/41000/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40999 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40999/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40999/comments | https://api.github.com/repos/huggingface/transformers/issues/40999/events | https://github.com/huggingface/transformers/pull/40999 | 3,433,756,073 | PR_kwDOCUB6oc6peUp9 | 40,999 | [docs] rm stray tf/flax autodocs references | {
"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 | [] | 2025-09-19T10:10:19 | 2025-09-19T11:04:21 | 2025-09-19T11:04:12 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40999",
"html_url": "https://github.com/huggingface/transformers/pull/40999",
"diff_url": "https://github.com/huggingface/transformers/pull/40999.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40999.patch",
"merged_at": "2025-09-19T11:04:12"
} | # What does this PR do?
- Removes stray TF/FLAX `[[autodocs]]` references. All languages can now be built (tested locally)
- Removes the `ms`, `te`, and `tr` languages:
- `ms` only contained the index, was not being built in our CI, and was a very incorrect `toc_tree`
- `te` only contained an outdated index and quicktour (many TF references, need a full rewrite)
- `tr` only contained an outdated index (many TF references, need a full rewrite) | {
"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/40999/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/40999/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40998 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40998/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40998/comments | https://api.github.com/repos/huggingface/transformers/issues/40998/events | https://github.com/huggingface/transformers/pull/40998 | 3,433,701,561 | PR_kwDOCUB6oc6peI25 | 40,998 | 🚨 [v5] remove generate output retrocompatibility aliases | {
"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 | [] | 2025-09-19T09:52:09 | 2025-09-19T14:56:46 | 2025-09-19T14:36:13 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40998",
"html_url": "https://github.com/huggingface/transformers/pull/40998",
"diff_url": "https://github.com/huggingface/transformers/pull/40998.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40998.patch",
"merged_at": "2025-09-19T14:36:13"
} | # What does this PR do?
See title :) | {
"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/40998/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/40998/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40997 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40997/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40997/comments | https://api.github.com/repos/huggingface/transformers/issues/40997/events | https://github.com/huggingface/transformers/pull/40997 | 3,433,681,056 | PR_kwDOCUB6oc6peEaH | 40,997 | 🚨 [v5] remove deprecated entry point | {
"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 | [] | 2025-09-19T09:45:15 | 2025-09-19T14:57:02 | 2025-09-19T14:40:27 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40997",
"html_url": "https://github.com/huggingface/transformers/pull/40997",
"diff_url": "https://github.com/huggingface/transformers/pull/40997.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40997.patch",
"merged_at": "2025-09-19T14:40:27"
} | # What does this PR do?
Removes the `transformers-cli` entrypoint, which is deprecated in favor of `transformers` | {
"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/40997/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/40997/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40996 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40996/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40996/comments | https://api.github.com/repos/huggingface/transformers/issues/40996/events | https://github.com/huggingface/transformers/pull/40996 | 3,433,520,164 | PR_kwDOCUB6oc6pdhln | 40,996 | Remove [[autodoc]] refs to TF/Flax objects | {
"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 | [] | 2025-09-19T08:58:18 | 2025-09-19T09:28:36 | 2025-09-19T09:28:34 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40996",
"html_url": "https://github.com/huggingface/transformers/pull/40996",
"diff_url": "https://github.com/huggingface/transformers/pull/40996.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40996.patch",
"merged_at": "2025-09-19T09:28:34"
} | # What does this PR do?
Should bring back to doc builder
cc @gante @vasqu
| {
"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/40996/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/40996/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40995 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40995/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40995/comments | https://api.github.com/repos/huggingface/transformers/issues/40995/events | https://github.com/huggingface/transformers/pull/40995 | 3,433,496,872 | PR_kwDOCUB6oc6pdceA | 40,995 | feat(trainer): Add support for multiple datasets and domain-specific loss | {
"login": "harshaljanjani",
"id": 75426551,
"node_id": "MDQ6VXNlcjc1NDI2NTUx",
"avatar_url": "https://avatars.githubusercontent.com/u/75426551?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/harshaljanjani",
"html_url": "https://github.com/harshaljanjani",
"followers_url": "https://api.github.com/users/harshaljanjani/followers",
"following_url": "https://api.github.com/users/harshaljanjani/following{/other_user}",
"gists_url": "https://api.github.com/users/harshaljanjani/gists{/gist_id}",
"starred_url": "https://api.github.com/users/harshaljanjani/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/harshaljanjani/subscriptions",
"organizations_url": "https://api.github.com/users/harshaljanjani/orgs",
"repos_url": "https://api.github.com/users/harshaljanjani/repos",
"events_url": "https://api.github.com/users/harshaljanjani/events{/privacy}",
"received_events_url": "https://api.github.com/users/harshaljanjani/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-19T08:51:48 | 2025-10-10T16:55:46 | null | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40995",
"html_url": "https://github.com/huggingface/transformers/pull/40995",
"diff_url": "https://github.com/huggingface/transformers/pull/40995.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40995.patch",
"merged_at": null
} | ## What does this PR do?
This has been a long-standing feature request, and I’d love to hear your thoughts on how I approached resolving it. I’ve tested the implementation on the GLUE benchmark and added unit tests for the new logic; happy to make any changes based on feedback!
Just a bit of description: I’ve tried two distinct strategies for handling multiple datasets, with the `multi_dataset_strategy` arg:
1. `sampling` (default): Here, each training step samples a batch from one of the datasets. The sampling can be configured to be proportional to the dataset sizes (`proportional`) or to cycle through them (`round_robin`).
2. `aggregate`: In this mode, each training step fetches one batch from every dataset, computes a domain-specific loss for each, and then aggregates these losses into a single value for backprop.
Best,
Harshal
Might close #30725 and thereby complete #33345.
## Before submitting
* [x] Did you read the [contributor guidelines](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#create-a-pull-request), Pull Request section?
* [x] Did you update the documentation with your changes? See the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation).
* [x] Did you add the necessary tests?
## Who can review?
I’m not sure exactly who to tag here, so I’m pinging @qubvel and @ArthurZucker. | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40995/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/40995/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/40994 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40994/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40994/comments | https://api.github.com/repos/huggingface/transformers/issues/40994/events | https://github.com/huggingface/transformers/pull/40994 | 3,433,468,760 | PR_kwDOCUB6oc6pdWRn | 40,994 | Check model inputs - hidden states | {
"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 | [] | 2025-09-19T08:44:06 | 2025-10-06T09:48:52 | 2025-10-06T09:48:52 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40994",
"html_url": "https://github.com/huggingface/transformers/pull/40994",
"diff_url": "https://github.com/huggingface/transformers/pull/40994.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40994.patch",
"merged_at": "2025-10-06T09:48:52"
} | # What does this PR do?
Fixes https://github.com/huggingface/transformers/issues/40964
In most vision models the `output.hidden_states` are the hiddens right after encoder blocks, i.e. before the last layernorm. Therefore for these models `output.hidden_states != output.last_hidden_state`
Currently `check_model_inputs` assumes that last hidden state is the correct one to return which is true for language models only. This PR adds a kwarg for `check_model_inputs` which decides whether to replace last hidden state or not
TBH, i think the way it is done in LMs is the ultimate correct version and we probably need to "break" vision models. But I can't think of a way to obtain pre-norm last hidden states which are needed for some VLMs | {
"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/40994/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/40994/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40993 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40993/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40993/comments | https://api.github.com/repos/huggingface/transformers/issues/40993/events | https://github.com/huggingface/transformers/issues/40993 | 3,433,418,635 | I_kwDOCUB6oc7Mpc-L | 40,993 | HfArgumentParser cannot parse TRL Config | {
"login": "caoyang-sufe",
"id": 135610001,
"node_id": "U_kgDOCBU-kQ",
"avatar_url": "https://avatars.githubusercontent.com/u/135610001?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/caoyang-sufe",
"html_url": "https://github.com/caoyang-sufe",
"followers_url": "https://api.github.com/users/caoyang-sufe/followers",
"following_url": "https://api.github.com/users/caoyang-sufe/following{/other_user}",
"gists_url": "https://api.github.com/users/caoyang-sufe/gists{/gist_id}",
"starred_url": "https://api.github.com/users/caoyang-sufe/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/caoyang-sufe/subscriptions",
"organizations_url": "https://api.github.com/users/caoyang-sufe/orgs",
"repos_url": "https://api.github.com/users/caoyang-sufe/repos",
"events_url": "https://api.github.com/users/caoyang-sufe/events{/privacy}",
"received_events_url": "https://api.github.com/users/caoyang-sufe/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 | [] | 2025-09-19T08:29:48 | 2025-09-19T09:06:20 | 2025-09-19T09:06:20 | NONE | null | null | null | null | ### System Info
transformers==4.56.1
trl==0.17.0
I used to apply code below
```python
from transformers import HfArgumentParser
from trl import (
ScriptArguments, ModelConfig, SFTConfig
)
parser = HfArgumentParser((ScriptArguments, SFTConfig, ModelConfig))
script_arguments, trainer_config, model_config = parser.parse_args_into_dataclasses()
```
to parse training args, but after updating transformers to 4.56, it does not work:
```
Traceback (most recent call last):
File "D:\mytest.py", line 5, in <module>
parser = HfArgumentParser((ScriptArguments, SFTConfig, ModelConfig))
File "E:\Anaconda3\envs\myopenai\lib\site-packages\transformers\hf_argparser.py", line 143, in __init__
self._add_dataclass_arguments(dtype)
File "E:\Anaconda3\envs\myopenai\lib\site-packages\transformers\hf_argparser.py", line 260, in _add_dataclass_arguments
raise RuntimeError(
RuntimeError: Type resolution failed for <class 'trl.trainer.sft_config.SFTConfig'>. Try declaring the class in global scope or removing line of `from __future__ import annotations` which opts in Postponed Evaluation of Annotations (PEP 563)
```
How to fix it?
### Who can help?
_No response_
### 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
Run
```python
from transformers import HfArgumentParser
from trl import (
ScriptArguments, ModelConfig, SFTConfig
)
parser = HfArgumentParser((ScriptArguments, SFTConfig, ModelConfig))
script_arguments, trainer_config, model_config = parser.parse_args_into_dataclasses()
```
### Expected behavior
It should be work | {
"login": "caoyang-sufe",
"id": 135610001,
"node_id": "U_kgDOCBU-kQ",
"avatar_url": "https://avatars.githubusercontent.com/u/135610001?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/caoyang-sufe",
"html_url": "https://github.com/caoyang-sufe",
"followers_url": "https://api.github.com/users/caoyang-sufe/followers",
"following_url": "https://api.github.com/users/caoyang-sufe/following{/other_user}",
"gists_url": "https://api.github.com/users/caoyang-sufe/gists{/gist_id}",
"starred_url": "https://api.github.com/users/caoyang-sufe/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/caoyang-sufe/subscriptions",
"organizations_url": "https://api.github.com/users/caoyang-sufe/orgs",
"repos_url": "https://api.github.com/users/caoyang-sufe/repos",
"events_url": "https://api.github.com/users/caoyang-sufe/events{/privacy}",
"received_events_url": "https://api.github.com/users/caoyang-sufe/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40993/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/40993/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/40992 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40992/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40992/comments | https://api.github.com/repos/huggingface/transformers/issues/40992/events | https://github.com/huggingface/transformers/pull/40992 | 3,432,781,709 | PR_kwDOCUB6oc6pa_R7 | 40,992 | [testing] test `num_hidden_layers` being small in model tester | {
"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 | [] | 2025-09-19T04:56:30 | 2025-09-19T09:45:09 | 2025-09-19T09:45:07 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40992",
"html_url": "https://github.com/huggingface/transformers/pull/40992",
"diff_url": "https://github.com/huggingface/transformers/pull/40992.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40992.patch",
"merged_at": "2025-09-19T09:45:07"
} | # What does this PR do?
Follow up of #40982
Let's still try to check if the number of layers is small.
I don't have enough to dive into all exceptional cases, but for new models, let's try to be careful. | {
"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/40992/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/40992/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40991 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40991/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40991/comments | https://api.github.com/repos/huggingface/transformers/issues/40991/events | https://github.com/huggingface/transformers/pull/40991 | 3,432,531,945 | PR_kwDOCUB6oc6paIBC | 40,991 | Add LWDetr model | {
"login": "sbucaille",
"id": 24275548,
"node_id": "MDQ6VXNlcjI0Mjc1NTQ4",
"avatar_url": "https://avatars.githubusercontent.com/u/24275548?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/sbucaille",
"html_url": "https://github.com/sbucaille",
"followers_url": "https://api.github.com/users/sbucaille/followers",
"following_url": "https://api.github.com/users/sbucaille/following{/other_user}",
"gists_url": "https://api.github.com/users/sbucaille/gists{/gist_id}",
"starred_url": "https://api.github.com/users/sbucaille/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/sbucaille/subscriptions",
"organizations_url": "https://api.github.com/users/sbucaille/orgs",
"repos_url": "https://api.github.com/users/sbucaille/repos",
"events_url": "https://api.github.com/users/sbucaille/events{/privacy}",
"received_events_url": "https://api.github.com/users/sbucaille/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-19T02:54:30 | 2025-10-29T16:47:46 | null | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40991",
"html_url": "https://github.com/huggingface/transformers/pull/40991",
"diff_url": "https://github.com/huggingface/transformers/pull/40991.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40991.patch",
"merged_at": null
} | # What does this PR do?
Adds LWDetr model.
In #36895 I started working on adding RFDetr, but after putting some work I realized that it relies a LOT on LWDetr.
Adding RFDetr will essentially replace the ViT encoder by Dino, so the biggest part of the work is on the implementation of LWDetr, which could also be a good alternative for people to use for their use cases.
## Who can review?
Still work in progress but since @yonigozlan asked for an update here it is.
All the inference code is implemented. A lot of refactoring/renaming is still needed and I'm writing the tests to be able to do that safely. In the meantime you can check the code and let me know if you have comments.
@qubvel | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40991/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/40991/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/40990 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40990/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40990/comments | https://api.github.com/repos/huggingface/transformers/issues/40990/events | https://github.com/huggingface/transformers/issues/40990 | 3,432,292,570 | I_kwDOCUB6oc7MlKDa | 40,990 | Extremely high perplexity on openai/gpt-oss-20b with WikiText-2 (raw) | {
"login": "kuantuna",
"id": 66808459,
"node_id": "MDQ6VXNlcjY2ODA4NDU5",
"avatar_url": "https://avatars.githubusercontent.com/u/66808459?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/kuantuna",
"html_url": "https://github.com/kuantuna",
"followers_url": "https://api.github.com/users/kuantuna/followers",
"following_url": "https://api.github.com/users/kuantuna/following{/other_user}",
"gists_url": "https://api.github.com/users/kuantuna/gists{/gist_id}",
"starred_url": "https://api.github.com/users/kuantuna/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/kuantuna/subscriptions",
"organizations_url": "https://api.github.com/users/kuantuna/orgs",
"repos_url": "https://api.github.com/users/kuantuna/repos",
"events_url": "https://api.github.com/users/kuantuna/events{/privacy}",
"received_events_url": "https://api.github.com/users/kuantuna/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 | [] | 2025-09-19T00:40:14 | 2025-09-27T00:28:00 | 2025-09-22T10:07:32 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.56.1
- Platform: Linux-6.5.0-1025-gcp-x86_64-with-glibc2.35
- Python version: 3.11.10
- Huggingface_hub version: 0.35.0
- Safetensors version: 0.6.2
- Accelerate version: 1.10.1
- Accelerate config: not found
- DeepSpeed version: 0.17.3+cu126.pt27.v0.17.3.recogni2
- PyTorch version (accelerator?): 2.7.1+cu126 (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?: no
- Using GPU in script?: yes
- GPU type: NVIDIA A100-SXM4-40GB
### Who can help?
@ArthurZucker @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
Script:
```python
#!/usr/bin/env python
import math
import torch
from datasets import load_dataset
from transformers import AutoModelForCausalLM, AutoTokenizer
# Config
MODEL_NAME = "openai/gpt-oss-20b"
SPLIT = "test" # WikiText-2 (raw) test split
CONTEXT_LENGTH = 2048 # evaluation window size
DTYPE = torch.bfloat16
DEVICE_MAP = "auto"
def main():
# Load tokenizer & model
tok = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=DTYPE, device_map=DEVICE_MAP).eval()
# Load dataset and build one long token stream (no special tokens)
ds = load_dataset("wikitext", "wikitext-2-raw-v1", split=SPLIT)
encs = tok([row["text"] for row in ds], add_special_tokens=False)
flat_ids = [tid for seq in encs["input_ids"] for tid in seq]
ids = torch.tensor(flat_ids, dtype=torch.long)
# Keep first 10% of tokens
n_keep = max(1, int(0.10 * ids.numel()))
ids = ids[:n_keep]
# Keep only full CONTEXT_LENGTH windows
n_windows = ids.numel() // CONTEXT_LENGTH
if n_windows == 0:
raise ValueError(f"Not enough tokens ({ids.numel()}) for a single {CONTEXT_LENGTH}-token window.")
ids = ids[: n_windows * CONTEXT_LENGTH].view(n_windows, CONTEXT_LENGTH)
# Forward passes
total_nll, total_tokens = 0.0, 0
with torch.no_grad():
for i in range(n_windows):
x = ids[i : i + 1].to(model.device) # [1, L]
out = model(input_ids=x, labels=x) # HF shifts labels internally
contrib = x.size(1) - 1 # L-1 positions contribute
total_nll += out.loss.item() * contrib # sum NLL
total_tokens += contrib
avg_nll = total_nll / total_tokens
ppl = math.exp(avg_nll)
# Detailed prints
print("\n=== Repro Config ===")
print(f"model_name: {MODEL_NAME}")
print(f"split: {SPLIT}")
print(f"context_length: {CONTEXT_LENGTH}")
print(f"dtype: {DTYPE}")
print(f"device_map: {DEVICE_MAP}")
print(f"tokens_total: {ids.numel()}")
print(f"num_segments: {n_windows}")
print(f"bos/eos/pad: {tok.bos_token}/{tok.eos_token}/{tok.pad_token}")
print("\n=== Results ===")
print(f"tokens_scored: {total_tokens}")
print(f"avg_nll: {avg_nll:.6f}")
print(f"perplexity: {ppl:.3f}\n")
if __name__ == "__main__":
main()
```
Output:
```
=== Repro Config ===
model_name: openai/gpt-oss-20b
split: test
context_length: 2048
dtype: torch.bfloat16
device_map: auto
tokens_total: 28672
num_segments: 14
bos/eos/pad: <|startoftext|>/<|return|>/<|endoftext|>
=== Results ===
tokens_scored: 28658
avg_nll: 5.977535
perplexity: 394.467
```
### Expected behavior
When evaluating `openai/gpt-oss-20b` on the WikiText-2 (raw) test split with a standard perplexity script, the reported perplexity is extremely high (~394). This is surprising, as a 20B parameter GPT-class model should normally achieve much lower perplexity on this benchmark.
