repo string | github_id int64 | github_node_id string | number int64 | html_url string | api_url string | title string | body string | state string | state_reason string | locked bool | comments_count int64 | labels list | assignees list | created_at string | updated_at string | closed_at string | author_association string | milestone_title string | snapshot_id string | extracted_at string | author_login string | author_id int64 | author_node_id string | author_type string | author_site_admin bool |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
huggingface/transformers | 4,026,192,695 | I_kwDOCUB6oc7v-tM3 | 44,457 | https://github.com/huggingface/transformers/issues/44457 | https://api.github.com/repos/huggingface/transformers/issues/44457 | LORA权重合并保存在本地之后重新加载出来，两个模型输出结果不一致 | ### System Info
Debian GNU/Linux 12 (bookworm)+5090-32G
```
absl-py 2.3.1
accelerate 1.12.0
aiohappyeyeballs 2.6.1
aiohttp 3.13.2
aiosignal 1.4.0
altgraph 0.17.5
annotated-doc ... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-05T06:40:39Z | 2026-04-12T08:13:59Z | 2026-04-12T08:13:58Z | NONE | null | 20260413T085906Z | 2026-04-13T08:59:06Z | fish-kong | 54,362,165 | MDQ6VXNlcjU0MzYyMTY1 | User | false |
huggingface/transformers | 4,026,426,872 | I_kwDOCUB6oc7v_mX4 | 44,458 | https://github.com/huggingface/transformers/issues/44458 | https://api.github.com/repos/huggingface/transformers/issues/44458 | Mllama compile failed after new attn mask | ### System Info
torch 2.10.0+cpu
regression PR: #42848
### Who can help?
@vasqu
### 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 (giv... | open | null | false | 4 | [
"bug"
] | [] | 2026-03-05T07:33:29Z | 2026-04-12T08:13:57Z | null | CONTRIBUTOR | null | 20260413T085906Z | 2026-04-13T08:59:06Z | jiqing-feng | 107,918,818 | U_kgDOBm614g | User | false |
huggingface/transformers | 4,027,812,864 | I_kwDOCUB6oc7wE4wA | 44,462 | https://github.com/huggingface/transformers/issues/44462 | https://api.github.com/repos/huggingface/transformers/issues/44462 | AutoTokenizer ignores tokenizer.json from the repository | ### System Info
- `transformers` version: 5.3.0
- Python version: 3.10.12
- Huggingface_hub version: 1.5.0
### Who can help?
@ArthurZucker and @itazap
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder (such as GL... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-05T12:04:31Z | 2026-03-20T12:37:12Z | 2026-03-20T12:37:12Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | apaniukov | 51,917,466 | MDQ6VXNlcjUxOTE3NDY2 | User | false |
huggingface/transformers | 4,028,028,902 | I_kwDOCUB6oc7wFtfm | 44,464 | https://github.com/huggingface/transformers/issues/44464 | https://api.github.com/repos/huggingface/transformers/issues/44464 | Chunked generation produces inconsistent outputs when using compiled forward | ### System Info
- `transformers` version: 4.57.6
- Platform: Linux-6.18.13-arch1-1-x86_64-with-glibc2.43
- Python version: 3.12.12
- Huggingface_hub version: 0.36.0
- Safetensors version: 0.7.0
- Accelerate version: 1.12.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelera... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-05T12:50:02Z | 2026-03-06T16:50:02Z | 2026-03-06T16:50:02Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | mrTsjolder | 3,738,082 | MDQ6VXNlcjM3MzgwODI= | User | false |
huggingface/transformers | 4,028,229,153 | I_kwDOCUB6oc7wGeYh | 44,466 | https://github.com/huggingface/transformers/issues/44466 | https://api.github.com/repos/huggingface/transformers/issues/44466 | [v5] Inconsistent serialization of `lm_head.weight` (tied weights?) depending on model device in v5/`main`, while v4.57 behaves correctly | ### System Info
```
- `transformers` version: 5.3.0.dev0
- Platform: Linux-6.8.0-100-generic-x86_64-with-glibc2.39
- Python version: 3.12.12
- Huggingface_hub version: 1.5.0
- Safetensors version: 0.7.0
- Accelerate version: 1.12.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version ... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-05T13:26:59Z | 2026-03-09T15:00:24Z | 2026-03-09T15:00:24Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | fxmarty-amd | 180,171,742 | U_kgDOCr0z3g | User | false |
huggingface/transformers | 4,029,975,044 | I_kwDOCUB6oc7wNIoE | 44,479 | https://github.com/huggingface/transformers/issues/44479 | https://api.github.com/repos/huggingface/transformers/issues/44479 | [`bug`] v5.3.0 video input regression for `qwen2_5_vl`, `qwen3_vl`, `qwen3_5`, and `qwen3_5_moe` | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Windows-10-10.0.26200-SP0
- Python version: 3.11.6
- Huggingface_hub version: 1.5.0
- Safetensors version: 0.6.2
- Accelerate version: 1.13.0.dev0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.10... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-05T18:47:53Z | 2026-03-10T09:57:36Z | 2026-03-10T09:57:35Z | MEMBER | null | 20260325T173244Z | 2026-03-25T17:32:44Z | tomaarsen | 37,621,491 | MDQ6VXNlcjM3NjIxNDkx | User | false |
huggingface/transformers | 4,031,866,225 | I_kwDOCUB6oc7wUWVx | 44,483 | https://github.com/huggingface/transformers/issues/44483 | https://api.github.com/repos/huggingface/transformers/issues/44483 | [Critical v5.3] /v1/chat/completions would not accept request as usual | ### System Info
Linux
### 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
Run `transfor... | closed | completed | false | 4 | [
"bug"
] | [
"LysandreJik"
] | 2026-03-06T02:39:54Z | 2026-03-16T05:26:19Z | 2026-03-16T05:26:18Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | zhangwei217245 | 346,451 | MDQ6VXNlcjM0NjQ1MQ== | User | false |
huggingface/transformers | 4,031,894,714 | I_kwDOCUB6oc7wUdS6 | 44,484 | https://github.com/huggingface/transformers/issues/44484 | https://api.github.com/repos/huggingface/transformers/issues/44484 | Why max_shard_size in PreTrainedModel.save_pretrained() is 50GB? | ### System Info
In old version like 4.57.1, max_shard_size in PreTrainedModel.save_pretrained() is 5GB, but in new version, max_shard_size is '50GB'. Is it normal?
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- [ ] My own modified scripts
### Tasks
- [ ] An officially suppor... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-06T02:51:48Z | 2026-03-06T13:34:27Z | 2026-03-06T13:34:27Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | Silencezong | 92,312,888 | U_kgDOBYCVOA | User | false |
huggingface/transformers | 4,032,430,401 | I_kwDOCUB6oc7wWgFB | 44,485 | https://github.com/huggingface/transformers/issues/44485 | https://api.github.com/repos/huggingface/transformers/issues/44485 | [Bug/Discusion] GLM-5 RoPE Implementation | ### System Info
https://huggingface.co/zai-org/GLM-5/blob/main/config.json#L45
hi! I see that the rope setting here is rope_interleave true.
However, looking at the transformers implementation, the logic here is false, refer to these
https://github.com/huggingface/transformers/blob/d5e555a632682555332c3c8e938461efd... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-06T06:04:18Z | 2026-04-13T08:37:44Z | 2026-04-13T08:37:44Z | CONTRIBUTOR | null | 20260413T085906Z | 2026-04-13T08:59:06Z | Jintao-Huang | 45,290,347 | MDQ6VXNlcjQ1MjkwMzQ3 | User | false |
huggingface/transformers | 4,032,643,658 | I_kwDOCUB6oc7wXUJK | 44,486 | https://github.com/huggingface/transformers/issues/44486 | https://api.github.com/repos/huggingface/transformers/issues/44486 | KubeflowCallback: Native progress reporting for Kubernetes-based Kubeflow training | ### Feature request
Add a `KubeflowCallback` to enable automatic progress and metrics reporting for training jobs running on [Kubeflow Trainer](https://github.com/kubeflow/trainer), the Kubernetes-native platform for distributed AI/ML training.
**Context:** This is part of a coordinated effort with the Kubeflow comm... | closed | completed | false | 1 | [
"Feature request"
] | [] | 2026-03-06T07:07:19Z | 2026-03-18T14:58:25Z | 2026-03-18T14:58:25Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | abhijeet-dhumal | 84,722,973 | MDQ6VXNlcjg0NzIyOTcz | User | false |
huggingface/transformers | 4,032,973,273 | I_kwDOCUB6oc7wYknZ | 44,488 | https://github.com/huggingface/transformers/issues/44488 | https://api.github.com/repos/huggingface/transformers/issues/44488 | Current version also does not load "cjvt/sleng-bert" | ### System Info
broken config:
Python 3.13.5
tokenizers 0.22.2
transformers 5.2.0
torch 2.7.1+cu118
working config:
Python 3.13.5
tokenizers 0.22.1
transformers 4.57.1
torch 2.8.0+cu129
### Who can help?
@ArthurZucker @Cyrilvallez
### Information
- [x] The official example scripts
- [ ] My own m... | closed | completed | false | 14 | [
"bug"
] | [] | 2026-03-06T08:36:44Z | 2026-03-24T10:05:50Z | 2026-03-23T08:52:03Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | AngledLuffa | 3,411,033 | MDQ6VXNlcjM0MTEwMzM= | User | false |
huggingface/transformers | 4,033,842,729 | I_kwDOCUB6oc7wb44p | 44,492 | https://github.com/huggingface/transformers/issues/44492 | https://api.github.com/repos/huggingface/transformers/issues/44492 | Typo in Cache strategies | ### System Info
I’m sorry if I am writing this issue in a wrong subsection but your chatbot let me here. The document Cache Strategies, https://huggingface.co/docs/transformers/kv_cache , says that the JIT maximizes latency while it should “minimizes” the latency .
I basically just greaduated, start of my career so p... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-06T12:00:54Z | 2026-03-06T20:05:38Z | 2026-03-06T20:05:38Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | aryan221102 | 160,477,158 | U_kgDOCZCv5g | User | false |
huggingface/transformers | 4,033,865,199 | I_kwDOCUB6oc7wb-Xv | 44,493 | https://github.com/huggingface/transformers/issues/44493 | https://api.github.com/repos/huggingface/transformers/issues/44493 | Many models started showing UNEXPECTED Key with position id | ### System Info
- `transformers` version: 5.0.0
- Platform: Linux-6.6.113+-x86_64-with-glibc2.35
- Python version: 3.12.12
- Huggingface_hub version: 1.5.0
- Safetensors version: 0.7.0
- Accelerate version: 1.12.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-06T12:06:44Z | 2026-03-18T18:30:50Z | 2026-03-18T18:30:50Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | weathon | 41,298,844 | MDQ6VXNlcjQxMjk4ODQ0 | User | false |
huggingface/transformers | 4,034,746,383 | I_kwDOCUB6oc7wfVgP | 44,496 | https://github.com/huggingface/transformers/issues/44496 | https://api.github.com/repos/huggingface/transformers/issues/44496 | ValueError: Unrecognized model in allenai/Olmo-Hybrid-Instruct-SFT-7B. Should have a `model_type` key in its config.json. | ### System Info
- `transformers` version: 5.3.0.dev0 (main)
- Platform: macOS-15.7.4-arm64-arm-64bit-Mach-O
- Python version: 3.13.2
- Huggingface_hub version: 1.3.1
- Safetensors version: 0.5.3
- Accelerate version: 1.12.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (acceler... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-03-06T15:17:24Z | 2026-03-08T23:07:07Z | 2026-03-08T23:07:06Z | MEMBER | null | 20260325T173244Z | 2026-03-25T17:32:44Z | xenova | 26,504,141 | MDQ6VXNlcjI2NTA0MTQx | User | false |
huggingface/transformers | 4,036,647,258 | I_kwDOCUB6oc7wmlla | 44,509 | https://github.com/huggingface/transformers/issues/44509 | https://api.github.com/repos/huggingface/transformers/issues/44509 | Docs still mention text2text-generation / summarization / translation pipeline tasks which were removed in v5 | ### System Info
Transformers version: 5.x
The issue is related to the documentation rather than a specific runtime environment.
### Who can help?
@Rocketknight1 @stevhliu
### Information
- [x] The official example scripts
- [ ] My own modified scripts
### Tasks
- [x] An officially supported task in the `exampl... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-03-06T23:34:44Z | 2026-03-09T19:00:15Z | 2026-03-09T19:00:15Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | math-hiyoko | 56,009,584 | MDQ6VXNlcjU2MDA5NTg0 | User | false |
huggingface/transformers | 4,038,808,878 | I_kwDOCUB6oc7wu1Uu | 44,512 | https://github.com/huggingface/transformers/issues/44512 | https://api.github.com/repos/huggingface/transformers/issues/44512 | Docs still mention transformers run command which was removed in v5 | ### System Info
Transformers version: 5.x
The issue is related to the documentation rather than a specific runtime environment.