Clarification would be helpful to determine whether this behavior indicates a bug in the Transformers integration or if GPT-OSS models are not intended to be directly evaluated as causal LMs without special formatting.
Note: The model card mentions a “harmony” chat template for usage, but it is unclear whether special formatting is required when performing perplexity evaluation on a corpus like WikiText. | {
"login": "kuantuna",
"id": 66808459,
"node_id": "MDQ6VXNlcjY2ODA4NDU5",
"avatar_url": "https://avatars.githubusercontent.com/u/66808459?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/kuantuna",
"html_url": "https://github.com/kuantuna",
"followers_url": "https://api.github.com/users/kuantuna/followers",
"following_url": "https://api.github.com/users/kuantuna/following{/other_user}",
"gists_url": "https://api.github.com/users/kuantuna/gists{/gist_id}",
"starred_url": "https://api.github.com/users/kuantuna/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/kuantuna/subscriptions",
"organizations_url": "https://api.github.com/users/kuantuna/orgs",
"repos_url": "https://api.github.com/users/kuantuna/repos",
"events_url": "https://api.github.com/users/kuantuna/events{/privacy}",
"received_events_url": "https://api.github.com/users/kuantuna/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40990/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/40990/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/40989 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40989/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40989/comments | https://api.github.com/repos/huggingface/transformers/issues/40989/events | https://github.com/huggingface/transformers/pull/40989 | 3,431,987,006 | PR_kwDOCUB6oc6pYTZ0 | 40,989 | enable flex attention ut cases 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 | [] | 2025-09-18T21:14:03 | 2025-09-29T15:51:16 | 2025-09-29T14:30:49 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40989",
"html_url": "https://github.com/huggingface/transformers/pull/40989",
"diff_url": "https://github.com/huggingface/transformers/pull/40989.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40989.patch",
"merged_at": "2025-09-29T14:30:49"
} | enable flex_attention test cases on XPU, it's supported by XPU torch now.
@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/40989/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/40989/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40988 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40988/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40988/comments | https://api.github.com/repos/huggingface/transformers/issues/40988/events | https://github.com/huggingface/transformers/issues/40988 | 3,431,951,512 | I_kwDOCUB6oc7Mj2yY | 40,988 | Qwen Model does not free CPU Memory when moving it to GPU | {
"login": "cehongwang",
"id": 123616592,
"node_id": "U_kgDOB149UA",
"avatar_url": "https://avatars.githubusercontent.com/u/123616592?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/cehongwang",
"html_url": "https://github.com/cehongwang",
"followers_url": "https://api.github.com/users/cehongwang/followers",
"following_url": "https://api.github.com/users/cehongwang/following{/other_user}",
"gists_url": "https://api.github.com/users/cehongwang/gists{/gist_id}",
"starred_url": "https://api.github.com/users/cehongwang/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/cehongwang/subscriptions",
"organizations_url": "https://api.github.com/users/cehongwang/orgs",
"repos_url": "https://api.github.com/users/cehongwang/repos",
"events_url": "https://api.github.com/users/cehongwang/events{/privacy}",
"received_events_url": "https://api.github.com/users/cehongwang/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 | [] | 2025-09-18T20:57:33 | 2025-10-27T08:03:08 | 2025-10-27T08:03:08 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.51.3
- Platform: Linux-5.15.0-88-generic-x86_64-with-glibc2.39
- Python version: 3.12.3
- Huggingface_hub version: 0.34.5
- Safetensors version: 0.6.2
- Accelerate version: 1.3.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (GPU?): 2.8.0+cu128 (True)
- 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?: No
- Using GPU in script?: No
- GPU type: NVIDIA H200
### Who can help?
CPU usage is typically expected to be released after doing model.cuda(). This is not the case of Qwen1.5 MoE or Qwen 3 MoE.
### 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 this script
```py
from transformers import AutoModelForCausalLM
import torch
import psutil
process = psutil.Process()
print("Before loading model")
print(process.memory_info().rss / 1024 / 1024, "MB")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen1.5-MoE-A2.7B-Chat", torch_dtype=torch.float16).eval()
print("After loading model")
print(process.memory_info().rss / 1024 / 1024, "MB")
model.cuda()
print("After loading model to GPU")
print(process.memory_info().rss / 1024 / 1024, "MB")
```
### Expected behavior
The CPU memory should be released | {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40988/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/40988/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/40987 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40987/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40987/comments | https://api.github.com/repos/huggingface/transformers/issues/40987/events | https://github.com/huggingface/transformers/pull/40987 | 3,431,845,039 | PR_kwDOCUB6oc6pX0g1 | 40,987 | New blueberry model and tokenizer | {
"login": "dustinwloring1988",
"id": 21135165,
"node_id": "MDQ6VXNlcjIxMTM1MTY1",
"avatar_url": "https://avatars.githubusercontent.com/u/21135165?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/dustinwloring1988",
"html_url": "https://github.com/dustinwloring1988",
"followers_url": "https://api.github.com/users/dustinwloring1988/followers",
"following_url": "https://api.github.com/users/dustinwloring1988/following{/other_user}",
"gists_url": "https://api.github.com/users/dustinwloring1988/gists{/gist_id}",
"starred_url": "https://api.github.com/users/dustinwloring1988/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/dustinwloring1988/subscriptions",
"organizations_url": "https://api.github.com/users/dustinwloring1988/orgs",
"repos_url": "https://api.github.com/users/dustinwloring1988/repos",
"events_url": "https://api.github.com/users/dustinwloring1988/events{/privacy}",
"received_events_url": "https://api.github.com/users/dustinwloring1988/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T20:17:10 | 2025-09-18T20:55:06 | 2025-09-18T20:55:06 | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40987",
"html_url": "https://github.com/huggingface/transformers/pull/40987",
"diff_url": "https://github.com/huggingface/transformers/pull/40987.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40987.patch",
"merged_at": null
} | # What does this PR do?
This PR introduces the foundational elements for the new `Blueberry` model. It sets up the basic directory structure (`src/transformers/models/blueberry`) and implements the `BlueberryConfig` class.
Key features of the `BlueberryConfig` include:
- Definition of a small, non-MoE decoder-only transformer architecture.
- Support for a hybrid NoPE/RoPE attention mechanism, controlled by the `layer_types` parameter, allowing for alternating sliding window and full attention layers.
- Integration of YaRN RoPE scaling, configured via the `rope_scaling` dictionary.
- Custom default parameters for various model components as specified in the project objective.
This PR lays the groundwork for the full model implementation (`modeling_blueberry.py`) and tokenizer (`tokenization_blueberry.py`) which will follow.
## 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.
@ArthurZucker @gante
| {
"login": "dustinwloring1988",
"id": 21135165,
"node_id": "MDQ6VXNlcjIxMTM1MTY1",
"avatar_url": "https://avatars.githubusercontent.com/u/21135165?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/dustinwloring1988",
"html_url": "https://github.com/dustinwloring1988",
"followers_url": "https://api.github.com/users/dustinwloring1988/followers",
"following_url": "https://api.github.com/users/dustinwloring1988/following{/other_user}",
"gists_url": "https://api.github.com/users/dustinwloring1988/gists{/gist_id}",
"starred_url": "https://api.github.com/users/dustinwloring1988/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/dustinwloring1988/subscriptions",
"organizations_url": "https://api.github.com/users/dustinwloring1988/orgs",
"repos_url": "https://api.github.com/users/dustinwloring1988/repos",
"events_url": "https://api.github.com/users/dustinwloring1988/events{/privacy}",
"received_events_url": "https://api.github.com/users/dustinwloring1988/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40987/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/40987/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40986 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40986/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40986/comments | https://api.github.com/repos/huggingface/transformers/issues/40986/events | https://github.com/huggingface/transformers/issues/40986 | 3,431,586,930 | I_kwDOCUB6oc7Midxy | 40,986 | Rewrite audio feature processors to accept torch tensors | {
"login": "jackzhxng",
"id": 32371937,
"node_id": "MDQ6VXNlcjMyMzcxOTM3",
"avatar_url": "https://avatars.githubusercontent.com/u/32371937?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/jackzhxng",
"html_url": "https://github.com/jackzhxng",
"followers_url": "https://api.github.com/users/jackzhxng/followers",
"following_url": "https://api.github.com/users/jackzhxng/following{/other_user}",
"gists_url": "https://api.github.com/users/jackzhxng/gists{/gist_id}",
"starred_url": "https://api.github.com/users/jackzhxng/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/jackzhxng/subscriptions",
"organizations_url": "https://api.github.com/users/jackzhxng/orgs",
"repos_url": "https://api.github.com/users/jackzhxng/repos",
"events_url": "https://api.github.com/users/jackzhxng/events{/privacy}",
"received_events_url": "https://api.github.com/users/jackzhxng/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"
},
{
"id": 6470596964,
"node_id": "LA_kwDOCUB6oc8AAAABga15ZA",
"url": "https://api.github.com/repos/huggingface/transformers/labels/Audio",
"name": "Audio",
"color": "760453",
"default": false,
"description": ""
}
] | open | false | null | [] | null | [] | 2025-09-18T18:57:40 | 2025-09-19T16:58:45 | null | CONTRIBUTOR | null | null | null | null | ### Feature request
We would like to be able to torch.export the audio feature processors such as:
- [GraniteSpeechFeatureExtractor](https://github.com/huggingface/transformers/blob/5f6e278a5177d8b85945a2cdb6b776dacee34914/src/transformers/models/granite_speech/feature_extraction_granite_speech.py#L38)
- [WhisperFeatureExtractor](https://github.com/huggingface/transformers/blob/5f6e278a5177d8b85945a2cdb6b776dacee34914/src/transformers/models/whisper/feature_extraction_whisper.py#L36) used for Voxtral and Whisper
@ArthurZucker @zucchini-nlp @Cyrilvallez @larryliu0820
### Motivation
This allows us to torch.export the whole E2E needed to generate output for Transformers audio models. Then in ExecuTorch, we can run everything without needing to write any custom C++ code for audio preprocessing.
### Your contribution
Can help with code / provide context (?) | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40986/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/40986/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/40985 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40985/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40985/comments | https://api.github.com/repos/huggingface/transformers/issues/40985/events | https://github.com/huggingface/transformers/pull/40985 | 3,431,417,311 | PR_kwDOCUB6oc6pWV-D | 40,985 | Cursor/create blueberry language model and tokenizer b996 | {
"login": "dustinwloring1988",
"id": 21135165,
"node_id": "MDQ6VXNlcjIxMTM1MTY1",
"avatar_url": "https://avatars.githubusercontent.com/u/21135165?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/dustinwloring1988",
"html_url": "https://github.com/dustinwloring1988",
"followers_url": "https://api.github.com/users/dustinwloring1988/followers",
"following_url": "https://api.github.com/users/dustinwloring1988/following{/other_user}",
"gists_url": "https://api.github.com/users/dustinwloring1988/gists{/gist_id}",
"starred_url": "https://api.github.com/users/dustinwloring1988/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/dustinwloring1988/subscriptions",
"organizations_url": "https://api.github.com/users/dustinwloring1988/orgs",
"repos_url": "https://api.github.com/users/dustinwloring1988/repos",
"events_url": "https://api.github.com/users/dustinwloring1988/events{/privacy}",
"received_events_url": "https://api.github.com/users/dustinwloring1988/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T18:01:16 | 2025-09-18T18:04:54 | 2025-09-18T18:04:54 | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40985",
"html_url": "https://github.com/huggingface/transformers/pull/40985",
"diff_url": "https://github.com/huggingface/transformers/pull/40985.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40985.patch",
"merged_at": null
} | # What does this PR do?
This PR introduces the new "Blueberry" model to the `transformers` library. Blueberry is a small, decoder-only language model featuring a novel hybrid attention mechanism that alternates between RoPE with YaRN scaling (sliding window) and NoPE (full attention) layers. It also includes a custom GPT-2-like tokenizer with a 100K vocabulary and implements the Harmony Chat Format via a Jinja2 template.
The PR adds:
* `src/transformers/models/blueberry/`: Model configuration, core modeling, and tokenizer implementations.
* `tests/models/blueberry/`: Unit tests for the model's architecture and tokenizer functionality, including the chat template.
* `docs/source/en/model_doc/blueberry.md`: Comprehensive documentation for the model.
## 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. (This was a direct task, assuming approval)
- [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?
@ArthurZucker
| {
"login": "dustinwloring1988",
"id": 21135165,
"node_id": "MDQ6VXNlcjIxMTM1MTY1",
"avatar_url": "https://avatars.githubusercontent.com/u/21135165?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/dustinwloring1988",
"html_url": "https://github.com/dustinwloring1988",
"followers_url": "https://api.github.com/users/dustinwloring1988/followers",
"following_url": "https://api.github.com/users/dustinwloring1988/following{/other_user}",
"gists_url": "https://api.github.com/users/dustinwloring1988/gists{/gist_id}",
"starred_url": "https://api.github.com/users/dustinwloring1988/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/dustinwloring1988/subscriptions",
"organizations_url": "https://api.github.com/users/dustinwloring1988/orgs",
"repos_url": "https://api.github.com/users/dustinwloring1988/repos",
"events_url": "https://api.github.com/users/dustinwloring1988/events{/privacy}",
"received_events_url": "https://api.github.com/users/dustinwloring1988/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40985/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/40985/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40984 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40984/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40984/comments | https://api.github.com/repos/huggingface/transformers/issues/40984/events | https://github.com/huggingface/transformers/issues/40984 | 3,431,415,151 | I_kwDOCUB6oc7Mhz1v | 40,984 | Adding `logits_to_keep` to older models | {
"login": "philiproeleveld",
"id": 25742753,
"node_id": "MDQ6VXNlcjI1NzQyNzUz",
"avatar_url": "https://avatars.githubusercontent.com/u/25742753?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/philiproeleveld",
"html_url": "https://github.com/philiproeleveld",
"followers_url": "https://api.github.com/users/philiproeleveld/followers",
"following_url": "https://api.github.com/users/philiproeleveld/following{/other_user}",
"gists_url": "https://api.github.com/users/philiproeleveld/gists{/gist_id}",
"starred_url": "https://api.github.com/users/philiproeleveld/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/philiproeleveld/subscriptions",
"organizations_url": "https://api.github.com/users/philiproeleveld/orgs",
"repos_url": "https://api.github.com/users/philiproeleveld/repos",
"events_url": "https://api.github.com/users/philiproeleveld/events{/privacy}",
"received_events_url": "https://api.github.com/users/philiproeleveld/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [
{
"id": 1990918270,
"node_id": "MDU6TGFiZWwxOTkwOTE4Mjcw",
"url": "https://api.github.com/repos/huggingface/transformers/labels/Good%20First%20Issue",
"name": "Good First Issue",
"color": "bbf794",
"default": false,
"description": ""
}
] | closed | false | null | [] | null | [] | 2025-09-18T18:00:30 | 2025-10-15T09:56:03 | 2025-10-15T09:56:03 | CONTRIBUTOR | null | null | null | null | Hi, while working with BioGPT I came across #39016 to not compute logits of the entire sequence when it's not needed. I was wondering if you'd be open to a PR applying the same improvement to BioGPT and, while I'm at it, any other `GenerationMixin` models that would benefit from this. I understand some of these models might be a bit obsolete and some could benefit from a refactor to modular instead, but I figure while they're there it's worth doing. | {
"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/40984/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/40984/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/40983 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40983/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40983/comments | https://api.github.com/repos/huggingface/transformers/issues/40983/events | https://github.com/huggingface/transformers/issues/40983 | 3,431,409,709 | I_kwDOCUB6oc7Mhygt | 40,983 | [`RFC v5`] Encoder-decoder support for Bert-related models and GPT2 | {
"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
} | [
{
"id": 9105758243,
"node_id": "LA_kwDOCUB6oc8AAAACHr7YIw",
"url": "https://api.github.com/repos/huggingface/transformers/labels/for_v5?",
"name": "for_v5?",
"color": "35BC94",
"default": false,
"description": ""
}
] | open | false | null | [] | null | [] | 2025-09-18T17:58:38 | 2025-10-27T09:18:12 | null | CONTRIBUTOR | null | null | null | null | Background:
- Encoder-decoder models have been introduced via incorporating already trained encoder- and decoder-only models. See https://huggingface.co/papers/1908.08345
- It relies on the fact that these models can be customized on the fly.
Issues/Motivation:
- Transformers tries to actively keep a clean code of modeling files so that users can see the direct architecture without much bloat.
- With the aforementioned feature, a lot of bloat code was introduced with multiple branches which makes maintenance harder.
- These models are now overloaded and dont have the essentials only anymore.
Proposal:
- There are 3 paths we can go from here
- 1. Keep everything as is
- 2. Split the modeling files properly to have a Bert encoder-only, decoder-only etc
- 3. Drop support for encoder-decoder models composed of different models
Nothing is set yet but I'm interested in the community's usage/view on these related features, e.g. see https://huggingface.co/docs/transformers/model_doc/encoder-decoder - suggestions are greatly appreciated!
cc @ArthurZucker @Cyrilvallez @LysandreJik | {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40983/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/40983/timeline | null | reopened | {
"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/40982 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40982/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40982/comments | https://api.github.com/repos/huggingface/transformers/issues/40982/events | https://github.com/huggingface/transformers/pull/40982 | 3,431,176,173 | PR_kwDOCUB6oc6pVg6k | 40,982 | Remove `set_model_tester_for_less_flaky_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 | [] | 2025-09-18T16:46:14 | 2025-09-18T16:56:12 | 2025-09-18T16:56:10 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40982",
"html_url": "https://github.com/huggingface/transformers/pull/40982",
"diff_url": "https://github.com/huggingface/transformers/pull/40982.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40982.patch",
"merged_at": "2025-09-18T16:56:10"
} | # What does this PR do?