### Who can help?
@stevhliu
### Information
- [x] The official example scripts
- [ ] My own modified scripts
### Tasks
- [x] An officially supported task in the `examples` folder (suc... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-03-07T16:10:16Z | 2026-03-09T15:37:17Z | 2026-03-09T15:37:17Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | math-hiyoko | 56,009,584 | MDQ6VXNlcjU2MDA5NTg0 | User | false |
huggingface/transformers | 4,038,940,374 | I_kwDOCUB6oc7wvVbW | 44,514 | https://github.com/huggingface/transformers/issues/44514 | https://api.github.com/repos/huggingface/transformers/issues/44514 | `Qwen2_5_VLProcessor.apply_chat_template` crashes on batched input when `padding=False` | ### System Info
transformers 5.3.0
### 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
... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-07T17:03:20Z | 2026-03-25T11:33:47Z | 2026-03-25T11:33:47Z | MEMBER | null | 20260325T173244Z | 2026-03-25T17:32:44Z | qgallouedec | 45,557,362 | MDQ6VXNlcjQ1NTU3MzYy | User | false |
huggingface/transformers | 4,040,512,605 | I_kwDOCUB6oc7w1VRd | 44,521 | https://github.com/huggingface/transformers/issues/44521 | https://api.github.com/repos/huggingface/transformers/issues/44521 | apply_chat_template returns all-zero assistant_masks for multimodal inputs | ### System Info
transformers==5.3.0
### Who can help?
@ArthurZucker and @itazap
### 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)
### R... | open | null | false | 8 | [
"bug"
] | [] | 2026-03-08T05:03:12Z | 2026-04-18T08:11:39Z | null | NONE | null | 20260418T090534Z | 2026-04-18T09:05:34Z | renhouxing | 36,196,749 | MDQ6VXNlcjM2MTk2NzQ5 | User | false |
huggingface/transformers | 4,041,937,165 | I_kwDOCUB6oc7w6xEN | 44,530 | https://github.com/huggingface/transformers/issues/44530 | https://api.github.com/repos/huggingface/transformers/issues/44530 | [CB] PagedAttentionCache crashes with "Invalid group type: linear_attention" on Qwen3.5 models | ### System Info
- `transformers` version: 5.2.0
- Platform: Windows-11-10.0.26200-SP0
- Python version: 3.13.3
- Huggingface_hub version: 1.5.0
- Safetensors version: 0.5.3
- Accelerate version: 1.12.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.10.0+cu128 (... | closed | completed | false | 7 | [
"bug",
"Code agent slop"
] | [] | 2026-03-08T18:49:55Z | 2026-03-13T11:58:36Z | 2026-03-10T22:42:08Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | mxchampagne | 96,957,776 | U_kgDOBcd1UA | User | false |
huggingface/transformers | 4,043,153,614 | I_kwDOCUB6oc7w_aDO | 44,534 | https://github.com/huggingface/transformers/issues/44534 | https://api.github.com/repos/huggingface/transformers/issues/44534 | Transformers v5 fills non-persistent buffers with junk | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux-6.17.0-14-generic-x86_64-with-glibc2.39
- Python version: 3.12.3
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.7.0
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accel... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-09T03:43:59Z | 2026-03-09T12:56:33Z | 2026-03-09T12:56:33Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | umarbutler | 8,473,183 | MDQ6VXNlcjg0NzMxODM= | User | false |
huggingface/transformers | 4,044,272,162 | I_kwDOCUB6oc7xDrIi | 44,537 | https://github.com/huggingface/transformers/issues/44537 | https://api.github.com/repos/huggingface/transformers/issues/44537 | About check_model_inputs after version 5.2.0, Where is it!? | Hei bros, I found the func transformers.utils.generic.check_model_inputs in version <=5.1.0 turn to transformers.utils.generic.merge_with_config_defaults in version >=5.2.0, but I did not found any notes about it, or I am a blind person......TAT | closed | completed | false | 2 | [] | [] | 2026-03-09T08:56:37Z | 2026-03-10T02:15:57Z | 2026-03-10T02:15:57Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | QYQTexas | 169,536,866 | U_kgDOChrtYg | User | false |
huggingface/transformers | 4,044,495,369 | I_kwDOCUB6oc7xEhoJ | 44,541 | https://github.com/huggingface/transformers/issues/44541 | https://api.github.com/repos/huggingface/transformers/issues/44541 | Can not deploy SFT Qwen3.5-9B model | ### System Info
I SFT Qwen3.5-9B model with transformers==5.2.0
### Reproduction
When I try to deploy my model with vllm==0.17.0, it reports:
TypeError: Invalid type of HuggingFace config. Expected type: <class 'vllm.transformers_utils.configs.qwen3_5.Qwen3_5Config'>, but found type: <class 'transformers.models.qwe... | closed | completed | false | 17 | [
"bug"
] | [] | 2026-03-09T09:39:19Z | 2026-04-19T01:15:17Z | 2026-04-10T09:27:56Z | NONE | null | 20260419T020535Z | 2026-04-19T02:05:35Z | zouYC2021 | 66,997,535 | MDQ6VXNlcjY2OTk3NTM1 | User | false |
huggingface/transformers | 4,045,354,786 | I_kwDOCUB6oc7xHzci | 44,545 | https://github.com/huggingface/transformers/issues/44545 | https://api.github.com/repos/huggingface/transformers/issues/44545 | Qwen2_5_VLProcessor.apply_chat_template crashes on batched input when padding=False | ## Bug Description
`Qwen2_5_VLProcessor.apply_chat_template` raises `ValueError: setting an array element with a sequence` when processing a batch of ≥2 conversations that include images, under the default `padding=False` setting.
**Root cause:** `mm_token_type_ids` was built by calling `np.array(text_inputs["input_i... | closed | completed | false | 2 | [] | [] | 2026-03-09T12:37:17Z | 2026-03-25T11:33:46Z | 2026-03-25T11:33:46Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | Anakintano | 126,348,043 | U_kgDOB4frCw | User | false |
huggingface/transformers | 4,048,763,527 | I_kwDOCUB6oc7xUzqH | 44,554 | https://github.com/huggingface/transformers/issues/44554 | https://api.github.com/repos/huggingface/transformers/issues/44554 | [MPS] Upstream correctness issue in attention when value head dim differs from query | ### System Info
There is a correctness issue with attention in PyTorch when using the MPS backend with value head dims different from the query head (see https://github.com/pytorch/pytorch/issues/176767).
Consider the following reproducer
```python
import torch
import torch.nn.functional as F
q = torch.rand(1, 1, 8... | open | reopened | false | 10 | [
"WIP",
"bug"
] | [] | 2026-03-10T01:13:32Z | 2026-04-17T11:56:26Z | null | CONTRIBUTOR | null | 20260417T180542Z | 2026-04-17T18:05:42Z | hvaara | 1,535,968 | MDQ6VXNlcjE1MzU5Njg= | User | false |
huggingface/transformers | 4,049,397,875 | I_kwDOCUB6oc7xXOhz | 44,556 | https://github.com/huggingface/transformers/issues/44556 | https://api.github.com/repos/huggingface/transformers/issues/44556 | Loading checkpoint trained on v4.57 cannot be reload after upgraded to v5.2 & v5.3 | ### System Info
I have trained models using Qwen3 using v4.57 but the ckpt loading will hang forever after Loading weights: 100%|██████████| 708/708 [00:09<00:00, 78.04it/s] .
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [ ] An offic... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-10T04:56:49Z | 2026-03-11T20:27:04Z | 2026-03-11T20:27:04Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | ihungalexhsu | 8,895,697 | MDQ6VXNlcjg4OTU2OTc= | User | false |
huggingface/transformers | 4,050,043,835 | I_kwDOCUB6oc7xZsO7 | 44,559 | https://github.com/huggingface/transformers/issues/44559 | https://api.github.com/repos/huggingface/transformers/issues/44559 | flash-attn-4 (flash_attn.cute) is not supported by attn_implementation="flash_attention_2" | ### Feature request
# Support `flash-attn-4` (`flash_attn.cute`) in Transformers attention backend selection
## System Info
- `transformers==5.3.0`
- `torch==2.10.0+cu128`
- `flash-attn-4==4.0.0b4`
- `accelerate==1.13.0`
- `trl==0.29.0`
- `peft==0.18.0`
- `deepspeed==0.18.7`
- `tokenizers==0.22.2`
- `huggingface_hub=... | closed | completed | false | 2 | [
"Feature request"
] | [] | 2026-03-10T07:47:09Z | 2026-03-11T06:53:38Z | 2026-03-11T06:53:38Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | DimensionSTP | 65,501,090 | MDQ6VXNlcjY1NTAxMDkw | User | false |
huggingface/transformers | 4,050,236,929 | I_kwDOCUB6oc7xabYB | 44,560 | https://github.com/huggingface/transformers/issues/44560 | https://api.github.com/repos/huggingface/transformers/issues/44560 | Qwen3-vl-embedding Video Error "StopIteration" in transformers 5.3.0 | ### System Info
transformers version: 5.3.0
Platform: Windows11, WSL2, uv, vscode
Python 3.12.13 (main, Mar 3 2026, 14:59:34) [Clang 21.1.4 ] on linux
### Who can help?
@zucchini-nlp
### Information
- [x] The official example scripts
- [ ] My own modified scripts
### Tasks
- [ ] An officially supported task in... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-10T08:30:00Z | 2026-03-10T10:15:25Z | 2026-03-10T09:57:21Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | QYQTexas | 169,536,866 | U_kgDOChrtYg | User | false |
huggingface/transformers | 4,050,489,633 | I_kwDOCUB6oc7xbZEh | 44,561 | https://github.com/huggingface/transformers/issues/44561 | https://api.github.com/repos/huggingface/transformers/issues/44561 | Removal of `is_torch_fx_available` in v5.0 breaks `trust_remote_code` models | ### System Info
- `transformers` version: 5.0.0
- PyTorch version: 2.10.0
- Python version: 3.12
### Who can help?
@ArthurZucker @Rocketknight1
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder
- [x] My own task ... | closed | not_planned | false | 1 | [] | [] | 2026-03-10T09:22:21Z | 2026-03-25T17:48:04Z | 2026-03-25T17:48:04Z | NONE | null | 20260325T200011Z | 2026-03-25T20:00:11Z | janbernloehr | 1,050,099 | MDQ6VXNlcjEwNTAwOTk= | User | false |
huggingface/transformers | 4,051,220,122 | I_kwDOCUB6oc7xeLaa | 44,568 | https://github.com/huggingface/transformers/issues/44568 | https://api.github.com/repos/huggingface/transformers/issues/44568 | [BUG] add_special_tokens=True doesn't add BOS/EOS tokens for microsoft/mdeberta-v3-base tokenizer in transformers >=5.0 | ### System Info
## Version Details
- Working version: transformers==4.48.0
- Broken versions: transformers==5.0.0, 5.1.0, 5.2.0, 5.3.0
## Environment
- transformers: 5.2.0
- tokenizers: 0.22.2
- Python: 3.12
- Platform: Linux
### Who can help?
@ArthurZucker and @itazap
### Information
- [ ] The official example sc... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-03-10T11:43:59Z | 2026-03-24T09:40:46Z | 2026-03-24T09:40:46Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | Abdullahaml1 | 25,988,048 | MDQ6VXNlcjI1OTg4MDQ4 | User | false |
huggingface/transformers | 4,052,196,712 | I_kwDOCUB6oc7xh51o | 44,572 | https://github.com/huggingface/transformers/issues/44572 | https://api.github.com/repos/huggingface/transformers/issues/44572 | <spam> | <spam> | closed | completed | false | 0 | [] | [] | 2026-03-10T14:28:30Z | 2026-03-11T13:20:33Z | 2026-03-11T13:20:19Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | aiqing20230305-bot | 257,078,371 | U_kgDOD1K0Yw | User | false |
huggingface/transformers | 4,052,252,831 | I_kwDOCUB6oc7xiHif | 44,573 | https://github.com/huggingface/transformers/issues/44573 | https://api.github.com/repos/huggingface/transformers/issues/44573 | <spam> | <spam> | closed | completed | false | 0 | [] | [] | 2026-03-10T14:37:42Z | 2026-03-11T13:20:49Z | 2026-03-11T13:20:37Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | aiqing20230305-bot | 257,078,371 | U_kgDOD1K0Yw | User | false |
huggingface/transformers | 4,052,267,788 | I_kwDOCUB6oc7xiLMM | 44,574 | https://github.com/huggingface/transformers/issues/44574 | https://api.github.com/repos/huggingface/transformers/issues/44574 | Title: Clarification and consistency of "Transformers" terminology in README | <img width="1258" height="366" alt="Image" src="https://github.com/user-attachments/assets/b71a8eca-da64-4464-a28e-452dc21ca569" />
<img width="1434" height="399" alt="Image" src="https://github.com/user-attachments/assets/f324f04f-abbc-4bec-92a9-9ffde2f42034" />
While reading the **README as a new user**, I found th... | closed | completed | false | 2 | [] | [] | 2026-03-10T14:40:08Z | 2026-03-12T04:38:08Z | 2026-03-11T13:24:42Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | rabbierabbie | 161,629,097 | U_kgDOCaJDqQ | User | false |
huggingface/transformers | 4,055,271,864 | I_kwDOCUB6oc7xtom4 | 44,589 | https://github.com/huggingface/transformers/issues/44589 | https://api.github.com/repos/huggingface/transformers/issues/44589 | TypeError: couldn't find storage object Float8_e4m3fnStorage | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Linux-6.8.0-101-generic-x86_64-with-glibc2.35
- Python version: 3.12.13
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (acc... | closed | completed | false | 6 | [
"bug"
] | [] | 2026-03-11T02:16:53Z | 2026-03-20T13:12:32Z | 2026-03-20T13:12:32Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | xin3he | 83,260,933 | MDQ6VXNlcjgzMjYwOTMz | User | false |
huggingface/transformers | 4,057,458,915 | I_kwDOCUB6oc7x1-jj | 44,593 | https://github.com/huggingface/transformers/issues/44593 | https://api.github.com/repos/huggingface/transformers/issues/44593 | Support for sequence-level custom metrics with decoder-only models | ### Feature request
Hi Hugging Face team,
I’m trying to compute custom metrics at the sequence level for a decoder-only Transformer model, but I ran into an issue. The Seq2SeqTrainer class provides the predict_with_generate option, but it is primarily designed for encoder-decoder architectures. As a result, using it ... | open | null | false | 6 | [
"Feature request"
] | [] | 2026-03-11T10:51:47Z | 2026-03-13T11:57:16Z | null | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | l-k-11235 | 57,141,057 | MDQ6VXNlcjU3MTQxMDU3 | User | false |
huggingface/transformers | 4,060,145,187 | I_kwDOCUB6oc7yAOYj | 44,609 | https://github.com/huggingface/transformers/issues/44609 | https://api.github.com/repos/huggingface/transformers/issues/44609 | StaticSlidingWindowLayer triggers torch.compile/dynamo recompilations every decode step | ### Feature request
### Summary
Could there be added support for StaticSlidingWindowLayer to be fully compatible with torch.compile() by avoiding it's dynamic control flow?