As per the title. It's a very bad idea to dynamically change the whole `tester`, and brings a lot of issues for models with `layer_types`. @gante removed most of its usage already, just cleaning up what's left! | {
"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/40982/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/40982/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40981 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40981/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40981/comments | https://api.github.com/repos/huggingface/transformers/issues/40981/events | https://github.com/huggingface/transformers/pull/40981 | 3,430,821,120 | PR_kwDOCUB6oc6pUS2p | 40,981 | Track the CI (model) jobs that don't produce test output files (process being killed etc.) | {
"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 | [] | 2025-09-18T15:09:36 | 2025-09-18T16:27:29 | 2025-09-18T16:27:27 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40981",
"html_url": "https://github.com/huggingface/transformers/pull/40981",
"diff_url": "https://github.com/huggingface/transformers/pull/40981.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40981.patch",
"merged_at": "2025-09-18T16:27:27"
} | # What does this PR do?
For a few CI jobs, the pytest process is killed, not producing test output files. In this case, currently, this is not tracked, so we somehow think those jobs have no failing tests.
This PR try to track such situation, and show at least on slack:
> 🚨 There were {self.n_jobs_errored_out} jobs errored out (not producing test output files). 🚨
When we see this on Slack, we can check the artifact `model_results.json` by searching `["error"]` to find which jobs have this issue, click the `["job_link"]` to get to the job run page for more details.
Might be better to extend this work to other jobs (deepspeed, pipeline), but let's see how it works with model jobs first.
<img width="906" height="337" alt="Screenshot 2025-09-18 182213" src="https://github.com/user-attachments/assets/aaa96116-36af-46b1-864f-9ff2cf1b5273" /> | {
"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/40981/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/40981/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40980 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40980/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40980/comments | https://api.github.com/repos/huggingface/transformers/issues/40980/events | https://github.com/huggingface/transformers/pull/40980 | 3,430,662,187 | PR_kwDOCUB6oc6pTvoz | 40,980 | [tests] update `test_left_padding_compatibility` (and minimize overwrites) | {
"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 | [] | 2025-09-18T14:33:05 | 2025-09-19T14:56:56 | 2025-09-19T14:36:26 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40980",
"html_url": "https://github.com/huggingface/transformers/pull/40980",
"diff_url": "https://github.com/huggingface/transformers/pull/40980.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40980.patch",
"merged_at": "2025-09-19T14:36:26"
} | # What does this PR do?
This PR updates `test_left_padding_compatibility` so that:
1. We now test left-padding support against all model inputs, as opposed to text-only inputs. This makes the test stricter: we now test that the model behaves well with left-padded text inputs, together with other (often optional) model inputs. Example: image inputs + padded text inputs may need a corresponding padded cross-attention mask 🔒
2. No true overwrites exist in the library. We can now define custom inputs to the test, to set model-specific inputs that may need to be padded (e.g. see example above) 🧘
(requirement for #40917 -- the changes made there need the updated test for a happy CI)
| {
"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/40980/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/40980/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40979 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40979/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40979/comments | https://api.github.com/repos/huggingface/transformers/issues/40979/events | https://github.com/huggingface/transformers/pull/40979 | 3,430,529,987 | PR_kwDOCUB6oc6pTSYh | 40,979 | Remove require_torch_bf16_gpu | {
"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 | [] | 2025-09-18T14:01:44 | 2025-10-17T10:50:33 | 2025-10-17T10:35:19 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40979",
"html_url": "https://github.com/huggingface/transformers/pull/40979",
"diff_url": "https://github.com/huggingface/transformers/pull/40979.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40979.patch",
"merged_at": "2025-10-17T10:35:19"
} | # What does this PR do?
Remove `require_torch_bf16_gpu` because it is now unused. | {
"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/40979/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/40979/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40978 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40978/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40978/comments | https://api.github.com/repos/huggingface/transformers/issues/40978/events | https://github.com/huggingface/transformers/pull/40978 | 3,430,427,258 | PR_kwDOCUB6oc6pS7sM | 40,978 | Ci utils | {
"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 | [] | 2025-09-18T13:36:11 | 2025-09-22T14:16:20 | 2025-09-22T14:16:19 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40978",
"html_url": "https://github.com/huggingface/transformers/pull/40978",
"diff_url": "https://github.com/huggingface/transformers/pull/40978.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40978.patch",
"merged_at": "2025-09-22T14:16:19"
} | This PR adds an util script to run tests as they would be in the CI and generate reports.
This is quite useful when fixing the CI and enables devs to use something akin to `run-slow` locally. Also useful when doing big sweeping changes and we want to run the CI locally or on a new hardware.
Also adds the `reports` file to the .gitignore as a QoL improvement for people that runs the CI locally a lot. | {
"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/40978/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/40978/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40977 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40977/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40977/comments | https://api.github.com/repos/huggingface/transformers/issues/40977/events | https://github.com/huggingface/transformers/issues/40977 | 3,430,306,974 | I_kwDOCUB6oc7MdlSe | 40,977 | Whisper Finetuning Issue | {
"login": "AbhijithMallya",
"id": 91281454,
"node_id": "MDQ6VXNlcjkxMjgxNDU0",
"avatar_url": "https://avatars.githubusercontent.com/u/91281454?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/AbhijithMallya",
"html_url": "https://github.com/AbhijithMallya",
"followers_url": "https://api.github.com/users/AbhijithMallya/followers",
"following_url": "https://api.github.com/users/AbhijithMallya/following{/other_user}",
"gists_url": "https://api.github.com/users/AbhijithMallya/gists{/gist_id}",
"starred_url": "https://api.github.com/users/AbhijithMallya/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/AbhijithMallya/subscriptions",
"organizations_url": "https://api.github.com/users/AbhijithMallya/orgs",
"repos_url": "https://api.github.com/users/AbhijithMallya/repos",
"events_url": "https://api.github.com/users/AbhijithMallya/events{/privacy}",
"received_events_url": "https://api.github.com/users/AbhijithMallya/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [
{
"id": 1990918270,
"node_id": "MDU6TGFiZWwxOTkwOTE4Mjcw",
"url": "https://api.github.com/repos/huggingface/transformers/labels/Good%20First%20Issue",
"name": "Good First Issue",
"color": "bbf794",
"default": false,
"description": ""
}
] | closed | false | null | [] | null | [] | 2025-09-18T13:07:44 | 2025-09-30T08:38:33 | 2025-09-30T08:38:33 | NONE | null | null | null | null | Getting `RuntimeError: Dataset scripts are no longer supported, but found common_voice_11_0.py` when finetuning whisper.
Reference : [Whisper Finetuning](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition#single-gpu-whisper-training)
File : `examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py`
Logs
```
09/18/2025 18:32:15 - WARNING - __main__ - Process rank: 0, device: cuda:0, n_gpu: 1, distributed training: False, 16-bits training: True
09/18/2025 18:32:15 - INFO - __main__ - Training/evaluation parameters Seq2SeqTrainingArguments(
_n_gpu=1,
accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False},
adafactor=False,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
auto_find_batch_size=False,
average_tokens_across_devices=False,
batch_eval_metrics=False,
bf16=False,
bf16_full_eval=False,
data_seed=None,
dataloader_drop_last=False,
dataloader_num_workers=0,
dataloader_persistent_workers=False,
dataloader_pin_memory=True,
dataloader_prefetch_factor=None,
ddp_backend=None,
ddp_broadcast_buffers=None,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
ddp_timeout=1800,
debug=[],
deepspeed=None,
disable_tqdm=False,
do_eval=True,
do_predict=False,
do_train=True,
eval_accumulation_steps=None,
eval_delay=0,
eval_do_concat_batches=True,
eval_on_start=False,
eval_steps=1000,
eval_strategy=steps,
eval_use_gather_object=False,
fp16=True,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
fsdp=[],
fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False},
fsdp_min_num_params=0,
fsdp_transformer_layer_cls_to_wrap=None,
full_determinism=False,
generation_config=None,
generation_max_length=225,
generation_num_beams=None,
gradient_accumulation_steps=2,
gradient_checkpointing=True,
gradient_checkpointing_kwargs=None,
greater_is_better=None,
group_by_length=False,
half_precision_backend=auto,
hub_always_push=False,
hub_model_id=None,
hub_private_repo=None,
hub_revision=None,
hub_strategy=every_save,
hub_token=<HUB_TOKEN>,
ignore_data_skip=False,
include_for_metrics=[],
include_inputs_for_metrics=False,
include_num_input_tokens_seen=no,
include_tokens_per_second=False,
jit_mode_eval=False,
label_names=None,
label_smoothing_factor=0.0,
learning_rate=1e-05,
length_column_name=length,
liger_kernel_config=None,
load_best_model_at_end=False,
local_rank=0,
log_level=passive,
log_level_replica=warning,
log_on_each_node=True,
logging_dir=./whisper-small-hi/runs/Sep18_18-32-15_M626DTRV,
logging_first_step=False,
logging_nan_inf_filter=True,
logging_steps=25,
logging_strategy=steps,
lr_scheduler_kwargs={},
lr_scheduler_type=linear,
max_grad_norm=1.0,
max_steps=5000,
metric_for_best_model=None,
mp_parameters=,
neftune_noise_alpha=None,
no_cuda=False,
num_train_epochs=3.0,
optim=adamw_torch_fused,
optim_args=None,
optim_target_modules=None,
output_dir=./whisper-small-hi,
overwrite_output_dir=True,
parallelism_config=None,
past_index=-1,
per_device_eval_batch_size=16,
per_device_train_batch_size=16,
predict_with_generate=True,
prediction_loss_only=False,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
ray_scope=last,
remove_unused_columns=True,
report_to=[],
restore_callback_states_from_checkpoint=False,
resume_from_checkpoint=None,
run_name=None,
save_on_each_node=False,
save_only_model=False,
save_safetensors=True,
save_steps=1000,
save_strategy=steps,
save_total_limit=None,
seed=42,
skip_memory_metrics=True,
sortish_sampler=False,
tf32=None,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
torch_empty_cache_steps=None,
torchdynamo=None,
tpu_metrics_debug=False,
tpu_num_cores=None,
use_cpu=False,
use_legacy_prediction_loop=False,
use_liger_kernel=False,
use_mps_device=False,
warmup_ratio=0.0,
warmup_steps=500,
weight_decay=0.0,
)
09/18/2025 18:32:15 - INFO - __main__ - Training/evaluation parameters Seq2SeqTrainingArguments(
_n_gpu=1,
accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False},
adafactor=False,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
auto_find_batch_size=False,
average_tokens_across_devices=False,
batch_eval_metrics=False,
bf16=False,
bf16_full_eval=False,
data_seed=None,
dataloader_drop_last=False,
dataloader_num_workers=0,
dataloader_persistent_workers=False,
dataloader_pin_memory=True,
dataloader_prefetch_factor=None,
ddp_backend=None,
ddp_broadcast_buffers=None,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
ddp_timeout=1800,
debug=[],
deepspeed=None,
disable_tqdm=False,
do_eval=True,
do_predict=False,
do_train=True,
eval_accumulation_steps=None,
eval_delay=0,
eval_do_concat_batches=True,
eval_on_start=False,
eval_steps=1000,
eval_strategy=steps,
eval_use_gather_object=False,
fp16=True,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
fsdp=[],
fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False},
fsdp_min_num_params=0,
fsdp_transformer_layer_cls_to_wrap=None,
full_determinism=False,
generation_config=None,
generation_max_length=225,
generation_num_beams=None,
gradient_accumulation_steps=2,
gradient_checkpointing=True,
gradient_checkpointing_kwargs=None,
greater_is_better=None,
group_by_length=False,
half_precision_backend=auto,
hub_always_push=False,
hub_model_id=None,
hub_private_repo=None,
hub_revision=None,
hub_strategy=every_save,
hub_token=<HUB_TOKEN>,
ignore_data_skip=False,
include_for_metrics=[],
include_inputs_for_metrics=False,
include_num_input_tokens_seen=no,
include_tokens_per_second=False,
jit_mode_eval=False,
label_names=None,
label_smoothing_factor=0.0,
learning_rate=1e-05,
length_column_name=length,
liger_kernel_config=None,
load_best_model_at_end=False,
local_rank=0,
log_level=passive,
log_level_replica=warning,
log_on_each_node=True,
logging_dir=./whisper-small-hi/runs/Sep18_18-32-15_M626DTRV,
logging_first_step=False,
logging_nan_inf_filter=True,
logging_steps=25,
logging_strategy=steps,
lr_scheduler_kwargs={},
lr_scheduler_type=linear,
max_grad_norm=1.0,
max_steps=5000,
metric_for_best_model=None,
mp_parameters=,
neftune_noise_alpha=None,
no_cuda=False,
num_train_epochs=3.0,
optim=adamw_torch_fused,
optim_args=None,
optim_target_modules=None,
output_dir=./whisper-small-hi,
overwrite_output_dir=True,
parallelism_config=None,
past_index=-1,
per_device_eval_batch_size=16,
per_device_train_batch_size=16,
predict_with_generate=True,
prediction_loss_only=False,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
ray_scope=last,
remove_unused_columns=True,
report_to=[],
restore_callback_states_from_checkpoint=False,
resume_from_checkpoint=None,
run_name=None,
save_on_each_node=False,
save_only_model=False,
save_safetensors=True,
save_steps=1000,
save_strategy=steps,
save_total_limit=None,
seed=42,
skip_memory_metrics=True,
sortish_sampler=False,
tf32=None,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
torch_empty_cache_steps=None,
torchdynamo=None,
tpu_metrics_debug=False,
tpu_num_cores=None,
use_cpu=False,
use_legacy_prediction_loop=False,
use_liger_kernel=False,
use_mps_device=False,
warmup_ratio=0.0,
warmup_steps=500,
weight_decay=0.0,
)
README.md: 14.4kB [00:00, 20.9MB/s]
common_voice_11_0.py: 8.13kB [00:00, 18.2MB/s]
Traceback (most recent call last):
File "/data2/abhijith/transformers/examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py", line 646, in <module>
main()
File "/data2/abhijith/transformers/examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py", line 356, in main
raw_datasets["train"] = load_dataset(
File "/data2/abhijith/transformers/examples/pytorch/speech-recognition/.venv/lib/python3.10/site-packages/datasets/load.py", line 1392, in load_dataset
builder_instance = load_dataset_builder(
File "/data2/abhijith/transformers/examples/pytorch/speech-recognition/.venv/lib/python3.10/site-packages/datasets/load.py", line 1132, in load_dataset_builder
dataset_module = dataset_module_factory(
File "/data2/abhijith/transformers/examples/pytorch/speech-recognition/.venv/lib/python3.10/site-packages/datasets/load.py", line 1031, in dataset_module_factory
raise e1 from None
File "/data2/abhijith/transformers/examples/pytorch/speech-recognition/.venv/lib/python3.10/site-packages/datasets/load.py", line 989, in dataset_module_factory
raise RuntimeError(f"Dataset scripts are no longer supported, but found {filename}")
RuntimeError: Dataset scripts are no longer supported, but found common_voice_11_0.py
``` | {
"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/40977/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/40977/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/40976 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40976/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40976/comments | https://api.github.com/repos/huggingface/transformers/issues/40976/events | https://github.com/huggingface/transformers/pull/40976 | 3,430,276,269 | PR_kwDOCUB6oc6pSaPX | 40,976 | Better defaults for assisted generation | {
"login": "manueldeprada",
"id": 6536835,
"node_id": "MDQ6VXNlcjY1MzY4MzU=",
"avatar_url": "https://avatars.githubusercontent.com/u/6536835?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/manueldeprada",
"html_url": "https://github.com/manueldeprada",
"followers_url": "https://api.github.com/users/manueldeprada/followers",
"following_url": "https://api.github.com/users/manueldeprada/following{/other_user}",
"gists_url": "https://api.github.com/users/manueldeprada/gists{/gist_id}",
"starred_url": "https://api.github.com/users/manueldeprada/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/manueldeprada/subscriptions",
"organizations_url": "https://api.github.com/users/manueldeprada/orgs",
"repos_url": "https://api.github.com/users/manueldeprada/repos",
"events_url": "https://api.github.com/users/manueldeprada/events{/privacy}",
"received_events_url": "https://api.github.com/users/manueldeprada/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-18T13:00:20 | 2025-09-18T15:51:58 | null | CONTRIBUTOR | null | null | true | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40976",
"html_url": "https://github.com/huggingface/transformers/pull/40976",
"diff_url": "https://github.com/huggingface/transformers/pull/40976.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40976.patch",
"merged_at": null
} | #40657 inadvertently changed an implicit algorithmic bias: candidate_generator (the assistant model) was getting logits_processor while the decoding method (main model) was getting prepared_logits_processor. This meant that the assistant was running with T=1 while the main model was using lower temp.
We investigated and its good for speculation to have a hotter assistant model (so it was a good bug that we were not applying the lower temp to the assistant),
<img width="703" height="720" alt="image" src="https://github.com/user-attachments/assets/bd2e5459-fed1-4227-babb-2e5eca52ef2f" />
<img width="715" height="720" alt="image" src="https://github.com/user-attachments/assets/98b3f060-1673-42cb-9b00-c46d71a339a2" />
<img width="1034" height="1055" alt="image" src="https://github.com/user-attachments/assets/c2aa4d0f-61ed-4183-963e-ba5f9768cebc" />
But it should be explicitly set and not a hidden argument forwarding consequence. This PR does that, setting it by default to 1.5.
This PR also fixes:
```
FAILED tests/generation/test_utils.py::GenerationIntegrationTests::test_generated_length_assisted_generation - ValueError: Passing `MinLengthLogitsProcessor` when using `assisted_generation is disabled. Please pass in `min_length` into `.generate()` instead
```
which originates from the same change in LogitsProcessor passing. | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40976/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/40976/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/40975 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40975/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40975/comments | https://api.github.com/repos/huggingface/transformers/issues/40975/events | https://github.com/huggingface/transformers/pull/40975 | 3,430,265,325 | PR_kwDOCUB6oc6pSXza | 40,975 | Use torch.autocast | {
"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 | [] | 2025-09-18T12:57:39 | 2025-09-22T12:53:24 | 2025-09-22T12:18:24 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40975",
"html_url": "https://github.com/huggingface/transformers/pull/40975",
"diff_url": "https://github.com/huggingface/transformers/pull/40975.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40975.patch",
"merged_at": "2025-09-22T12:18:24"
} | # What does this PR do?