When using a StaticCache with StaticSlidingWindowLayer (e.g. GPT-OSS, Mistral) and torch.compile(), the compiled graph recompiles on every deco... | closed | completed | false | 5 | [
"Feature request"
] | [] | 2026-03-11T19:15:20Z | 2026-03-12T13:03:20Z | 2026-03-12T13:01:42Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | jazpurTT | 200,671,331 | U_kgDOC_YAYw | User | false |
huggingface/transformers | 4,060,354,689 | I_kwDOCUB6oc7yBBiB | 44,610 | https://github.com/huggingface/transformers/issues/44610 | https://api.github.com/repos/huggingface/transformers/issues/44610 | [BUG] OmDet-Turbo processor produces 640px inputs but the model expects 224px | ### System Info
* `transformers` version: `5.0.0.dev0`
* Platform: `Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.39`
* Python version: `3.12.3`
* `huggingface_hub` version: `1.3.2`
* `safetensors` version: `0.7.0`
* `accelerate` version: `1.12.0`
* Accelerate config: `not installed`
* DeepSpeed version:... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-03-11T19:58:13Z | 2026-04-18T09:07:17Z | 2026-03-13T11:55:55Z | CONTRIBUTOR | null | 20260418T100536Z | 2026-04-18T10:05:36Z | harshaljanjani | 75,426,551 | MDQ6VXNlcjc1NDI2NTUx | User | false |
huggingface/transformers | 4,061,867,605 | I_kwDOCUB6oc7yGy5V | 44,617 | https://github.com/huggingface/transformers/issues/44617 | https://api.github.com/repos/huggingface/transformers/issues/44617 | Sam3Video: CUDA out of memory | ### System Info
transformers 5.3.0
Python 3.10.12
torch 2.4.0+cu124
Tracking multiple targets simultaneously, typically numbering in the dozens, results in out of memory.
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [ ] An official... | open | null | false | 3 | [
"bug"
] | [] | 2026-03-12T03:29:05Z | 2026-03-27T10:37:08Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | middleknight | 39,006,812 | MDQ6VXNlcjM5MDA2ODEy | User | false |
huggingface/transformers | 4,062,402,313 | I_kwDOCUB6oc7yI1cJ | 44,619 | https://github.com/huggingface/transformers/issues/44619 | https://api.github.com/repos/huggingface/transformers/issues/44619 | Plug model rule in development flow and extend it | This is a follow-up from https://github.com/huggingface/transformers/pull/44174
We're now plugging the tool into developer flow:
- add an opt-in github hook for checking the model
- automatically run `make check-model-rules` on PRs to generate reports
- add a ref to the CLI in the AI agents files
- fix models that ar... | open | null | false | 1 | [] | [
"tarekziade"
] | 2026-03-12T06:00:58Z | 2026-04-14T08:29:54Z | null | MEMBER | null | 20260414T122001Z | 2026-04-14T12:20:01Z | tarekziade | 250,019 | MDQ6VXNlcjI1MDAxOQ== | User | false |
huggingface/transformers | 4,063,277,021 | I_kwDOCUB6oc7yMK_d | 44,623 | https://github.com/huggingface/transformers/issues/44623 | https://api.github.com/repos/huggingface/transformers/issues/44623 | [BUG] processor.save_pretrained(...) missing files | ### System Info
transformers 4.57.6
<img width="688" height="176" alt="Image" src="https://github.com/user-attachments/assets/7106b5ce-6ef8-4bb6-ba7a-889821d02f8f" />
transformers 5.3.0
<img width="664" height="101" alt="Image" src="https://github.com/user-attachments/assets/b54d5186-280b-405e-b0bb-4caee99f2a11" /... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-12T09:20:10Z | 2026-03-13T12:04:23Z | 2026-03-13T12:04:23Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | Jintao-Huang | 45,290,347 | MDQ6VXNlcjQ1MjkwMzQ3 | User | false |
huggingface/transformers | 4,063,886,834 | I_kwDOCUB6oc7yOf3y | 44,625 | https://github.com/huggingface/transformers/issues/44625 | https://api.github.com/repos/huggingface/transformers/issues/44625 | Qwen3.5 `num_labels` not propagated from core config to text config | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Windows-10-10.0.26200-SP0
- Python version: 3.11.6
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.6.2
- Accelerate version: 1.13.0.dev0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.10... | open | null | false | 3 | [
"bug"
] | [] | 2026-03-12T11:09:38Z | 2026-04-08T12:22:10Z | null | MEMBER | null | 20260411T144729Z | 2026-04-11T14:47:29Z | tomaarsen | 37,621,491 | MDQ6VXNlcjM3NjIxNDkx | User | false |
huggingface/transformers | 4,065,597,475 | I_kwDOCUB6oc7yVBgj | 44,637 | https://github.com/huggingface/transformers/issues/44637 | https://api.github.com/repos/huggingface/transformers/issues/44637 | load_best_model_at_end reloads PEFT adapter weights onto CUDA and can OOM under low remaining GPU memory | ## System Info
- `transformers` version: local current checkout (5.3.0.dev0)
- Python: `3.12.12`
- PyTorch: `2.10.0+cu128`
- CUDA available: `True`
- CUDA device count: `8`
- `torchvision`: `0.25.0+cu128`
- `Pillow`: `12.1.1`
- PEFT: `0.18.1`
I can also provide the full `transformers env` output if needed.
## Who ca... | open | null | false | 7 | [] | [] | 2026-03-12T15:47:16Z | 2026-04-02T14:14:48Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | DogWala | 111,959,103 | U_kgDOBqxcPw | User | false |
huggingface/transformers | 4,067,983,074 | I_kwDOCUB6oc7yeH7i | 44,643 | https://github.com/huggingface/transformers/issues/44643 | https://api.github.com/repos/huggingface/transformers/issues/44643 | Qwen3.5 + `flash_attention_2` crashes: 3D M-RoPE position_ids leak to `_is_packed_sequence` | ## Qwen3.5 + `flash_attention_2` crashes: 3D M-RoPE position_ids leak to `_is_packed_sequence`
### System Info
- `transformers`: 5.3.0, PyTorch: 2.6.0+cu124, flash-attn: 2.8.3, Python: 3.10, Linux
### Reproduction
Fine-tuning `Qwen3.5-9B` with `attn_implementation="flash_attention_2"` crashes with `CUDA error: an i... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-13T00:11:25Z | 2026-03-13T11:21:05Z | 2026-03-13T11:21:05Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | ritwickchaudhry | 11,583,361 | MDQ6VXNlcjExNTgzMzYx | User | false |
huggingface/transformers | 4,069,534,970 | I_kwDOCUB6oc7ykCz6 | 44,655 | https://github.com/huggingface/transformers/issues/44655 | https://api.github.com/repos/huggingface/transformers/issues/44655 | Unable to save Pipeline objects with save_pretrained. | ### System Info
**transformers**: 5.3.0, **python**: 3.13.12, **OS**: macOS 15.6
### Who can help?
@Rocketknight1
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [x] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [ ] My own task or ... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-03-13T07:53:25Z | 2026-03-13T14:08:01Z | 2026-03-13T14:08:01Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | VishnuHaasan | 66,861,385 | MDQ6VXNlcjY2ODYxMzg1 | User | false |
huggingface/transformers | 4,070,986,225 | I_kwDOCUB6oc7yplHx | 44,661 | https://github.com/huggingface/transformers/issues/44661 | https://api.github.com/repos/huggingface/transformers/issues/44661 | `add-new-model-like` fails if model is inside the `TOKENIZER_MAPPING_NAMES` | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Linux-6.14.0-1013-nvidia-aarch64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.7.0
- Accelerate version: 1.14.0.dev0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version... | closed | completed | false | 7 | [
"bug"
] | [] | 2026-03-13T13:00:26Z | 2026-03-13T17:13:29Z | 2026-03-13T17:13:29Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | michalrzak | 66,417,283 | MDQ6VXNlcjY2NDE3Mjgz | User | false |
huggingface/transformers | 4,071,819,790 | I_kwDOCUB6oc7yswoO | 44,671 | https://github.com/huggingface/transformers/issues/44671 | https://api.github.com/repos/huggingface/transformers/issues/44671 | CamemBERT produces incorrect masked LM predictions in v5 | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux-6.6.113+-x86_64-with-glibc2.35
- Python version: 3.12.12
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.10... | closed | completed | false | 5 | [
"bug"
] | [] | 2026-03-13T15:30:48Z | 2026-03-23T10:47:50Z | 2026-03-23T10:47:50Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | math-hiyoko | 56,009,584 | MDQ6VXNlcjU2MDA5NTg0 | User | false |
huggingface/transformers | 4,072,494,799 | I_kwDOCUB6oc7yvVbP | 44,677 | https://github.com/huggingface/transformers/issues/44677 | https://api.github.com/repos/huggingface/transformers/issues/44677 | Add base_model_tp_plan to OlmoeConfig | ## Description
`OlmoeConfig` is missing a `base_model_tp_plan` class attribute, which means `from_pretrained(tp_plan="auto")` does not work for OLMoE models.
Other MoE models like Qwen3-MoE already have this. OLMoE needs its own plan with a key difference: `q_norm` and `k_norm` must use `"colwise"` (not `"replicated_... | closed | completed | false | 0 | [] | [] | 2026-03-13T17:37:56Z | 2026-03-26T13:58:59Z | 2026-03-26T13:58:59Z | MEMBER | null | 20260407T090028Z | 2026-04-07T09:00:28Z | dacorvo | 1,910,518 | MDQ6VXNlcjE5MTA1MTg= | User | false |
huggingface/transformers | 4,072,495,180 | I_kwDOCUB6oc7yvVhM | 44,678 | https://github.com/huggingface/transformers/issues/44678 | https://api.github.com/repos/huggingface/transformers/issues/44678 | Use index_select instead of fancy indexing in batched_mm_experts_forward | ## Description
`batched_mm_experts_forward` in `transformers/integrations/moe.py` uses fancy indexing (`self.gate_up_proj[expert_ids]`) to select expert weights and biases. While semantically correct, fancy indexing is ambiguous for some compiler backends — it could be interpreted as gather, advanced indexing, or slic... | closed | completed | false | 1 | [] | [] | 2026-03-13T17:38:02Z | 2026-03-19T12:13:33Z | 2026-03-19T12:13:33Z | MEMBER | null | 20260325T173244Z | 2026-03-25T17:32:44Z | dacorvo | 1,910,518 | MDQ6VXNlcjE5MTA1MTg= | User | false |
huggingface/transformers | 4,072,495,877 | I_kwDOCUB6oc7yvVsF | 44,679 | https://github.com/huggingface/transformers/issues/44679 | https://api.github.com/repos/huggingface/transformers/issues/44679 | Allow kernel modules to declare their preferred mask function | ## Description
`load_and_register_attn_kernel` in `transformers/integrations/hub_kernels.py` hardcodes `flash_attention_2` as the mask function for all custom attention kernels:
```python
ALL_MASK_ATTENTION_FUNCTIONS.register(attn_implementation, ALL_MASK_ATTENTION_FUNCTIONS["flash_attention_2"])
```
This means cust... | open | null | false | 1 | [] | [] | 2026-03-13T17:38:08Z | 2026-04-13T08:37:38Z | null | MEMBER | null | 20260413T085906Z | 2026-04-13T08:59:06Z | dacorvo | 1,910,518 | MDQ6VXNlcjE5MTA1MTg= | User | false |
huggingface/transformers | 4,073,182,351 | I_kwDOCUB6oc7yx9SP | 44,683 | https://github.com/huggingface/transformers/issues/44683 | https://api.github.com/repos/huggingface/transformers/issues/44683 | Compiled flex_attention fails on torch >= 2.9 | ### System Info
All recent transformers versions -- impacts torch >= 2.9
### Who can help?
@vasqu @ArthurZucker @Cyrilvallez
### Information
- [ ] The official example scripts
- [ ] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [ ] My ow... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-03-13T20:09:58Z | 2026-03-18T11:44:20Z | 2026-03-18T11:44:20Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | ntenenz | 8,411,908 | MDQ6VXNlcjg0MTE5MDg= | User | false |
huggingface/transformers | 4,075,806,271 | I_kwDOCUB6oc7y794_ | 44,701 | https://github.com/huggingface/transformers/issues/44701 | https://api.github.com/repos/huggingface/transformers/issues/44701 | Example: Handling imbalanced text classification with F1-score evaluation using Trainer API | Many real-world NLP classification tasks have imbalanced label distributions.