The new API is available since PT 2.2 | {
"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/40975/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/40975/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40974 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40974/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40974/comments | https://api.github.com/repos/huggingface/transformers/issues/40974/events | https://github.com/huggingface/transformers/pull/40974 | 3,430,259,237 | PR_kwDOCUB6oc6pSWcu | 40,974 | Testing - DO NOT review or merge | {
"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 | [] | 2025-09-18T12:56:13 | 2025-09-18T13:42:10 | 2025-09-18T13:08:23 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40974",
"html_url": "https://github.com/huggingface/transformers/pull/40974",
"diff_url": "https://github.com/huggingface/transformers/pull/40974.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40974.patch",
"merged_at": null
} | # What does this PR do?
It's getting impossible to track down why CI is failing in a PR with many diffs, so I will use this one as minimalistic version | {
"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/40974/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/40974/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40973 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40973/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40973/comments | https://api.github.com/repos/huggingface/transformers/issues/40973/events | https://github.com/huggingface/transformers/pull/40973 | 3,430,193,983 | PR_kwDOCUB6oc6pSIH7 | 40,973 | [Fix] Fix test file error | {
"login": "YangKai0616",
"id": 103475281,
"node_id": "U_kgDOBiroUQ",
"avatar_url": "https://avatars.githubusercontent.com/u/103475281?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/YangKai0616",
"html_url": "https://github.com/YangKai0616",
"followers_url": "https://api.github.com/users/YangKai0616/followers",
"following_url": "https://api.github.com/users/YangKai0616/following{/other_user}",
"gists_url": "https://api.github.com/users/YangKai0616/gists{/gist_id}",
"starred_url": "https://api.github.com/users/YangKai0616/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/YangKai0616/subscriptions",
"organizations_url": "https://api.github.com/users/YangKai0616/orgs",
"repos_url": "https://api.github.com/users/YangKai0616/repos",
"events_url": "https://api.github.com/users/YangKai0616/events{/privacy}",
"received_events_url": "https://api.github.com/users/YangKai0616/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T12:39:37 | 2025-10-09T15:30:54 | 2025-10-09T15:30:54 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40973",
"html_url": "https://github.com/huggingface/transformers/pull/40973",
"diff_url": "https://github.com/huggingface/transformers/pull/40973.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40973.patch",
"merged_at": "2025-10-09T15:30:54"
} | # 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.
-->
<!-- Remove if not applicable -->
Fixed a minor error in the test file. Otherwise, an error will be reported:
```
cls = <enum 'OptimizerNames'>
value = TrainingArguments(
_n_gpu=1,
accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': Tru...diction_loop=False,
use_liger_kernel=False,
use_mps_device=False,
warmup_ratio=0.0,
warmup_steps=0,
weight_decay=0.0,
)
@classmethod
def _missing_(cls, value):
> raise ValueError(
f"{value} is not a valid {cls.__name__}, please select one of {list(cls._value2member_map_.keys())}"
)
E ValueError: TrainingArguments(
E _n_gpu=1,
E accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False},
E adafactor=False,
E adam_beta1=0.9,
E adam_beta2=0.999,
E adam_epsilon=1e-08,
E auto_find_batch_size=False,
E average_tokens_across_devices=False,
E batch_eval_metrics=False,
E bf16=False,
E bf16_full_eval=False,
E data_seed=None,
E dataloader_drop_last=False,
E dataloader_num_workers=0,
E dataloader_persistent_workers=False,
E dataloader_pin_memory=True,
E dataloader_prefetch_factor=None,
E ddp_backend=None,
E ddp_broadcast_buffers=None,
E ddp_bucket_cap_mb=None,
E ddp_find_unused_parameters=None,
E ddp_timeout=1800,
E debug=[],
E deepspeed=None,
E disable_tqdm=False,
E do_eval=False,
E do_predict=False,
E do_train=False,
E eval_accumulation_steps=None,
E eval_delay=0,
E eval_do_concat_batches=True,
E eval_on_start=False,
E eval_steps=None,
E eval_strategy=IntervalStrategy.NO,
E eval_use_gather_object=False,
E fp16=False,
E fp16_backend=auto,
E fp16_full_eval=False,
E fp16_opt_level=O1,
E fsdp=[],
E fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False},
E fsdp_min_num_params=0,
E fsdp_transformer_layer_cls_to_wrap=None,
E full_determinism=False,
E gradient_accumulation_steps=1,
E gradient_checkpointing=False,
E gradient_checkpointing_kwargs=None,
E greater_is_better=None,
E group_by_length=False,
E half_precision_backend=auto,
E hub_always_push=False,
E hub_model_id=None,
E hub_private_repo=None,
E hub_revision=None,
E hub_strategy=HubStrategy.EVERY_SAVE,
E hub_token=<HUB_TOKEN>,
E ignore_data_skip=False,
E include_for_metrics=[],
E include_inputs_for_metrics=False,
E include_num_input_tokens_seen=no,
E include_tokens_per_second=False,
E jit_mode_eval=False,
E label_names=None,
E label_smoothing_factor=0.0,
E learning_rate=5e-05,
E length_column_name=length,
E liger_kernel_config=None,
E load_best_model_at_end=False,
E local_rank=0,
E log_level=passive,
E log_level_replica=warning,
E log_on_each_node=True,
E logging_dir=None/runs/Sep18_12-37-01_173afc988c7f,
E logging_first_step=False,
E logging_nan_inf_filter=True,
E logging_steps=500,
E logging_strategy=IntervalStrategy.STEPS,
E lr_scheduler_kwargs={},
E lr_scheduler_type=SchedulerType.LINEAR,
E max_grad_norm=1.0,
E max_steps=-1,
E metric_for_best_model=None,
E mp_parameters=,
E neftune_noise_alpha=None,
E no_cuda=False,
E num_train_epochs=3.0,
E optim=OptimizerNames.ADAMW_TORCH_8BIT,
E optim_args=None,
E optim_target_modules=None,
E output_dir=None,
E overwrite_output_dir=False,
E parallelism_config=None,
E past_index=-1,
E per_device_eval_batch_size=8,
E per_device_train_batch_size=8,
E prediction_loss_only=False,
E push_to_hub=False,
E push_to_hub_model_id=None,
E push_to_hub_organization=None,
E push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
E ray_scope=last,
E remove_unused_columns=True,
E report_to=['tensorboard'],
E restore_callback_states_from_checkpoint=False,
E resume_from_checkpoint=None,
E run_name=None,
E save_on_each_node=False,
E save_only_model=False,
E save_safetensors=True,
E save_steps=500,
E save_strategy=SaveStrategy.STEPS,
E save_total_limit=None,
E seed=42,
E skip_memory_metrics=True,
E tf32=None,
E torch_compile=False,
E torch_compile_backend=None,
E torch_compile_mode=None,
E torch_empty_cache_steps=None,
E torchdynamo=None,
E tpu_metrics_debug=False,
E tpu_num_cores=None,
E use_cpu=False,
E use_legacy_prediction_loop=False,
E use_liger_kernel=False,
E use_mps_device=False,
E warmup_ratio=0.0,
E warmup_steps=0,
E weight_decay=0.0,
E ) is not a valid OptimizerNames, please select one of ['adamw_torch', 'adamw_torch_fused', 'adamw_torch_xla', 'adamw_torch_npu_fused', 'adamw_apex_fused', 'adafactor', 'adamw_anyprecision', 'adamw_torch_4bit', 'adamw_torch_8bit', 'ademamix', 'sgd', 'adagrad', 'adamw_bnb_8bit', 'adamw_8bit', 'ademamix_8bit', 'lion_8bit', 'lion_32bit', 'paged_adamw_32bit', 'paged_adamw_8bit', 'paged_ademamix_32bit', 'paged_ademamix_8bit', 'paged_lion_32bit', 'paged_lion_8bit', 'rmsprop', 'rmsprop_bnb', 'rmsprop_bnb_8bit', 'rmsprop_bnb_32bit', 'galore_adamw', 'galore_adamw_8bit', 'galore_adafactor', 'galore_adamw_layerwise', 'galore_adamw_8bit_layerwise', 'galore_adafactor_layerwise', 'lomo', 'adalomo', 'grokadamw', 'schedule_free_radam', 'schedule_free_adamw', 'schedule_free_sgd', 'apollo_adamw', 'apollo_adamw_layerwise', 'stable_adamw']
/usr/local/lib/python3.11/dist-packages/transformers/utils/generic.py:478: ValueError
```
## 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| {
"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/40973/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/40973/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40972 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40972/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40972/comments | https://api.github.com/repos/huggingface/transformers/issues/40972/events | https://github.com/huggingface/transformers/issues/40972 | 3,430,170,833 | I_kwDOCUB6oc7MdEDR | 40,972 | Error when running Qwen3_Next Qwen3-Next model | {
"login": "qsstcl",
"id": 116471041,
"node_id": "U_kgDOBvE1AQ",
"avatar_url": "https://avatars.githubusercontent.com/u/116471041?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/qsstcl",
"html_url": "https://github.com/qsstcl",
"followers_url": "https://api.github.com/users/qsstcl/followers",
"following_url": "https://api.github.com/users/qsstcl/following{/other_user}",
"gists_url": "https://api.github.com/users/qsstcl/gists{/gist_id}",
"starred_url": "https://api.github.com/users/qsstcl/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/qsstcl/subscriptions",
"organizations_url": "https://api.github.com/users/qsstcl/orgs",
"repos_url": "https://api.github.com/users/qsstcl/repos",
"events_url": "https://api.github.com/users/qsstcl/events{/privacy}",
"received_events_url": "https://api.github.com/users/qsstcl/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
}
] | open | false | null | [] | null | [] | 2025-09-18T12:33:24 | 2025-10-10T16:11:38 | null | NONE | null | null | null | null | ### System Info
Copy-and-paste the text below in your GitHub issue and FILL OUT the two last points.
- `transformers` version: 4.57.0.dev0
- Platform: Linux-6.8.0-79-generic-x86_64-with-glibc2.39
- Python version: 3.12.11
- Huggingface_hub version: 0.35.0
- Safetensors version: 0.6.2
- Accelerate version: 1.10.1
- Accelerate config: not found
- DeepSpeed version: not installed
- 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?: <fill in>
- Using GPU in script?: <fill in>
- GPU type: NVIDIA GeForce RTX 4090
### Who can help?
@ArthurZucker @Cyrilvallez @gante
### 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
1、running with test.py given by Qwen official website.
```
from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Qwen/Qwen3-Next-80B-A3B-Instruct"
custom_cache_path = "/mnt/m2_4/models/Qwen3_Next_80B_A3B"
# load the tokenizer and the model from the local directory
tokenizer = AutoTokenizer.from_pretrained(model_name,cache_dir=custom_cache_path)
model = AutoModelForCausalLM.from_pretrained(
model_name,
dtype="auto",
device_map="auto",
cache_dir = custom_cache_path,
)
# prepare the model input
prompt = "Give me a short introduction to large language model."
messages = [
{"role": "user", "content": prompt},
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
# conduct text completion
generated_ids = model.generate(
**model_inputs,
max_new_tokens=16384,
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()
content = tokenizer.decode(output_ids, skip_special_tokens=True)
print("content:", content)
```
2、the error log as follows, error when loading "mtp.***" tensor weights.
<img width="924" height="695" alt="Image" src="https://github.com/user-attachments/assets/b57e7ca9-6794-4cdb-a85b-124724348e72" />
### Expected behavior
output correct answers | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40972/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/40972/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/40971 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40971/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40971/comments | https://api.github.com/repos/huggingface/transformers/issues/40971/events | https://github.com/huggingface/transformers/pull/40971 | 3,430,002,293 | PR_kwDOCUB6oc6pRd2N | 40,971 | Minor test addition for sdpa producing NaNs for pad tokens | {
"login": "DuyguA",
"id": 8277232,
"node_id": "MDQ6VXNlcjgyNzcyMzI=",
"avatar_url": "https://avatars.githubusercontent.com/u/8277232?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/DuyguA",
"html_url": "https://github.com/DuyguA",
"followers_url": "https://api.github.com/users/DuyguA/followers",
"following_url": "https://api.github.com/users/DuyguA/following{/other_user}",
"gists_url": "https://api.github.com/users/DuyguA/gists{/gist_id}",
"starred_url": "https://api.github.com/users/DuyguA/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/DuyguA/subscriptions",
"organizations_url": "https://api.github.com/users/DuyguA/orgs",
"repos_url": "https://api.github.com/users/DuyguA/repos",
"events_url": "https://api.github.com/users/DuyguA/events{/privacy}",
"received_events_url": "https://api.github.com/users/DuyguA/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T11:48:14 | 2025-09-23T10:00:23 | 2025-09-19T13:58:03 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40971",
"html_url": "https://github.com/huggingface/transformers/pull/40971",
"diff_url": "https://github.com/huggingface/transformers/pull/40971.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40971.patch",
"merged_at": null
} | The issue was fixed from PyTorch side, still I added a quick test to confirm the issue is indeed resolved.
Fixes #31035
## 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).
- [ x] Did you write any new necessary 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
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40971/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/40971/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40970 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40970/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40970/comments | https://api.github.com/repos/huggingface/transformers/issues/40970/events | https://github.com/huggingface/transformers/issues/40970 | 3,429,944,190 | I_kwDOCUB6oc7McMt- | 40,970 | Inconsistenly sized batches of images and text (Gemma 3) | {
"login": "KarelKenens",
"id": 143591762,
"node_id": "U_kgDOCI8JUg",
"avatar_url": "https://avatars.githubusercontent.com/u/143591762?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/KarelKenens",
"html_url": "https://github.com/KarelKenens",
"followers_url": "https://api.github.com/users/KarelKenens/followers",
"following_url": "https://api.github.com/users/KarelKenens/following{/other_user}",
"gists_url": "https://api.github.com/users/KarelKenens/gists{/gist_id}",
"starred_url": "https://api.github.com/users/KarelKenens/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/KarelKenens/subscriptions",
"organizations_url": "https://api.github.com/users/KarelKenens/orgs",
"repos_url": "https://api.github.com/users/KarelKenens/repos",
"events_url": "https://api.github.com/users/KarelKenens/events{/privacy}",
"received_events_url": "https://api.github.com/users/KarelKenens/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 | [] | 2025-09-18T11:32:18 | 2025-09-18T13:31:51 | 2025-09-18T13:31:51 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.56.1
- Platform: Windows-11-10.0.26100-SP0
- Python version: 3.13.7
- Huggingface_hub version: 0.35.0
- Safetensors version: 0.6.2
- Accelerate version: 1.10.1
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.8.0+
- 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?: No
- Using GPU in script?: NA
- GPU type: NVIDIA RTX 1000 Ada Generation Laptop GPU
### Who can help?
_No response_
### 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
```python
import numpy as np
import torch
from PIL import Image
from transformers import AutoProcessor, Gemma3ForConditionalGeneration
model_id = "google/gemma-3-4b-it"
model = Gemma3ForConditionalGeneration.from_pretrained(
model_id, device_map="auto", token=<token>
).eval()
processor = AutoProcessor.from_pretrained(
model_id, token=<token>
)
messages = [
{
"role": "user",
"content": [
{
"type": "image",
"image": Image.fromarray(
np.random.randint(0, 255, (512, 512, 3), dtype=np.uint8)
),
},
{"type": "text", "text": "Describe this image in detail."},
],
},
{"role": "assistant", "content": [{"type": "text", "text": " "}]},
{
"role": "user",
"content": [
{
"type": "image",
"image": Image.fromarray(
np.random.randint(0, 255, (512, 512, 3), dtype=np.uint8)
),
},
{"type": "text", "text": "Describe this image in detail."},
],
},
]
inputs = processor.apply_chat_template(
messages,
add_generation_prompt=True,
tokenize=True,
return_dict=True,
return_tensors="pt",
).to(model.device, dtype=torch.bfloat16)
input_len = inputs["input_ids"].shape[-1]
with torch.inference_mode():
generation = model.generate(**inputs, max_new_tokens=100, do_sample=False)
generation = generation[0][input_len:]
decoded = processor.decode(generation, skip_special_tokens=True)
print(decoded)
```
### Expected behavior
### Issue (Seemingly)
Using `ProcessorMixin.apply_chat_template` on multi-turn conversations (e.g. more than "user" message) where multiple turns/messages contain an image. Image batches are created as a batch per "turn". I.e. one "conversation" can correspond to multiple image batches.
In `Gemma3Processor` it is expected that the number of batches corresponds to the number of "conversations". | {
"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/40970/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/40970/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/40969 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40969/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40969/comments | https://api.github.com/repos/huggingface/transformers/issues/40969/events | https://github.com/huggingface/transformers/pull/40969 | 3,429,835,448 | PR_kwDOCUB6oc6pQ5L5 | 40,969 | Fix outdated version checks of accelerator | {
"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 | [] | 2025-09-18T11:03:28 | 2025-09-18T11:51:06 | 2025-09-18T11:49:14 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40969",
"html_url": "https://github.com/huggingface/transformers/pull/40969",
"diff_url": "https://github.com/huggingface/transformers/pull/40969.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40969.patch",
"merged_at": "2025-09-18T11:49:14"
} | # What does this PR do?
Because `ACCELERATE_MIN_VERSION` is 0.26 | {
"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/40969/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/40969/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40968 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40968/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40968/comments | https://api.github.com/repos/huggingface/transformers/issues/40968/events | https://github.com/huggingface/transformers/issues/40968 | 3,429,742,607 | I_kwDOCUB6oc7MbbgP | 40,968 | An error occurred when directly deploying the Qwen3-Next 80B-A3B-Thinking model saved by the save_pretrained of transformers using vllm. However, directly deploying the original model was normal | {
"login": "zhanlun150729",
"id": 34824176,
"node_id": "MDQ6VXNlcjM0ODI0MTc2",
"avatar_url": "https://avatars.githubusercontent.com/u/34824176?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/zhanlun150729",
"html_url": "https://github.com/zhanlun150729",
"followers_url": "https://api.github.com/users/zhanlun150729/followers",
"following_url": "https://api.github.com/users/zhanlun150729/following{/other_user}",
"gists_url": "https://api.github.com/users/zhanlun150729/gists{/gist_id}",
"starred_url": "https://api.github.com/users/zhanlun150729/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/zhanlun150729/subscriptions",
"organizations_url": "https://api.github.com/users/zhanlun150729/orgs",
"repos_url": "https://api.github.com/users/zhanlun150729/repos",
"events_url": "https://api.github.com/users/zhanlun150729/events{/privacy}",
"received_events_url": "https://api.github.com/users/zhanlun150729/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
}
] | open | false | null | [] | null | [] | 2025-09-18T10:37:28 | 2025-10-09T15:14:46 | null | NONE | null | null | null | null | ### System Info
System Info
### Who can help?
(zhihu_0210_bak_0403) root@nb-zhanlun-zl-0914-1-0:/home# python collect_env.py
Collecting environment information...