However, most example scripts in the repository evaluate models primarily using accuracy.
Accuracy can be misleading for imbalanced datasets, and metrics such as F1-score
or balanced accuracy are often more appropriate.
I would like to cont... | closed | completed | false | 1 | [
"Code agent slop"
] | [] | 2026-03-14T13:16:42Z | 2026-03-18T16:12:21Z | 2026-03-18T12:59:40Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | MdSaifAli786123 | 136,231,572 | U_kgDOCB66lA | User | false |
huggingface/transformers | 4,075,957,274 | I_kwDOCUB6oc7y8iwa | 44,704 | https://github.com/huggingface/transformers/issues/44704 | https://api.github.com/repos/huggingface/transformers/issues/44704 | AutoProcessor.from_pretrained not passing all kwargs to cached_file | Hi,
I believe `AutoProcessor.from_pretrained` is not forwarding arguments correctly to `cached_file`.
The `cached_file` function is [defined with `**kwargs`](https://github.com/huggingface/transformers/blob/main/src/transformers/utils/hub.py#L225). However, `AutoProcessor.from_pretrained` filters the provided kwargs... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-14T14:46:06Z | 2026-03-25T18:04:24Z | 2026-03-25T18:04:24Z | NONE | null | 20260325T200011Z | 2026-03-25T20:00:11Z | peacefulotter | 32,218,033 | MDQ6VXNlcjMyMjE4MDMz | User | false |
huggingface/transformers | 4,076,931,510 | I_kwDOCUB6oc7zAQm2 | 44,716 | https://github.com/huggingface/transformers/issues/44716 | https://api.github.com/repos/huggingface/transformers/issues/44716 | `PixioPatchEmbeddings.forward` supports `interpolate_pos_encoding` but it is not propagated through `PixioEmbeddings`/`PixioModel` | The `PixioPatchEmbeddings` module defines an `interpolate_pos_encoding` argument in its forward method, but this argument is never propagated from higher-level modules.
Current call chain:
`PixioPatchEmbeddings` → `PixioEmbeddings` → `PixioModel`
Since `PixioEmbeddings` and `PixioModel` do not pass or expose the `int... | open | null | false | 1 | [] | [] | 2026-03-14T22:41:54Z | 2026-03-27T02:44:21Z | null | CONTRIBUTOR | null | 20260407T090028Z | 2026-04-07T09:00:28Z | audioXD | 8,224,015 | MDQ6VXNlcjgyMjQwMTU= | User | false |
huggingface/transformers | 4,077,031,787 | I_kwDOCUB6oc7zApFr | 44,717 | https://github.com/huggingface/transformers/issues/44717 | https://api.github.com/repos/huggingface/transformers/issues/44717 | Support packed sequences for linear attention models (i.e. Qwen3.5) | ### Feature request
Currently, packing does not seem supported for text-based datasets (https://github.com/unslothai/unsloth/issues/4160). It would be good to support this.
### Motivation
Without packing, my training runs are approximately 3-5x more expensive with the dataset that I'd like to use, and also suffer fr... | open | null | false | 14 | [
"Feature request"
] | [] | 2026-03-14T23:22:19Z | 2026-03-26T21:04:34Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | kirawi | 67,773,714 | MDQ6VXNlcjY3NzczNzE0 | User | false |
huggingface/transformers | 4,080,096,496 | I_kwDOCUB6oc7zMVTw | 44,734 | https://github.com/huggingface/transformers/issues/44734 | https://api.github.com/repos/huggingface/transformers/issues/44734 | `transformers serve`: /v1/responses crashes on KV cache continuation due to wrong tensor indexing | ### System Info
- transformers version: 5.3.0 (latest main)
- Platform: Linux / macOS
- Python version: 3.12
### Who can help?
@LysandreJik @Rocketknight1
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [ ] An officially supported task in the `examples` folder (such a... | closed | completed | false | 1 | [] | [] | 2026-03-16T04:08:09Z | 2026-03-16T15:28:01Z | 2026-03-16T15:28:01Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | mango766 | 85,663,565 | MDQ6VXNlcjg1NjYzNTY1 | User | false |
huggingface/transformers | 4,080,523,655 | I_kwDOCUB6oc7zN9mH | 44,737 | https://github.com/huggingface/transformers/issues/44737 | https://api.github.com/repos/huggingface/transformers/issues/44737 | XLNet: relative_positional_encoding computes on CPU every forward pass (missing device= in torch.arange) | ### System Info
- `transformers` version: confirmed on 4.30.2, still present on `main` as of 2026-03-16
- Platform: Linux (tested on NVIDIA/AMD GPUs)
- PyTorch: 2.x
- Python: 3.10+
### Who can help?
@ArthurZucker @Rocketknight1
### Information
- [x] The official example scripts
- [x] My own modified scripts
### T... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-16T06:30:52Z | 2026-03-19T13:30:49Z | 2026-03-19T13:30:49Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | mvstrauss | 38,297,631 | MDQ6VXNlcjM4Mjk3NjMx | User | false |
huggingface/transformers | 4,080,877,033 | I_kwDOCUB6oc7zPT3p | 44,740 | https://github.com/huggingface/transformers/issues/44740 | https://api.github.com/repos/huggingface/transformers/issues/44740 | Run model linter on new models additions | - detect PRs with new models
- add a `new model` label
- run the linter and when failing comment the PR with errors | open | null | false | 2 | [] | [
"tarekziade"
] | 2026-03-16T08:04:16Z | 2026-04-17T08:29:51Z | null | MEMBER | null | 20260417T180542Z | 2026-04-17T18:05:42Z | tarekziade | 250,019 | MDQ6VXNlcjI1MDAxOQ== | User | false |
huggingface/transformers | 4,081,037,683 | I_kwDOCUB6oc7zP7Fz | 44,741 | https://github.com/huggingface/transformers/issues/44741 | https://api.github.com/repos/huggingface/transformers/issues/44741 | [Neuron] Improve transformers compatibility with AWS Neuron devices | ## Context
AWS Neuron devices (Trainium/Inferentia) compile a separate **NEFF** (Neuron Executable File Format) for every unique tensor shape. Any code path that changes tensor shapes between iterations — growing masks, `torch.cat` on outputs, variable padding — triggers expensive recompilations (2–60s per NEFF depend... | closed | completed | false | 2 | [] | [] | 2026-03-16T08:38:23Z | 2026-04-17T14:02:28Z | 2026-04-17T14:02:28Z | MEMBER | null | 20260417T180542Z | 2026-04-17T18:05:42Z | dacorvo | 1,910,518 | MDQ6VXNlcjE5MTA1MTg= | User | false |