==============================
System Info
==============================
OS : Ubuntu 22.04.4 LTS (x86_64)
GCC version : (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version : Could not collect
CMake version : version 3.28.3
Libc version : glibc-2.35
==============================
PyTorch Info
==============================
PyTorch version : 2.8.0+cu128
Is debug build : False
CUDA used to build PyTorch : 12.8
ROCM used to build PyTorch : N/A
==============================
Python Environment
==============================
Python version : 3.10.14 (main, May 6 2024, 19:42:50) [GCC 11.2.0] (64-bit runtime)
Python platform : Linux-5.4.0-42-generic-x86_64-with-glibc2.35
==============================
CUDA / GPU Info
==============================
Is CUDA available : True
CUDA runtime version : 12.4.99
CUDA_MODULE_LOADING set to : LAZY
GPU models and configuration :
GPU 0: NVIDIA A800-SXM4-80GB
GPU 1: NVIDIA A800-SXM4-80GB
GPU 2: NVIDIA A800-SXM4-80GB
GPU 3: NVIDIA A800-SXM4-80GB
GPU 4: NVIDIA A800-SXM4-80GB
GPU 5: NVIDIA A800-SXM4-80GB
GPU 6: NVIDIA A800-SXM4-80GB
GPU 7: NVIDIA A800-SXM4-80GB
<img width="1256" height="301" alt="Image" src="https://github.com/user-attachments/assets/4f90b53b-9296-4be5-a299-22ae9bfd1c79" />
### 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
An error occurred when directly deploying the Qwen3-Next 80B-A3B-Thinking model saved by the save_pretrained of transformers using vllm. However, directly deploying the original model was normal
<img width="901" height="206" alt="Image" src="https://github.com/user-attachments/assets/f0b0baab-5f7a-43be-bb6f-96e4323d6570" />
<img width="1691" height="832" alt="Image" src="https://github.com/user-attachments/assets/f77ed66b-226d-4c18-a9c3-cb03d2b86388" />
### Expected behavior
正常部署，实际报错 | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40968/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/40968/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/40967 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40967/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40967/comments | https://api.github.com/repos/huggingface/transformers/issues/40967/events | https://github.com/huggingface/transformers/pull/40967 | 3,429,538,857 | PR_kwDOCUB6oc6pP3jo | 40,967 | Update expected values for one more `test_speculative_generation` after #40949 | {
"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 | [] | 2025-09-18T09:42:42 | 2025-09-18T09:51:59 | 2025-09-18T09:47:14 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40967",
"html_url": "https://github.com/huggingface/transformers/pull/40967",
"diff_url": "https://github.com/huggingface/transformers/pull/40967.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40967.patch",
"merged_at": "2025-09-18T09:47:14"
} | # What does this PR do?
Missed this one
> tests/models/qwen3_moe/test_modeling_qwen3_moe.py::Qwen3MoeIntegrationTest::test_speculative_generation
in #40949 | {
"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/40967/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/40967/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40966 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40966/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40966/comments | https://api.github.com/repos/huggingface/transformers/issues/40966/events | https://github.com/huggingface/transformers/pull/40966 | 3,429,466,199 | PR_kwDOCUB6oc6pPoPJ | 40,966 | use `skip_predictor=True` in vjepa2 `get_vision_features` | {
"login": "hamishs",
"id": 41787553,
"node_id": "MDQ6VXNlcjQxNzg3NTUz",
"avatar_url": "https://avatars.githubusercontent.com/u/41787553?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/hamishs",
"html_url": "https://github.com/hamishs",
"followers_url": "https://api.github.com/users/hamishs/followers",
"following_url": "https://api.github.com/users/hamishs/following{/other_user}",
"gists_url": "https://api.github.com/users/hamishs/gists{/gist_id}",
"starred_url": "https://api.github.com/users/hamishs/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/hamishs/subscriptions",
"organizations_url": "https://api.github.com/users/hamishs/orgs",
"repos_url": "https://api.github.com/users/hamishs/repos",
"events_url": "https://api.github.com/users/hamishs/events{/privacy}",
"received_events_url": "https://api.github.com/users/hamishs/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T09:21:19 | 2025-09-18T11:52:10 | 2025-09-18T11:51:45 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40966",
"html_url": "https://github.com/huggingface/transformers/pull/40966",
"diff_url": "https://github.com/huggingface/transformers/pull/40966.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40966.patch",
"merged_at": "2025-09-18T11:51:45"
} | # What does this PR do?
`VJEPA2Model`'s `get_vision_features` method is meant to only forward pass through the encoder however it uses the model's `forward` method and omits `skip_predictor=True` meaning it also passes through the `predictor` which is wasted computation. This pr sets `skip_predictor=True` so that we only do the necessary calculations.
## 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?
Maybe @pcuenca, @LysandreJik or @koustuvsinha as reviewers of the original VJEPA-2 pr?
| {
"login": "qubvel",
"id": 31920396,
"node_id": "MDQ6VXNlcjMxOTIwMzk2",
"avatar_url": "https://avatars.githubusercontent.com/u/31920396?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/qubvel",
"html_url": "https://github.com/qubvel",
"followers_url": "https://api.github.com/users/qubvel/followers",
"following_url": "https://api.github.com/users/qubvel/following{/other_user}",
"gists_url": "https://api.github.com/users/qubvel/gists{/gist_id}",
"starred_url": "https://api.github.com/users/qubvel/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/qubvel/subscriptions",
"organizations_url": "https://api.github.com/users/qubvel/orgs",
"repos_url": "https://api.github.com/users/qubvel/repos",
"events_url": "https://api.github.com/users/qubvel/events{/privacy}",
"received_events_url": "https://api.github.com/users/qubvel/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40966/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/40966/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40965 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40965/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40965/comments | https://api.github.com/repos/huggingface/transformers/issues/40965/events | https://github.com/huggingface/transformers/pull/40965 | 3,429,272,667 | PR_kwDOCUB6oc6pO-L3 | 40,965 | Add captured actual outputs to CI artifacts | {
"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 | [] | 2025-09-18T08:26:40 | 2025-09-18T13:40:55 | 2025-09-18T13:40:53 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40965",
"html_url": "https://github.com/huggingface/transformers/pull/40965",
"diff_url": "https://github.com/huggingface/transformers/pull/40965.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40965.patch",
"merged_at": "2025-09-18T13:40:53"
} | # What does this PR do?
Follow up work of #40727:
- In the pytest step, add the following so we get the captured actual outputs (for patched methods)
> PATCH_TESTING_METHODS_TO_COLLECT_OUTPUTS=yes _PATCHED_TESTING_METHODS_OUTPUT_DIR=...
- Add a new step `Captured information` to show it on the job run page
- Add the link (of the new step `Captured information`) and the content of `captured_info.txt` to `model_results.json`
This could help the update of expected output values faster.
[Example run](https://github.com/huggingface/transformers/actions/runs/17821624855/job/50665488057#step:16:30) | {
"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/40965/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/40965/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40964 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40964/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40964/comments | https://api.github.com/repos/huggingface/transformers/issues/40964/events | https://github.com/huggingface/transformers/issues/40964 | 3,429,104,797 | I_kwDOCUB6oc7MY_yd | 40,964 | llava model compile output regression caused by `check_model_inputs` | {
"login": "jiqing-feng",
"id": 107918818,
"node_id": "U_kgDOBm614g",
"avatar_url": "https://avatars.githubusercontent.com/u/107918818?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/jiqing-feng",
"html_url": "https://github.com/jiqing-feng",
"followers_url": "https://api.github.com/users/jiqing-feng/followers",
"following_url": "https://api.github.com/users/jiqing-feng/following{/other_user}",
"gists_url": "https://api.github.com/users/jiqing-feng/gists{/gist_id}",
"starred_url": "https://api.github.com/users/jiqing-feng/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/jiqing-feng/subscriptions",
"organizations_url": "https://api.github.com/users/jiqing-feng/orgs",
"repos_url": "https://api.github.com/users/jiqing-feng/repos",
"events_url": "https://api.github.com/users/jiqing-feng/events{/privacy}",
"received_events_url": "https://api.github.com/users/jiqing-feng/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": ""
}
] | closed | false | null | [] | null | [] | 2025-09-18T07:43:29 | 2025-10-06T09:48:53 | 2025-10-06T09:48:53 | CONTRIBUTOR | null | null | null | null | ### System Info
torch 2.10.0.dev20250914+cpu
transformers 4.57.0.dev0
### Who can help?
@zucchini-nlp
### 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 the following code on CPU:
```python
import av
import cv2
import torch
import numpy as np
from PIL import Image
from huggingface_hub import hf_hub_download
from transformers import LlavaProcessor, LlavaForConditionalGeneration
model_id = "llava-hf/llava-interleave-qwen-7b-hf"
processor = LlavaProcessor.from_pretrained(model_id)
model = LlavaForConditionalGeneration.from_pretrained(model_id, dtype=torch.bfloat16)
def read_video_pyav(container, indices):
'''
Decode the video with PyAV decoder.
Args:
container (`av.container.input.InputContainer`): PyAV container.
indices (`List[int]`): List of frame indices to decode.
Returns:
result (np.ndarray): np array of decoded frames of shape (num_frames, height, width, 3).
'''
frames = []
container.seek(0)
start_index = indices[0]
end_index = indices[-1]
for i, frame in enumerate(container.decode(video=0)):
if i > end_index:
break
if i >= start_index and i in indices:
frames.append(frame)
return np.stack([x.to_ndarray(format="rgb24") for x in frames])
def sample_frames(path, num_frames):
video = cv2.VideoCapture(path)
total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
interval = total_frames // num_frames
frames = []
for i in range(total_frames):
ret, frame = video.read()
pil_img = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
if not ret:
continue
if i % interval == 0:
frames.append(pil_img)
video.release()
return frames[:num_frames]
# define a chat history and use `apply_chat_template` to get correctly formatted prompt
# Each value in "content" has to be a list of dicts with types ("text", "image", "video")
conversation = [
{
"role": "user",
"content": [
{"type": "text", "text": "Why is this video funny?"},
{"type": "video"},
],
},
]
prompt = processor.apply_chat_template(conversation, add_generation_prompt=True)
video_path = hf_hub_download(repo_id="raushan-testing-hf/videos-test", filename="sample_demo_1.mp4", repo_type="dataset")
container = av.open(video_path)
# sample uniformly 8 frames from the video, can sample more for longer videos
videos = sample_frames(video_path, 6)
user_prompt = conversation[0]["content"][0]["text"]
toks = "<image>" * 6
prompt = (
"<|im_start|>user"
+ toks
+ f"\n{user_prompt}<|im_end|><|im_start|>assistant"
)
inputs = processor(text=prompt, images=videos, return_tensors="pt").to(
model.device, model.dtype
)
generation_config = model.generation_config
generation_config.do_sample = False
generation_config.use_cache = True
generation_config.temperature = 1.0
generation_config.max_new_tokens = 10
generation_config.min_new_tokens = 10
generation_config.top_p = 1.0
generation_config.cache_implementation = "static"
output = model.generate(**inputs, generation_config=generation_config)
print("eager model output:")
print(processor.decode(output[0][2:], skip_special_tokens=True))
print("\n")
model.forward = torch.compile(model.forward)
output = model.generate(**inputs, generation_config=generation_config)
print("compile model output:")
print(processor.decode(output[0][2:], skip_special_tokens=True))
```
### Expected behavior
Output before the PR #40342
```
eager model output:
Why is this video funny?assistant The video is humorous because the baby is wearing oversized
compile model output:
Why is this video funny?assistant The video is humorous because the baby is wearing oversized
```
Output after the PR #40342
```
eager model output:
Why is this video funny?assistant The video is humorous because the baby is wearing oversized
compile model output:
Why is this video funny?assistant The video is humorous because it shows a baby attempting
``` | {
"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/40964/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/40964/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/40963 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40963/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40963/comments | https://api.github.com/repos/huggingface/transformers/issues/40963/events | https://github.com/huggingface/transformers/issues/40963 | 3,429,044,906 | I_kwDOCUB6oc7MYxKq | 40,963 | Qwen3-next feature dims are mismatched in torch_chunk_gated_delta_rule | {
"login": "jyC23333",
"id": 110331827,
"node_id": "U_kgDOBpOHsw",
"avatar_url": "https://avatars.githubusercontent.com/u/110331827?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/jyC23333",
"html_url": "https://github.com/jyC23333",
"followers_url": "https://api.github.com/users/jyC23333/followers",
"following_url": "https://api.github.com/users/jyC23333/following{/other_user}",
"gists_url": "https://api.github.com/users/jyC23333/gists{/gist_id}",
"starred_url": "https://api.github.com/users/jyC23333/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/jyC23333/subscriptions",
"organizations_url": "https://api.github.com/users/jyC23333/orgs",
"repos_url": "https://api.github.com/users/jyC23333/repos",
"events_url": "https://api.github.com/users/jyC23333/events{/privacy}",
"received_events_url": "https://api.github.com/users/jyC23333/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T07:27:50 | 2025-09-24T11:18:28 | 2025-09-24T11:18:28 | NONE | null | null | null | null | @Cyrilvallez @bozheng-hit
Hello, when I'm running the inference code of Qwen3-next during reviewing the `modeling_qwen3_next.py` file, I found the dimension is confused when the `hidden_states` are passed to the `torch_chunk_gated_delta_rule` function of `Qwen3NextGatedDeltaNet` class.
Please refer the code here: https://github.com/huggingface/transformers/blob/main/src/transformers/models/qwen3_next/modeling_qwen3_next.py#L461
My input query is `Give me a short introduction to large language model`, the same as the official example.
In this function, the shape of the input `key` is `[1， 18， 32， 128]`. Obviously this shape is `[(batch_size, seq_len, num_heads, head_k_dim)]`.
Then the feature are processed by this:
```
query, key, value, beta, g = [
x.transpose(1, 2).contiguous().to(torch.float32) for x in (query, key, value, beta, g)
]
```
and the shape of `key` will be `[1， 32， 18， 128]`,
However, the next line of the code is:
```
batch_size, sequence_length, num_heads, k_head_dim = key.shape
```
It seems that there is a mismatch in the code.
Thanks. | {
"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/40963/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/40963/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/40962 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40962/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40962/comments | https://api.github.com/repos/huggingface/transformers/issues/40962/events | https://github.com/huggingface/transformers/pull/40962 | 3,428,963,501 | PR_kwDOCUB6oc6pN6Ob | 40,962 | perceptron: Isaac-0.1 implementation | {
"login": "AkshatSh",
"id": 9097613,
"node_id": "MDQ6VXNlcjkwOTc2MTM=",
"avatar_url": "https://avatars.githubusercontent.com/u/9097613?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/AkshatSh",
"html_url": "https://github.com/AkshatSh",
"followers_url": "https://api.github.com/users/AkshatSh/followers",
"following_url": "https://api.github.com/users/AkshatSh/following{/other_user}",
"gists_url": "https://api.github.com/users/AkshatSh/gists{/gist_id}",
"starred_url": "https://api.github.com/users/AkshatSh/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/AkshatSh/subscriptions",
"organizations_url": "https://api.github.com/users/AkshatSh/orgs",
"repos_url": "https://api.github.com/users/AkshatSh/repos",
"events_url": "https://api.github.com/users/AkshatSh/events{/privacy}",
"received_events_url": "https://api.github.com/users/AkshatSh/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-18T07:05:39 | 2025-10-29T17:16:54 | null | NONE | null | null | true | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40962",
"html_url": "https://github.com/huggingface/transformers/pull/40962",
"diff_url": "https://github.com/huggingface/transformers/pull/40962.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40962.patch",
"merged_at": null
} | # Perceptron Isaac Implementation
Perceptron released open weight models [Isaac-0.1](https://huggingface.co/PerceptronAI/Isaac-0.1) and [Isaac-0.1-Base](https://huggingface.co/PerceptronAI/Isaac-0.1-Base) a 2B dense model for perception.
| null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40962/reactions",
"total_count": 7,
"+1": 0,
"-1": 0,
"laugh": 0,
"hooray": 0,
"confused": 0,
"heart": 0,
"rocket": 7,
"eyes": 0
} | https://api.github.com/repos/huggingface/transformers/issues/40962/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/40961 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40961/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40961/comments | https://api.github.com/repos/huggingface/transformers/issues/40961/events | https://github.com/huggingface/transformers/pull/40961 | 3,428,784,839 | PR_kwDOCUB6oc6pNTWy | 40,961 | Revert "Update expected values for some `test_speculative_generation`" | {
"login": "Rjayaprakashnarayana",
"id": 45266302,
"node_id": "MDQ6VXNlcjQ1MjY2MzAy",
"avatar_url": "https://avatars.githubusercontent.com/u/45266302?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Rjayaprakashnarayana",
"html_url": "https://github.com/Rjayaprakashnarayana",
"followers_url": "https://api.github.com/users/Rjayaprakashnarayana/followers",
"following_url": "https://api.github.com/users/Rjayaprakashnarayana/following{/other_user}",
"gists_url": "https://api.github.com/users/Rjayaprakashnarayana/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Rjayaprakashnarayana/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Rjayaprakashnarayana/subscriptions",
"organizations_url": "https://api.github.com/users/Rjayaprakashnarayana/orgs",
"repos_url": "https://api.github.com/users/Rjayaprakashnarayana/repos",
"events_url": "https://api.github.com/users/Rjayaprakashnarayana/events{/privacy}",
"received_events_url": "https://api.github.com/users/Rjayaprakashnarayana/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T06:03:08 | 2025-09-18T08:29:01 | 2025-09-18T08:29:00 | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40961",
"html_url": "https://github.com/huggingface/transformers/pull/40961",
"diff_url": "https://github.com/huggingface/transformers/pull/40961.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40961.patch",
"merged_at": null
} | Reverts huggingface/transformers#40949 | {
"login": "Rjayaprakashnarayana",
"id": 45266302,
"node_id": "MDQ6VXNlcjQ1MjY2MzAy",
"avatar_url": "https://avatars.githubusercontent.com/u/45266302?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Rjayaprakashnarayana",
"html_url": "https://github.com/Rjayaprakashnarayana",
"followers_url": "https://api.github.com/users/Rjayaprakashnarayana/followers",
"following_url": "https://api.github.com/users/Rjayaprakashnarayana/following{/other_user}",
"gists_url": "https://api.github.com/users/Rjayaprakashnarayana/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Rjayaprakashnarayana/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Rjayaprakashnarayana/subscriptions",
"organizations_url": "https://api.github.com/users/Rjayaprakashnarayana/orgs",
"repos_url": "https://api.github.com/users/Rjayaprakashnarayana/repos",
"events_url": "https://api.github.com/users/Rjayaprakashnarayana/events{/privacy}",
"received_events_url": "https://api.github.com/users/Rjayaprakashnarayana/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40961/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/40961/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40960 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40960/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40960/comments | https://api.github.com/repos/huggingface/transformers/issues/40960/events | https://github.com/huggingface/transformers/pull/40960 | 3,428,570,788 | PR_kwDOCUB6oc6pMlpO | 40,960 | Fix/issue 40560 train tokens per second | {
"login": "Rjayaprakashnarayana",
"id": 45266302,
"node_id": "MDQ6VXNlcjQ1MjY2MzAy",
"avatar_url": "https://avatars.githubusercontent.com/u/45266302?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Rjayaprakashnarayana",
"html_url": "https://github.com/Rjayaprakashnarayana",
"followers_url": "https://api.github.com/users/Rjayaprakashnarayana/followers",
"following_url": "https://api.github.com/users/Rjayaprakashnarayana/following{/other_user}",
"gists_url": "https://api.github.com/users/Rjayaprakashnarayana/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Rjayaprakashnarayana/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Rjayaprakashnarayana/subscriptions",
"organizations_url": "https://api.github.com/users/Rjayaprakashnarayana/orgs",
"repos_url": "https://api.github.com/users/Rjayaprakashnarayana/repos",
"events_url": "https://api.github.com/users/Rjayaprakashnarayana/events{/privacy}",
"received_events_url": "https://api.github.com/users/Rjayaprakashnarayana/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 | [] | 2025-09-18T04:06:14 | 2025-09-18T12:14:02 | 2025-09-18T11:41:26 | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40960",
"html_url": "https://github.com/huggingface/transformers/pull/40960",
"diff_url": "https://github.com/huggingface/transformers/pull/40960.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40960.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 #40560: `train_tokens_per_second` is wrong after continuing from checkpoint
This PR addresses the issue where `train_tokens_per_second` shows inflated values when resuming training from a checkpoint. The bug occurred because the calculation included cumulative tokens from previous training sessions rather than just the current session.
## Changes Made
- **src/transformers/trainer.py**: Reset training session start time when resuming from checkpoint
- **src/transformers/trainer_utils.py**: Add runtime bounds checking to prevent division errors
- **Training Loop**: Add session-specific token and sample counting for accurate metrics
- **Checkpoint Resume**: Ensure speed metrics use current session data only
## Root Cause
When resuming from checkpoint, the trainer was:
1. Using the original training start time instead of current session start time
2. Including cumulative token counts from all previous training sessions
3. Not properly resetting session-specific metrics
## Solution
1. Reset `_start_time` when resuming from checkpoint to current time
2. Track tokens processed in current session separately from global counts
3. Calculate metrics using session-relative data instead of cumulative totals
4. Add proper error handling for edge cases
## Testing
- ✅ Reproduced original issue with test script
- ✅ Verified fix resolves the problem
- ✅ Tested with `run_clm.py` script
- ✅ All existing tests pass
## Backward Compatibility
This change is fully backward compatible. It only affects the calculation of training metrics and does not change:
- Model weights or training behavior
- Checkpoint format or loading
- API interfaces
- Configuration options
## Additional Notes
This is marked as a "Good First Issue" and affects a commonly used training script (`run_clm.py`). The fix ensures that developers get accurate performance metrics when resuming training, which is important for:
- Performance monitoring and optimization
- Resource planning and scaling
- Training progress assessment
## 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?
- [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.
<!-- 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
contributed fixes by :- jayaprakashrayani@gmail.com
| {
"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/40960/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/40960/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40959 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40959/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40959/comments | https://api.github.com/repos/huggingface/transformers/issues/40959/events | https://github.com/huggingface/transformers/pull/40959 | 3,428,431,105 | PR_kwDOCUB6oc6pMHrS | 40,959 | Fix/issue 40560 wrong train tokens per second bug | {
"login": "Rjayaprakashnarayana",
"id": 45266302,
"node_id": "MDQ6VXNlcjQ1MjY2MzAy",
"avatar_url": "https://avatars.githubusercontent.com/u/45266302?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Rjayaprakashnarayana",
"html_url": "https://github.com/Rjayaprakashnarayana",
"followers_url": "https://api.github.com/users/Rjayaprakashnarayana/followers",
"following_url": "https://api.github.com/users/Rjayaprakashnarayana/following{/other_user}",
"gists_url": "https://api.github.com/users/Rjayaprakashnarayana/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Rjayaprakashnarayana/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Rjayaprakashnarayana/subscriptions",
"organizations_url": "https://api.github.com/users/Rjayaprakashnarayana/orgs",
"repos_url": "https://api.github.com/users/Rjayaprakashnarayana/repos",
"events_url": "https://api.github.com/users/Rjayaprakashnarayana/events{/privacy}",
"received_events_url": "https://api.github.com/users/Rjayaprakashnarayana/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T02:44:34 | 2025-09-18T03:58:31 | 2025-09-18T03:58:13 | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40959",
"html_url": "https://github.com/huggingface/transformers/pull/40959",
"diff_url": "https://github.com/huggingface/transformers/pull/40959.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40959.patch",
"merged_at": null
} | ## Description
Fixes #40560: `train_tokens_per_second` is wrong after continuing from checkpoint
This PR addresses the issue where `train_tokens_per_second` shows inflated values when resuming training from a checkpoint. The bug occurred because the calculation included cumulative tokens from previous training sessions rather than just the current session.
## Changes Made
- **src/transformers/trainer.py**: Reset training session start time when resuming from checkpoint
- **src/transformers/trainer_utils.py**: Add runtime bounds checking to prevent division errors
- **Training Loop**: Add session-specific token and sample counting for accurate metrics
- **Checkpoint Resume**: Ensure speed metrics use current session data only
## Root Cause
When resuming from checkpoint, the trainer was:
1. Using the original training start time instead of current session start time
2. Including cumulative token counts from all previous training sessions
3. Not properly resetting session-specific metrics
## Solution
1. Reset `_start_time` when resuming from checkpoint to current time
2. Track tokens processed in current session separately from global counts
3. Calculate metrics using session-relative data instead of cumulative totals
4. Add proper error handling for edge cases
## Testing
- ✅ Reproduced original issue with test script
- ✅ Verified fix resolves the problem
- ✅ Tested with `run_clm.py` script
- ✅ Added tests
## Backward Compatibility
This change is fully backward compatible. It only affects the calculation of training metrics and does not change:
- Model weights or training behavior
- Checkpoint format or loading
- API interfaces
- Configuration options
## Additional Notes
This is marked as a "Good First Issue" and affects a commonly used training script (`run_clm.py`). The fix ensures that developers get accurate performance metrics when resuming training, which is important for:
- Performance monitoring and optimization
- Resource planning and scaling
- Training progress assessment
# 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
- vision models: @amyeroberts, @qubvel
- speech models: @eustlb
- graph models: @clefourrier
Library:
- flax: @gante and @Rocketknight1
- generate: @zucchini-nlp (visual-language models) or @gante (all others)
- pipelines: @Rocketknight1
- tensorflow: @gante and @Rocketknight1
- tokenizers: @ArthurZucker
- trainer: @zach-huggingface, @SunMarc and @qgallouedec
- chat templates: @Rocketknight1
Integrations:
- deepspeed: HF Trainer/Accelerate: @SunMarc @zach-huggingface
- ray/raytune: @richardliaw, @amogkam
- Big Model Inference: @SunMarc
- quantization (bitsandbytes, autogpt): @SunMarc @MekkCyber
Documentation: @stevhliu
HF projects:
- accelerate: [different repo](https://github.com/huggingface/accelerate)
- datasets: [different repo](https://github.com/huggingface/datasets)
- diffusers: [different repo](https://github.com/huggingface/diffusers)
- rust tokenizers: [different repo](https://github.com/huggingface/tokenizers)
Maintained examples (not research project or legacy):
- Flax: @Rocketknight1
- PyTorch: See Models above and tag the person corresponding to the modality of the example.
- TensorFlow: @Rocketknight1
-->
| {
"login": "Rjayaprakashnarayana",
"id": 45266302,
"node_id": "MDQ6VXNlcjQ1MjY2MzAy",
"avatar_url": "https://avatars.githubusercontent.com/u/45266302?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Rjayaprakashnarayana",
"html_url": "https://github.com/Rjayaprakashnarayana",
"followers_url": "https://api.github.com/users/Rjayaprakashnarayana/followers",
"following_url": "https://api.github.com/users/Rjayaprakashnarayana/following{/other_user}",
"gists_url": "https://api.github.com/users/Rjayaprakashnarayana/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Rjayaprakashnarayana/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Rjayaprakashnarayana/subscriptions",
"organizations_url": "https://api.github.com/users/Rjayaprakashnarayana/orgs",
"repos_url": "https://api.github.com/users/Rjayaprakashnarayana/repos",
"events_url": "https://api.github.com/users/Rjayaprakashnarayana/events{/privacy}",
"received_events_url": "https://api.github.com/users/Rjayaprakashnarayana/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40959/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/40959/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40958 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40958/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40958/comments | https://api.github.com/repos/huggingface/transformers/issues/40958/events | https://github.com/huggingface/transformers/issues/40958 | 3,428,390,554 | I_kwDOCUB6oc7MWRaa | 40,958 | Variety of models not working sufficiently good on tool calling as has been advertised | {
"login": "dvn8weil",
"id": 190058927,
"node_id": "U_kgDOC1QRrw",
"avatar_url": "https://avatars.githubusercontent.com/u/190058927?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/dvn8weil",
"html_url": "https://github.com/dvn8weil",
"followers_url": "https://api.github.com/users/dvn8weil/followers",
"following_url": "https://api.github.com/users/dvn8weil/following{/other_user}",
"gists_url": "https://api.github.com/users/dvn8weil/gists{/gist_id}",
"starred_url": "https://api.github.com/users/dvn8weil/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/dvn8weil/subscriptions",
"organizations_url": "https://api.github.com/users/dvn8weil/orgs",
"repos_url": "https://api.github.com/users/dvn8weil/repos",
"events_url": "https://api.github.com/users/dvn8weil/events{/privacy}",
"received_events_url": "https://api.github.com/users/dvn8weil/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 | [] | 2025-09-18T02:23:37 | 2025-10-26T08:02:12 | 2025-10-26T08:02:12 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.53.3
- Platform: Linux-6.14.0-1014-gcp-x86_64-with-glibc2.39
- Python version: 3.12.3
- Huggingface_hub version: 0.34.4
- Safetensors version: 0.6.2
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.7.1+cu126 (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?: No
- Using GPU in script?: No
- GPU type: NVIDIA L4
### Who can help?
I have been trying out various small/medium size models for my tool calling usecase.
The models i tried out are :
- mistralai/Mistral-7B-Instruct-v0.3
- ibm-granite/granite-20b-functioncalling
- Team-ACE/ToolACE-2-Llama-3.1-8B
- HuggingFaceTB/SmolLM3-3B
I tried out these because these have been said to perform very good for tool calling capability/use-case but neither of these have worked good enough (with my setup) for my use-case which is consistent function calling , with `"tool_choice": "auto"`. On the other hand, Qwen3-12B has been working very well for me , w.r.t. tool calling capabilities.
### 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
I am using vLLM to run the models.
Steps to reproduce the issues :
- mistralai/Mistral-7B-Instruct-v0.3 : https://github.com/huggingface/transformers/issues/39852
- ibm-granite/granite-20b-functioncalling : https://github.com/huggingface/transformers/issues/40785
- HuggingFaceTB/SmolLM3-3B : https://github.com/huggingface/transformers/issues/40957
### Expected behavior
I am trying to understand if i have been setting up the models in an incorrect way. If someone can help me with understanding this, it will be really great. | {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40958/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/40958/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/40957 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40957/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40957/comments | https://api.github.com/repos/huggingface/transformers/issues/40957/events | https://github.com/huggingface/transformers/issues/40957 | 3,428,385,272 | I_kwDOCUB6oc7MWQH4 | 40,957 | Incomplete tool call arguments in HuggingFaceTB/SmolLM3-3B streaming output | {
"login": "dvn8weil",
"id": 190058927,
"node_id": "U_kgDOC1QRrw",
"avatar_url": "https://avatars.githubusercontent.com/u/190058927?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/dvn8weil",
"html_url": "https://github.com/dvn8weil",
"followers_url": "https://api.github.com/users/dvn8weil/followers",
"following_url": "https://api.github.com/users/dvn8weil/following{/other_user}",
"gists_url": "https://api.github.com/users/dvn8weil/gists{/gist_id}",
"starred_url": "https://api.github.com/users/dvn8weil/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/dvn8weil/subscriptions",
"organizations_url": "https://api.github.com/users/dvn8weil/orgs",
"repos_url": "https://api.github.com/users/dvn8weil/repos",
"events_url": "https://api.github.com/users/dvn8weil/events{/privacy}",
"received_events_url": "https://api.github.com/users/dvn8weil/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 | [] | 2025-09-18T02:21:08 | 2025-10-26T08:02:14 | 2025-10-26T08:02:14 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.53.3
- Platform: Linux-6.14.0-1014-gcp-x86_64-with-glibc2.39
- Python version: 3.12.3
- Huggingface_hub version: 0.34.4
- Safetensors version: 0.6.2
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.7.1+cu126 (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?: No
- Using GPU in script?: No
- GPU type: NVIDIA L4
### 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
vllm setup :
```
vllm serve HuggingFaceTB/SmolLM3-3B \
--enable-auto-tool-choice \
--tool-call-parser=hermes --reasoning-parser=qwen3
```
request as cURL :
```
curl --location 'http://0.0.0.0:8000/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer EMPTY' \
--data '{
"model": "HuggingFaceTB/SmolLM3-3B",
"messages": [
{
"role": "system",
"content": "You are a concise assistant that helps with tools."
},
{
"role": "user",
"content": "Tell me about Indian Democracy origin and then Can you add 23 and 51, and multiply 29 and 89"
}
],
"enable_thinking": false,
"chat_template_kwargs": {
"enable_thinking": false
},
"stream" : true,
"tools": [
{
"type": "function",
"function": {
"name": "get_current_weather",
"description": "Get the current weather in a given location",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and state, e.g. San Francisco, CA"
},
"unit": {
"type": "string",
"enum": [
"celsius",
"fahrenheit"
]
}
},
"required": [
"location"
]
}
}
},
{
"type": "function",
"function": {
"name": "add",
"description": "adds two numbers\n",
"parameters": {
"type": "object",
"properties": {
"x": {
"type": "number",
"description": ""
},
"y": {
"type": "number",
"description": ""
}
},
"required": [
"x",
"y"
]
}
}
},
{
"type": "function",
"function": {
"name": "multiply",
"description": "multiply two numbers\n",
"parameters": {
"type": "object",
"properties": {
"x": {
"type": "number",
"description": ""
},
"y": {
"type": "number",
"description": ""
}
},
"required": [
"x",
"y"
]
}
}
}
],
"tool_choice": "auto"
}'
```
### Expected behavior
Output is :
```
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"role":"assistant","content":""},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"content":null},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"id":"chatcmpl-tool-acf94c2655364429b96d82af14949d67","type":"function","index":0,"function":{"name":"get_current_weather"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"location\": \""}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"New"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Delhi"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\","}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"unit"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"c"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"elsius"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"content":"\n"},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"id":"chatcmpl-tool-94268bea54114b139426a9d2c62b6bec","type":"function","index":1,"function":{"name":"add"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"{\"x\":"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":" "}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"51"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"}"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"content":"\n"},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"id":"chatcmpl-tool-6146714702e44bd4afbe37d822332fc9","type":"function","index":2,"function":{"name":"multiply"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":"{\"x\":"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":" "}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":"89"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":"}"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"content":""},"logprobs":null,"finish_reason":"tool_calls","stop_reason":null}]}
data: [DONE]
```
The tool calls are incomplete , i.e. for tool calls that require two arguments, only one is provided. For instance, this is the multiple tool call part of the output :
```
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"id":"chatcmpl-tool-6146714702e44bd4afbe37d822332fc9","type":"function","index":2,"function":{"name":"multiply"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":"{\"x\":"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":" "}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":"89"}}]},"logprobs":null,"finish_reason":null}]}
data: {"id":"chatcmpl-2eaecc77b750452fb0a1f326d07334b5","object":"chat.completion.chunk","created":1758130251,"model":"HuggingFaceTB/SmolLM3-3B","choices":[{"index":0,"delta":{"tool_calls":[{"index":2,"function":{"arguments":"}"}}]},"logprobs":null,"finish_reason":null}]}
```
i.e. the tool call output is `multiply {x: 89}` , when it also needs the `y` parameter value
| {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40957/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/40957/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/40956 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40956/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40956/comments | https://api.github.com/repos/huggingface/transformers/issues/40956/events | https://github.com/huggingface/transformers/pull/40956 | 3,428,270,035 | PR_kwDOCUB6oc6pLk4y | 40,956 | Fix Issue #39030: AutoTokenizer.from_pretrained does not propagate token | {
"login": "brandenkmurray",
"id": 8684326,
"node_id": "MDQ6VXNlcjg2ODQzMjY=",
"avatar_url": "https://avatars.githubusercontent.com/u/8684326?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/brandenkmurray",
"html_url": "https://github.com/brandenkmurray",
"followers_url": "https://api.github.com/users/brandenkmurray/followers",
"following_url": "https://api.github.com/users/brandenkmurray/following{/other_user}",
"gists_url": "https://api.github.com/users/brandenkmurray/gists{/gist_id}",
"starred_url": "https://api.github.com/users/brandenkmurray/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/brandenkmurray/subscriptions",
"organizations_url": "https://api.github.com/users/brandenkmurray/orgs",
"repos_url": "https://api.github.com/users/brandenkmurray/repos",
"events_url": "https://api.github.com/users/brandenkmurray/events{/privacy}",
"received_events_url": "https://api.github.com/users/brandenkmurray/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-18T01:19:27 | 2025-09-18T14:36:52 | 2025-09-18T13:22:20 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40956",
"html_url": "https://github.com/huggingface/transformers/pull/40956",
"diff_url": "https://github.com/huggingface/transformers/pull/40956.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40956.patch",
"merged_at": "2025-09-18T13:22:20"
} | # What does this PR do?
Fixes #39030. Propagates token through `list_repo_templates` to `list_repo_tree`.
## Who can review?
@ArthurZucker @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/40956/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/40956/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40955 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40955/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40955/comments | https://api.github.com/repos/huggingface/transformers/issues/40955/events | https://github.com/huggingface/transformers/pull/40955 | 3,428,096,370 | PR_kwDOCUB6oc6pK-3n | 40,955 | Fix more dates in model cards and wrong modalities in _toctree.yml | {
"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 | [] | 2025-09-17T23:46:00 | 2025-09-19T13:47:28 | 2025-09-19T13:47:28 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40955",
"html_url": "https://github.com/huggingface/transformers/pull/40955",
"diff_url": "https://github.com/huggingface/transformers/pull/40955.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40955.patch",
"merged_at": "2025-09-19T13:47:28"
} | Missed a few in my previous PR.
Also some models were in the wrong modalities in `_toctree.yml`
| {
"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/40955/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/40955/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40954 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40954/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40954/comments | https://api.github.com/repos/huggingface/transformers/issues/40954/events | https://github.com/huggingface/transformers/pull/40954 | 3,427,969,782 | PR_kwDOCUB6oc6pKjXj | 40,954 | Fix Issue #40913: Respect user-provided chat_template parameter in processor creation | {
"login": "qizwiz",
"id": 211900,
"node_id": "MDQ6VXNlcjIxMTkwMA==",
"avatar_url": "https://avatars.githubusercontent.com/u/211900?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/qizwiz",
"html_url": "https://github.com/qizwiz",
"followers_url": "https://api.github.com/users/qizwiz/followers",
"following_url": "https://api.github.com/users/qizwiz/following{/other_user}",
"gists_url": "https://api.github.com/users/qizwiz/gists{/gist_id}",
"starred_url": "https://api.github.com/users/qizwiz/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/qizwiz/subscriptions",
"organizations_url": "https://api.github.com/users/qizwiz/orgs",
"repos_url": "https://api.github.com/users/qizwiz/repos",
"events_url": "https://api.github.com/users/qizwiz/events{/privacy}",
"received_events_url": "https://api.github.com/users/qizwiz/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | open | false | null | [] | null | [] | 2025-09-17T22:25:37 | 2025-09-18T08:32:09 | null | NONE | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40954",
"html_url": "https://github.com/huggingface/transformers/pull/40954",
"diff_url": "https://github.com/huggingface/transformers/pull/40954.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40954.patch",
"merged_at": null
} | This PR fixes a bug where user-provided `chat_template` parameters were being overwritten by model defaults when creating a processor. The fix ensures that user-provided values take precedence over model defaults.
Fixes #40913
## Problem
When creating a processor with a user-provided `chat_template` parameter, the value was being overwritten by the model's default chat template. This prevented users from customizing the chat template as intended.
**Reproduction:**
```python
from transformers import AutoProcessor
processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-3B-Instruct", chat_template="test")
print(processor.chat_template) # Should print "test" but prints the default template
```
## Root Cause
In `src/transformers/processing_utils.py` line 1086, the code was unconditionally setting:
```python
if chat_templates:
kwargs["chat_template"] = chat_templates
```
This overwrote any user-provided `chat_template` parameter.
## Solution
The fix modifies the condition to only set the chat template from model files if no user-provided value exists:
```python
if chat_templates and "chat_template" not in kwargs:
kwargs["chat_template"] = chat_templates
```
This ensures that:
1. User-provided `chat_template` values take precedence
2. Model defaults are still used when no user value is provided
3. Backward compatibility is maintained
## Testing
Added a test case that verifies:
1. User-provided chat_template values are preserved
2. Model defaults are used when no user value is provided
## Impact
This is a backward-compatible bug fix that allows users to override chat templates as expected when loading processors. | null | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40954/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/40954/timeline | null | null | null | null | true | false |
https://api.github.com/repos/huggingface/transformers/issues/40953 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40953/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40953/comments | https://api.github.com/repos/huggingface/transformers/issues/40953/events | https://github.com/huggingface/transformers/issues/40953 | 3,427,895,973 | I_kwDOCUB6oc7MUYql | 40,953 | gpt-oss-120b inference failed running on 16 GPUs, single node and with tp_plan="auto" | {
"login": "yuanhangsu1986",
"id": 185134695,
"node_id": "U_kgDOCwjuZw",
"avatar_url": "https://avatars.githubusercontent.com/u/185134695?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/yuanhangsu1986",
"html_url": "https://github.com/yuanhangsu1986",
"followers_url": "https://api.github.com/users/yuanhangsu1986/followers",
"following_url": "https://api.github.com/users/yuanhangsu1986/following{/other_user}",
"gists_url": "https://api.github.com/users/yuanhangsu1986/gists{/gist_id}",
"starred_url": "https://api.github.com/users/yuanhangsu1986/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/yuanhangsu1986/subscriptions",
"organizations_url": "https://api.github.com/users/yuanhangsu1986/orgs",
"repos_url": "https://api.github.com/users/yuanhangsu1986/repos",
"events_url": "https://api.github.com/users/yuanhangsu1986/events{/privacy}",
"received_events_url": "https://api.github.com/users/yuanhangsu1986/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T21:50:58 | 2025-10-26T08:02:15 | 2025-10-26T08:02:15 | NONE | null | null | null | null | ### System Info
- `transformers` version: 4.57.0.dev0
- Platform: Linux-5.4.0-80-generic-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 0.34.5
- Safetensors version: 0.5.3
- Accelerate version: 1.6.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.6.0+cu124 (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?: <fill in>
- Using GPU in script?: <fill in>
- GPU type: Tesla V100-SXM3-32GB-H
### Who can help?
@SunMarc @zucchini-nlp @vasqu @ArthurZucker @Cyrilvallez
Running inference with TP with gpt-oss-120b model on a node with 16 GPUs.
Got the following error in the cross attention layer:
```
[rank0]: File "/usr/local/lib/python3.10/dist-packages/transformers/models/gpt_oss/modeling_gpt_oss.py", line 314, in forward [rank0]: key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
[rank0]: RuntimeError: shape '[1, 89, -1, 64]' is invalid for input of size 2848
```
It seems the projected tensor shape, which is 2848, is less than the minimum allowed by the hidden_shape. This happens due to the TP partitioning of the projection matrix.
I'm wandering if changing the self.head_dim from 64 to 32 is a possible solution to this?
In other words, pseudo-code:
```
k_proj = self.k_proj(hidden_states)
if torch.numel(k_proj) < math.prod((*input_shape, 1, self.head_dim)):
head_dim = self.head_dim / math.prod((*input_shape, 1, self.head_dim)) * torch.numel(k_proj)
...
```
One thing to note that is the GPUs we used is v100, which is why we need that many GPUs to run gpt-oss-120b.
[Here ](https://huggingface.co/openai/gpt-oss-20b/discussions/61) is the official fix for running gpt-oss on older GPUs.
### Information
- [x] The official example scripts
- [ ] 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
To reproduce:
- python code
gpt-oss-120b.py
```
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, Mxfp4Config
tokenizer = AutoTokenizer.from_pretrained("openai/gpt-oss-120b")
quantization_config = Mxfp4Config(dequantize=False)
model_kwargs = dict(attn_implementation="eager", dtype=torch.bfloat16, use_cache=True, tp_plan="auto", quantization_config=quantization_config)
model = AutoModelForCausalLM.from_pretrained("openai/gpt-oss-120b", **model_kwargs).cuda()
SYSTEM_PROMPT = f"Please answer the following question in English."
USER_PROMPT = "What is the capital of Australia?"
messages = [
{"role": "system", "content": SYSTEM_PROMPT},
{"role": "user", "content": USER_PROMPT},
]
input_ids = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
return_tensors="pt",
).to(model.device)
gen_kwargs = {"max_new_tokens": 512, "do_sample": True, "temperature": 0.6, "top_p": None, "top_k": None}
output_ids = model.generate(input_ids, **gen_kwargs)
response = tokenizer.batch_decode(output_ids)[0]
print(response)
```
- Launch job with torchrun
```
torchrun --nproc-per-node 16 gpt-oss-120b.py
```
### Expected behavior
Should run without error | {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40953/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/40953/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/40952 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40952/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40952/comments | https://api.github.com/repos/huggingface/transformers/issues/40952/events | https://github.com/huggingface/transformers/issues/40952 | 3,427,769,404 | I_kwDOCUB6oc7MT5w8 | 40,952 | InstructBLIP requiring eager attention specified to work properly | {
"login": "nhatkhtn",
"id": 61368343,
"node_id": "MDQ6VXNlcjYxMzY4MzQz",
"avatar_url": "https://avatars.githubusercontent.com/u/61368343?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/nhatkhtn",
"html_url": "https://github.com/nhatkhtn",
"followers_url": "https://api.github.com/users/nhatkhtn/followers",
"following_url": "https://api.github.com/users/nhatkhtn/following{/other_user}",
"gists_url": "https://api.github.com/users/nhatkhtn/gists{/gist_id}",
"starred_url": "https://api.github.com/users/nhatkhtn/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/nhatkhtn/subscriptions",
"organizations_url": "https://api.github.com/users/nhatkhtn/orgs",
"repos_url": "https://api.github.com/users/nhatkhtn/repos",
"events_url": "https://api.github.com/users/nhatkhtn/events{/privacy}",
"received_events_url": "https://api.github.com/users/nhatkhtn/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 | [] | 2025-09-17T20:53:15 | 2025-09-19T08:45:39 | 2025-09-19T08:45:39 | NONE | null | null | null | null | ### System Info
The model `instructblip-vicuna-7b` gives wildly different results when specifying `attn_implementation='eager'` and not specifying anything. Furthermore, when specifying eager attention, the model seems to work correctly, while not specifying gives valid but noticeably worse results. For example, when captioning a COCO image:
- Running with specifying eager: "The image features a giraffe standing near a tree, with its reflection visible in the water. The giraffe is positioned in the center of the scene, with its head and neck stretched upwards towards the tree. There are several trees scattered throughout the scene, with one tree located on the left side of the image and another on the right side. The water body is located near the giraffe, with its reflection visible in the water. The overall scene appears to be a peaceful and serene setting, with the giraffe standing tall and majestic amidst the trees and water"
- Running without specifying eager: "a reflection of a tree in a lake with a blue sky and clouds"
System info:
- `transformers` version: 4.52.4
- Platform: Linux-5.14.0-503.40.1.el9_5.x86_64-x86_64-with-glibc2.34
- Python version: 3.12.11
- Huggingface_hub version: 0.33.0
- Safetensors version: 0.5.3
- Accelerate version: 1.7.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (GPU?): 2.7.1+cu128 (True)
- 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?: No
- Using GPU in script?: Yes
- GPU type: NVIDIA B200
### Who can help?
@zucchini-nlp
### 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
Run the following code, with and without the `attn_implementation="eager",` line
```
import requests
import torch
from PIL import Image
from transformers import InstructBlipProcessor, InstructBlipForConditionalGeneration
device = "cuda"
model = InstructBlipForConditionalGeneration.from_pretrained(
"Salesforce/instructblip-vicuna-7b",
attn_implementation="eager",
).to(device)
processor = InstructBlipProcessor.from_pretrained("Salesforce/instructblip-vicuna-7b")
url1 = "http://images.cocodataset.org/val2014/COCO_val2014_000000030925.jpg"
image1 = Image.open(requests.get(url1, stream=True).raw).convert("RGB")
prompt = "Please help me describe the image in detail."
inputs = processor(images=image1, text=prompt, return_tensors="pt", padding=True).to(device)
with torch.inference_mode():
outputs = model.generate(
**inputs,
max_length=256,
)
texts = processor.batch_decode(outputs, skip_special_tokens=True)
texts
```
### Expected behavior
Since InstructBLIP does not support any attention implementation other than eager, the output when specifying attn_implementation='eager' and when not specifying anything should be the same. | {
"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/40952/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/40952/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/40951 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40951/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40951/comments | https://api.github.com/repos/huggingface/transformers/issues/40951/events | https://github.com/huggingface/transformers/pull/40951 | 3,427,713,267 | PR_kwDOCUB6oc6pJrR_ | 40,951 | fix(timm): Catch 'Unknown model' RuntimeError in the Gemma 3n MobileNetV5 vision encoder | {
"login": "harshaljanjani",
"id": 75426551,
"node_id": "MDQ6VXNlcjc1NDI2NTUx",
"avatar_url": "https://avatars.githubusercontent.com/u/75426551?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/harshaljanjani",
"html_url": "https://github.com/harshaljanjani",
"followers_url": "https://api.github.com/users/harshaljanjani/followers",
"following_url": "https://api.github.com/users/harshaljanjani/following{/other_user}",
"gists_url": "https://api.github.com/users/harshaljanjani/gists{/gist_id}",
"starred_url": "https://api.github.com/users/harshaljanjani/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/harshaljanjani/subscriptions",
"organizations_url": "https://api.github.com/users/harshaljanjani/orgs",
"repos_url": "https://api.github.com/users/harshaljanjani/repos",
"events_url": "https://api.github.com/users/harshaljanjani/events{/privacy}",
"received_events_url": "https://api.github.com/users/harshaljanjani/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T20:29:13 | 2025-09-18T13:15:08 | 2025-09-18T13:09:09 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40951",
"html_url": "https://github.com/huggingface/transformers/pull/40951",
"diff_url": "https://github.com/huggingface/transformers/pull/40951.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40951.patch",
"merged_at": "2025-09-18T13:09:08"
} | A first time contributor to HF here!
## What does this PR do?
- Fixes a `RuntimeError` that occurs when loading the Gemma 3n model with an outdated version of the `timm` library; it's caused by the absence of the required `mobilenetv5_300m_enc` vision model in older `timm` versions.
- The newer error explicitly tells the user that the vision model for Gemma 3n is missing and provides them with the exact command to upgrade timm to a compatible version, resolving the issue.
Closes #39208.
cc: @ArthurZucker @qubvel
## 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). | {
"login": "qubvel",
"id": 31920396,
"node_id": "MDQ6VXNlcjMxOTIwMzk2",
"avatar_url": "https://avatars.githubusercontent.com/u/31920396?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/qubvel",
"html_url": "https://github.com/qubvel",
"followers_url": "https://api.github.com/users/qubvel/followers",
"following_url": "https://api.github.com/users/qubvel/following{/other_user}",
"gists_url": "https://api.github.com/users/qubvel/gists{/gist_id}",
"starred_url": "https://api.github.com/users/qubvel/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/qubvel/subscriptions",
"organizations_url": "https://api.github.com/users/qubvel/orgs",
"repos_url": "https://api.github.com/users/qubvel/repos",
"events_url": "https://api.github.com/users/qubvel/events{/privacy}",
"received_events_url": "https://api.github.com/users/qubvel/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40951/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/40951/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40950 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40950/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40950/comments | https://api.github.com/repos/huggingface/transformers/issues/40950/events | https://github.com/huggingface/transformers/pull/40950 | 3,427,500,070 | PR_kwDOCUB6oc6pI8n3 | 40,950 | Deprecate Trackio environment variables and deploy to Spaces by default | {
"login": "qgallouedec",
"id": 45557362,
"node_id": "MDQ6VXNlcjQ1NTU3MzYy",
"avatar_url": "https://avatars.githubusercontent.com/u/45557362?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/qgallouedec",
"html_url": "https://github.com/qgallouedec",
"followers_url": "https://api.github.com/users/qgallouedec/followers",
"following_url": "https://api.github.com/users/qgallouedec/following{/other_user}",
"gists_url": "https://api.github.com/users/qgallouedec/gists{/gist_id}",
"starred_url": "https://api.github.com/users/qgallouedec/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/qgallouedec/subscriptions",
"organizations_url": "https://api.github.com/users/qgallouedec/orgs",
"repos_url": "https://api.github.com/users/qgallouedec/repos",
"events_url": "https://api.github.com/users/qgallouedec/events{/privacy}",
"received_events_url": "https://api.github.com/users/qgallouedec/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T19:14:45 | 2025-10-27T23:13:27 | 2025-10-02T10:37:55 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40950",
"html_url": "https://github.com/huggingface/transformers/pull/40950",
"diff_url": "https://github.com/huggingface/transformers/pull/40950.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40950.patch",
"merged_at": "2025-10-02T10:37:55"
} | This PR
- Deploys Trackio by default (instead of keeping local). See internal discussion here https://huggingface.slack.com/archives/C08SW1X12C8/p1758056496519289?thread_ts=1757706409.065919&cid=C08SW1X12C8 and
- Drops environment variables in favour of explicit args. See https://github.com/gradio-app/trackio/pull/212#pullrequestreview-3234300545
```diff
- import os
import numpy as np
from datasets import Dataset
from transformers import Trainer, AutoModelForCausalLM, TrainingArguments
- os.environ["TRACKIO_PROJECT"] = "my-awesome-project"
- os.environ["TRACKIO_SPACE_ID"] = "my-awesome-username/my-awesome-space"
# Create a fake dataset
data = np.random.randint(0, 1000, (8192, 64)).tolist()
dataset = Dataset.from_dict({"input_ids": data, "labels": data})
# Train a model using the Trainer API
trainer = Trainer(
model=AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B"),
- args=TrainingArguments(run_name="Qwen3-0.6B-training", report_to="trackio"),
+ args=TrainingArguments(run_name="Qwen3-0.6B-training", report_to="trackio", project="my-awesome-project", trackio_space_id="my-awesome-username/my-awesome-space"),
train_dataset=dataset,
)
trainer.train()
``` | {
"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/40950/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/40950/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40949 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40949/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40949/comments | https://api.github.com/repos/huggingface/transformers/issues/40949/events | https://github.com/huggingface/transformers/pull/40949 | 3,427,413,909 | PR_kwDOCUB6oc6pIprV | 40,949 | Update expected values for some `test_speculative_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 | [] | 2025-09-17T18:47:30 | 2025-09-17T18:56:40 | 2025-09-17T18:50:38 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40949",
"html_url": "https://github.com/huggingface/transformers/pull/40949",
"diff_url": "https://github.com/huggingface/transformers/pull/40949.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40949.patch",
"merged_at": "2025-09-17T18:50:38"
} | # What does this PR do?
These tests are failing after #40657.
As discussed offline [here](https://huggingface.slack.com/archives/C01NE71C4F7/p1758117410373159?thread_ts=1758031736.841469&cid=C01NE71C4F7), it's expected
TL:DR
> The logits are updated due to changes in assistant temperature, as we move from implicit T=1 to explicit calibrated defaults.
Details
> turns out the PR inadvertently changed an implicit algorithmic bias: candidate_generator (the assistant model) was getting logits_processor while the decoding method (main model) was getting prepared_logits_processor. This meant that the assistant was running with T=1 while the main model was using lower temp.
We investigated and its good for speculation to have a hotter assistant model (so it was a good bug that we were not applying the lower temp to the assistant),
But it should be explicitly set and not a hidden argument forwarding consequence. So its correct as is for now, and in the future we will bring better defaults for assisted generation, after benchmarking a little more.
<img width="904" height="575" alt="image" src="https://github.com/user-attachments/assets/b4e329ea-9b23-4629-8bba-2d1515a73289" />
| {
"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/40949/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/40949/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40948 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40948/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40948/comments | https://api.github.com/repos/huggingface/transformers/issues/40948/events | https://github.com/huggingface/transformers/pull/40948 | 3,427,385,878 | PR_kwDOCUB6oc6pIjYp | 40,948 | Allow private Space id for Trackio | {
"login": "qgallouedec",
"id": 45557362,
"node_id": "MDQ6VXNlcjQ1NTU3MzYy",
"avatar_url": "https://avatars.githubusercontent.com/u/45557362?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/qgallouedec",
"html_url": "https://github.com/qgallouedec",
"followers_url": "https://api.github.com/users/qgallouedec/followers",
"following_url": "https://api.github.com/users/qgallouedec/following{/other_user}",
"gists_url": "https://api.github.com/users/qgallouedec/gists{/gist_id}",
"starred_url": "https://api.github.com/users/qgallouedec/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/qgallouedec/subscriptions",
"organizations_url": "https://api.github.com/users/qgallouedec/orgs",
"repos_url": "https://api.github.com/users/qgallouedec/repos",
"events_url": "https://api.github.com/users/qgallouedec/events{/privacy}",
"received_events_url": "https://api.github.com/users/qgallouedec/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T18:39:46 | 2025-10-02T10:38:28 | 2025-10-02T10:38:25 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40948",
"html_url": "https://github.com/huggingface/transformers/pull/40948",
"diff_url": "https://github.com/huggingface/transformers/pull/40948.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40948.patch",
"merged_at": "2025-10-02T10:38:25"
} | requires https://github.com/gradio-app/trackio/pull/236
| {
"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/40948/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/40948/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40947 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40947/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40947/comments | https://api.github.com/repos/huggingface/transformers/issues/40947/events | https://github.com/huggingface/transformers/pull/40947 | 3,427,006,084 | PR_kwDOCUB6oc6pHRSZ | 40,947 | Fix `Glm4vModelTest::test_eager_matches_fa2_generate` | {
"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 | [] | 2025-09-17T16:31:31 | 2025-09-17T17:54:01 | 2025-09-17T17:54:00 | COLLABORATOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40947",
"html_url": "https://github.com/huggingface/transformers/pull/40947",
"diff_url": "https://github.com/huggingface/transformers/pull/40947.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40947.patch",
"merged_at": "2025-09-17T17:54:00"
} | # What does this PR do?
This test is failing on GPU due to
> FAILED tests/models/glm4v/test_modeling_glm4v.py::Glm4vModelTest::test_eager_matches_fa2_generate - RuntimeError: cu_seqlens_q must be on CUDA
see [here](https://github.com/huggingface/transformers/actions/runs/17784986682/job/50551078682)
We need to change from
> "image_grid_thw": torch.tensor([[1, patches_per_side, patches_per_side]] * self.batch_size),
to
> "image_grid_thw": torch.tensor([[1, patches_per_side, patches_per_side]] * self.batch_size, device=torch_device),it
| {
"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/40947/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/40947/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40946 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40946/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40946/comments | https://api.github.com/repos/huggingface/transformers/issues/40946/events | https://github.com/huggingface/transformers/issues/40946 | 3,426,997,724 | I_kwDOCUB6oc7MQ9Xc | 40,946 | Docs (zh-hans): Revert certain terms to English for clarity and professional consistency | {
"login": "Ri-Nai",
"id": 72967260,
"node_id": "MDQ6VXNlcjcyOTY3MjYw",
"avatar_url": "https://avatars.githubusercontent.com/u/72967260?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Ri-Nai",
"html_url": "https://github.com/Ri-Nai",
"followers_url": "https://api.github.com/users/Ri-Nai/followers",
"following_url": "https://api.github.com/users/Ri-Nai/following{/other_user}",
"gists_url": "https://api.github.com/users/Ri-Nai/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Ri-Nai/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Ri-Nai/subscriptions",
"organizations_url": "https://api.github.com/users/Ri-Nai/orgs",
"repos_url": "https://api.github.com/users/Ri-Nai/repos",
"events_url": "https://api.github.com/users/Ri-Nai/events{/privacy}",
"received_events_url": "https://api.github.com/users/Ri-Nai/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T16:28:40 | 2025-10-26T08:02:17 | 2025-10-26T08:02:17 | CONTRIBUTOR | null | null | null | null | Hello Hugging Face Team,
First of all, thank you for the great work on the `transformers` library and for providing multi-language documentation.
I am writing to suggest some improvements for the Simplified Chinese README file (`i18n/README_zh-hans.md`). As a Chinese-speaking user, I believe that reverting a few specific translated terms back to their original English form would significantly improve the document's professionalism and readability for the target audience (Chinese developers).
**The Core Problem:**
Certain terms are either translated into an overly informal nickname or into a Chinese phrase that is less common in a technical context than the English original. This can lead to ambiguity and a less professional tone.
**Specific Suggestions:**
I propose the following changes:
| Current Translation (原文) | Suggested Change (建议修改为) | Justification (理由) |
| :--- | :--- | :--- |
| `抱抱脸` | `Hugging Face` | While "抱抱脸" is a cute and friendly nickname, Chinese developers almost exclusively refer to the company and project as "Hugging Face". Using the official English name in technical documentation is more professional and aligns with community usage. |
| `流水线` (or other translations) | `pipeline` | The term `pipeline` is a core concept in the library. Chinese developers are very familiar with the English term. Translating it to "流水线" (assembly line) can sometimes be awkward or less precise than simply using the original term `pipeline`, which is common practice for key technical terms in Chinese documentation. |
**Why this is an improvement:**
1. **Clarity:** Using the original English for core concepts and names avoids any potential translation ambiguity.
2. **Professionalism:** It aligns the documentation with how these terms are actually used in technical discussions, tutorials, and code within the Chinese developer community.
3. **Consistency:** It maintains consistency with the code itself, where developers will encounter `pipeline`, not `流水线`.
I would be happy to open a Pull Request to apply these changes if this proposal is accepted.
Thank you for your consideration | {
"login": "github-actions[bot]",
"id": 41898282,
"node_id": "MDM6Qm90NDE4OTgyODI=",
"avatar_url": "https://avatars.githubusercontent.com/in/15368?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/github-actions%5Bbot%5D",
"html_url": "https://github.com/apps/github-actions",
"followers_url": "https://api.github.com/users/github-actions%5Bbot%5D/followers",
"following_url": "https://api.github.com/users/github-actions%5Bbot%5D/following{/other_user}",
"gists_url": "https://api.github.com/users/github-actions%5Bbot%5D/gists{/gist_id}",
"starred_url": "https://api.github.com/users/github-actions%5Bbot%5D/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/github-actions%5Bbot%5D/subscriptions",
"organizations_url": "https://api.github.com/users/github-actions%5Bbot%5D/orgs",
"repos_url": "https://api.github.com/users/github-actions%5Bbot%5D/repos",
"events_url": "https://api.github.com/users/github-actions%5Bbot%5D/events{/privacy}",
"received_events_url": "https://api.github.com/users/github-actions%5Bbot%5D/received_events",
"type": "Bot",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40946/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/40946/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/40945 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40945/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40945/comments | https://api.github.com/repos/huggingface/transformers/issues/40945/events | https://github.com/huggingface/transformers/pull/40945 | 3,426,983,086 | PR_kwDOCUB6oc6pHMQT | 40,945 | [tests] Really use small models in all fast 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 | [] | 2025-09-17T16:24:00 | 2025-09-18T17:51:09 | 2025-09-18T13:24:13 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40945",
"html_url": "https://github.com/huggingface/transformers/pull/40945",
"diff_url": "https://github.com/huggingface/transformers/pull/40945.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40945.patch",
"merged_at": "2025-09-18T13:24:13"
} | Use small models in all tests. Only `timm_backbone` cannot enforce it, because it cannot propagate kwargs. ALL other models should never override this tests, otherwise the CI becomes super slow.
cc @ydshieh fir viz! | {
"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/40945/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/40945/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40944 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40944/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40944/comments | https://api.github.com/repos/huggingface/transformers/issues/40944/events | https://github.com/huggingface/transformers/pull/40944 | 3,426,969,597 | PR_kwDOCUB6oc6pHJUw | 40,944 | add more activation kernels, follow up | {
"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 | [] | 2025-09-17T16:19:53 | 2025-10-02T06:45:07 | 2025-10-02T06:45:05 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40944",
"html_url": "https://github.com/huggingface/transformers/pull/40944",
"diff_url": "https://github.com/huggingface/transformers/pull/40944.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40944.patch",
"merged_at": "2025-10-02T06:45:05"
} | # What does this PR do?
Follow up to add more activation kernels to https://github.com/huggingface/transformers/pull/40890
<img width="1918" height="756" alt="image" src="https://github.com/user-attachments/assets/51acd3ae-5328-45bc-8533-9cf275e8a9ab" />
| {
"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/40944/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/40944/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40943 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40943/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40943/comments | https://api.github.com/repos/huggingface/transformers/issues/40943/events | https://github.com/huggingface/transformers/pull/40943 | 3,426,965,946 | PR_kwDOCUB6oc6pHIjA | 40,943 | Docs(zh-hans): Refine wording for professionalism in README | {
"login": "Ri-Nai",
"id": 72967260,
"node_id": "MDQ6VXNlcjcyOTY3MjYw",
"avatar_url": "https://avatars.githubusercontent.com/u/72967260?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/Ri-Nai",
"html_url": "https://github.com/Ri-Nai",
"followers_url": "https://api.github.com/users/Ri-Nai/followers",
"following_url": "https://api.github.com/users/Ri-Nai/following{/other_user}",
"gists_url": "https://api.github.com/users/Ri-Nai/gists{/gist_id}",
"starred_url": "https://api.github.com/users/Ri-Nai/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/Ri-Nai/subscriptions",
"organizations_url": "https://api.github.com/users/Ri-Nai/orgs",
"repos_url": "https://api.github.com/users/Ri-Nai/repos",
"events_url": "https://api.github.com/users/Ri-Nai/events{/privacy}",
"received_events_url": "https://api.github.com/users/Ri-Nai/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T16:18:34 | 2025-10-20T15:39:49 | 2025-10-20T15:39:49 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40943",
"html_url": "https://github.com/huggingface/transformers/pull/40943",
"diff_url": "https://github.com/huggingface/transformers/pull/40943.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40943.patch",
"merged_at": "2025-10-20T15:39:49"
} | # What does this PR do?
This PR improves the localization quality of the Simplified Chinese README (`README_zh-hans.md`) by replacing a colloquial term with more professional vocabulary.
Specifically, the informal term `魔改` (mó gǎi), which is a slang for "heavy modification," has been replaced with the standard and formal term `修改` (xiū gǎi), meaning "to modify."
This subtle but important change enhances the document's professionalism and ensures the language is accessible and appropriate for a broad technical audience.
<!-- Remove if not applicable -->
<!-- This is a minor documentation improvement, so no associated issue is required. -->
## 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).
- [ ] 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.
@stevhliu | {
"login": "stevhliu",
"id": 59462357,
"node_id": "MDQ6VXNlcjU5NDYyMzU3",
"avatar_url": "https://avatars.githubusercontent.com/u/59462357?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/stevhliu",
"html_url": "https://github.com/stevhliu",
"followers_url": "https://api.github.com/users/stevhliu/followers",
"following_url": "https://api.github.com/users/stevhliu/following{/other_user}",
"gists_url": "https://api.github.com/users/stevhliu/gists{/gist_id}",
"starred_url": "https://api.github.com/users/stevhliu/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/stevhliu/subscriptions",
"organizations_url": "https://api.github.com/users/stevhliu/orgs",
"repos_url": "https://api.github.com/users/stevhliu/repos",
"events_url": "https://api.github.com/users/stevhliu/events{/privacy}",
"received_events_url": "https://api.github.com/users/stevhliu/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40943/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/40943/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40942 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40942/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40942/comments | https://api.github.com/repos/huggingface/transformers/issues/40942/events | https://github.com/huggingface/transformers/pull/40942 | 3,426,891,778 | PR_kwDOCUB6oc6pG4tT | 40,942 | Raise error instead of warning when using meta device in from_pretrained | {
"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 | [] | 2025-09-17T15:54:41 | 2025-09-17T16:23:39 | 2025-09-17T16:23:37 | MEMBER | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40942",
"html_url": "https://github.com/huggingface/transformers/pull/40942",
"diff_url": "https://github.com/huggingface/transformers/pull/40942.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40942.patch",
"merged_at": "2025-09-17T16:23:37"
} | # 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/40942/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/40942/timeline | null | null | null | null | true | true |
https://api.github.com/repos/huggingface/transformers/issues/40941 | https://api.github.com/repos/huggingface/transformers | https://api.github.com/repos/huggingface/transformers/issues/40941/labels{/name} | https://api.github.com/repos/huggingface/transformers/issues/40941/comments | https://api.github.com/repos/huggingface/transformers/issues/40941/events | https://github.com/huggingface/transformers/pull/40941 | 3,426,717,786 | PR_kwDOCUB6oc6pGSlq | 40,941 | docs(i18n): Correct the descriptive text in the README_zh-hans.md | {
"login": "lilin-1",
"id": 177207022,
"node_id": "U_kgDOCo_27g",
"avatar_url": "https://avatars.githubusercontent.com/u/177207022?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/lilin-1",
"html_url": "https://github.com/lilin-1",
"followers_url": "https://api.github.com/users/lilin-1/followers",
"following_url": "https://api.github.com/users/lilin-1/following{/other_user}",
"gists_url": "https://api.github.com/users/lilin-1/gists{/gist_id}",
"starred_url": "https://api.github.com/users/lilin-1/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/lilin-1/subscriptions",
"organizations_url": "https://api.github.com/users/lilin-1/orgs",
"repos_url": "https://api.github.com/users/lilin-1/repos",
"events_url": "https://api.github.com/users/lilin-1/events{/privacy}",
"received_events_url": "https://api.github.com/users/lilin-1/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | [] | closed | false | null | [] | null | [] | 2025-09-17T15:07:26 | 2025-09-17T15:48:38 | 2025-09-17T15:48:38 | CONTRIBUTOR | null | null | false | {
"url": "https://api.github.com/repos/huggingface/transformers/pulls/40941",
"html_url": "https://github.com/huggingface/transformers/pull/40941",
"diff_url": "https://github.com/huggingface/transformers/pull/40941.diff",
"patch_url": "https://github.com/huggingface/transformers/pull/40941.patch",
"merged_at": "2025-09-17T15:48:38"
} |
# What does this PR do?
This PR corrects an incomplete Simplified Chinese description in the documentation by adding the missing term **"函数库" (library)**. The change ensures consistency with the Traditional Chinese version and fixes grammatical/semantic ambiguities.
### Background & Motivation
The original Simplified Chinese text was incomplete, lacking a clear object to clarify the subject of the description. The corresponding **Traditional Chinese reference text** (which serves as the consistency baseline) includes the term `為 Jax、PyTorch 以及 TensorFlow 打造的先進自然語言處理函式庫` (where "函式庫" = Traditional Chinese for "library").
### Change Details
| Text Type | Original (Incomplete) | Corrected (Complete) |
|---------------------|-------------------------|------------------------|
| Simplified Chinese | `为 Jax、PyTorch 和 TensorFlow 打造的先进的自然语言处理` | `为 Jax、PyTorch 和 TensorFlow 打造的先进的自然语言处理函数库` |
### Key Improvements
1. **Cross-Chinese Variant Consistency**: Aligns Simplified Chinese with Traditional Chinese by matching the core term "函数库" (Simplified) ↔ "函式庫" (Traditional), eliminating translation discrepancies.
2. **Grammatical Completeness**: Fixes the original sentence’s lack of an object (the original phrase "打造的先进的自然语言处理" failed to specify *what* is "built" for the frameworks).
3. **Semantic Clarity**: Explicitly confirms the subject is an **NLP library** (not vague "NLP technology"), fully aligning with the original description’s intent.
No dependencies are required for this change.
<!-- Remove if not applicable -->
Fixes #N/A (No corresponding issue; this is a direct documentation accuracy improvement)
## 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.
*Note: Not applicable—this change addresses a direct documentation inconsistency found during review, with no prior discussion needed.*
- [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).
*Update made: Modified the Simplified Chinese README’s descriptive text to include "函数库".*
- [ ] Did you write any new necessary tests?
*Note: Not applicable—this is a documentation-only change that does not impact code logic or functionality.*
## 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 @ -->
@stevhliu
*(Per the template guidelines, tagged the documentation-focused reviewer for Chinese docs accuracy checks.)* | {
"login": "stevhliu",
"id": 59462357,
"node_id": "MDQ6VXNlcjU5NDYyMzU3",
"avatar_url": "https://avatars.githubusercontent.com/u/59462357?v=4",
"gravatar_id": "",
"url": "https://api.github.com/users/stevhliu",
"html_url": "https://github.com/stevhliu",
"followers_url": "https://api.github.com/users/stevhliu/followers",
"following_url": "https://api.github.com/users/stevhliu/following{/other_user}",
"gists_url": "https://api.github.com/users/stevhliu/gists{/gist_id}",
"starred_url": "https://api.github.com/users/stevhliu/starred{/owner}{/repo}",
"subscriptions_url": "https://api.github.com/users/stevhliu/subscriptions",
"organizations_url": "https://api.github.com/users/stevhliu/orgs",
"repos_url": "https://api.github.com/users/stevhliu/repos",
"events_url": "https://api.github.com/users/stevhliu/events{/privacy}",
"received_events_url": "https://api.github.com/users/stevhliu/received_events",
"type": "User",
"user_view_type": "public",
"site_admin": false
} | {
"url": "https://api.github.com/repos/huggingface/transformers/issues/40941/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/40941/timeline | null | null | null | null | true | true |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.