huggingface/transformers | 4,081,060,880 | I_kwDOCUB6oc7zQAwQ | 44,742 | https://github.com/huggingface/transformers/issues/44742 | https://api.github.com/repos/huggingface/transformers/issues/44742 | [Neuron] Static-shape generation loop for compilation-friendly inference | ## Context
`GenerationMixin._sample` grows `input_ids`, `attention_mask`, and `position_ids` via `torch.cat` on every decode step. This is problematic for any backend where dynamic tensor shapes carry a cost:
* **XLA/torch.compile backends:** Static shapes are required for graph caching — dynamic shapes cause retraci... | open | null | false | 3 | [] | [] | 2026-03-16T08:43:33Z | 2026-04-17T13:47:38Z | null | MEMBER | null | 20260417T180542Z | 2026-04-17T18:05:42Z | dacorvo | 1,910,518 | MDQ6VXNlcjE5MTA1MTg= | User | false |
huggingface/transformers | 4,081,078,375 | I_kwDOCUB6oc7zQFBn | 44,743 | https://github.com/huggingface/transformers/issues/44743 | https://api.github.com/repos/huggingface/transformers/issues/44743 | transformers modular_qwen3_5.py: Recurrent states always reset when using cache and seq_len>1 | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux-5.15.0-171-generic-x86_64-with-glibc2.35
- Python version: 3.12.11
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (acceler... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-16T08:47:36Z | 2026-03-17T06:34:36Z | 2026-03-16T09:46:10Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | yansongzhou | 237,394,923 | U_kgDODiZb6w | User | false |
huggingface/transformers | 4,081,249,005 | I_kwDOCUB6oc7zQurt | 44,744 | https://github.com/huggingface/transformers/issues/44744 | https://api.github.com/repos/huggingface/transformers/issues/44744 | 推理模型的缺陷 | 这种架构的推理模型无法解决非线性推理模型下的问题。
例如：1+2，2+1=6，推理模型不会把目标设置成如何获得6缺失了什么条件，而是基于训练数据在疯狂地做小学数学。
第二个共性特征解析：在开发ASR,TTS的编程任务中，推理模型只会质疑代码文件历史中的潜在问题，并不会察觉到数据流中隐藏的线索是反向演绎法，和诊断问题的最短路径是：文本A→TTS→Wav→ASR→文本A。
如果推理模型没有实质性地研发进展，那我们只是再次造了一个用于搜索向量数据库的Google.甚至都没有Google提供的情报有参考性。
此架构下的数据库收录了大多数人类意识资产，却没有真正还原人类推理模型的一半能力。 | closed | completed | false | 0 | [] | [] | 2026-03-16T09:22:56Z | 2026-03-18T15:11:18Z | 2026-03-18T15:11:18Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | QuickerStudio | 26,364,832 | MDQ6VXNlcjI2MzY0ODMy | User | false |
huggingface/transformers | 4,081,508,778 | I_kwDOCUB6oc7zRuGq | 44,746 | https://github.com/huggingface/transformers/issues/44746 | https://api.github.com/repos/huggingface/transformers/issues/44746 | 最理想的Transformes二代生态环境的设想，100B的推理引擎模型会成为未来标准。 | 定律：
1.针对任务类型而优化的不到100B的“闭源”推理模型采用二进制加密部署到客户本地，后来以客户为中心形成的云计算生态将大幅度平摊算力和能源局限性。（增量更新型的大模型才是未来标准，虽然现在看起来很荒谬，但是人类意识就是在不停的增量更新，直到事实合理）
2.公共开放式地增量向量数据库收录系统。（每人抱着有各种缺陷的向量数据库，不如动用所有人的力量去维护一个完美的向量数据库）
3.小而精准的本地资料库负责搜索，管理的“开源”模型。（简单推理需求本地化）
4.基于开源合作下的功能研发类型的软件生态环境诞生的商业经济。（企业研发的试错场）
进入这个时代的人们向量数据库共享，知识共享，大量的人才会有更多的精力去校准推理能力和应... | closed | completed | false | 1 | [] | [] | 2026-03-16T10:08:38Z | 2026-03-18T15:15:22Z | 2026-03-18T15:15:22Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | QuickerStudio | 26,364,832 | MDQ6VXNlcjI2MzY0ODMy | User | false |
huggingface/transformers | 4,081,656,083 | I_kwDOCUB6oc7zSSET | 44,748 | https://github.com/huggingface/transformers/issues/44748 | https://api.github.com/repos/huggingface/transformers/issues/44748 | [Neuron] Auto-select StaticCache when device is Neuron | ## Context
On Neuron devices, `StaticCache` is required for correct generation — dynamic tensor shapes trigger per-step recompilations. Currently, users must explicitly pass `past_key_values=StaticCache(...)` or `cache_implementation="static"` to `model.generate()`. Ideally, `model.generate()` should auto-select `Stat... | open | null | false | 3 | [] | [] | 2026-03-16T10:38:15Z | 2026-04-17T14:05:50Z | null | MEMBER | null | 20260417T180542Z | 2026-04-17T18:05:42Z | dacorvo | 1,910,518 | MDQ6VXNlcjE5MTA1MTg= | User | false |
huggingface/transformers | 4,081,994,049 | I_kwDOCUB6oc7zTklB | 44,749 | https://github.com/huggingface/transformers/issues/44749 | https://api.github.com/repos/huggingface/transformers/issues/44749 | Transformer 从4.57.3 升级到5.3.0 后过滤数据时长变慢十倍以上 | ### System Info
H20
### 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
过滤函数实现：对应链接 ht... | open | null | false | 9 | [
"bug"
] | [] | 2026-03-16T11:48:25Z | 2026-04-18T08:11:32Z | null | NONE | null | 20260418T090534Z | 2026-04-18T09:05:34Z | chenjiaoAngel | 38,650,344 | MDQ6VXNlcjM4NjUwMzQ0 | User | false |
huggingface/transformers | 4,082,686,570 | I_kwDOCUB6oc7zWNpq | 44,754 | https://github.com/huggingface/transformers/issues/44754 | https://api.github.com/repos/huggingface/transformers/issues/44754 | I really need help regarding to meta device issue | I installed TRELLIS 2 and getting the error for meta device, I followed the instructions here for installing this release of transformers, but when I launch TRELLIS 2 I still get the same error. Can someone please help me Fix this step by step because I spend 5 days trying without any solution | open | null | false | 2 | [] | [] | 2026-03-16T14:03:28Z | 2026-04-17T08:29:48Z | null | NONE | null | 20260417T180542Z | 2026-04-17T18:05:42Z | Hany138078 | 230,026,885 | U_kgDODbXuhQ | User | false |
huggingface/transformers | 4,082,832,596 | I_kwDOCUB6oc7zWxTU | 44,756 | https://github.com/huggingface/transformers/issues/44756 | https://api.github.com/repos/huggingface/transformers/issues/44756 | Disable mmap on Strix Halo to avoid OOM | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux-6.19.0-9-generic-x86_64-with-glibc2.42
- Python version: 3.13.12
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?... | open | null | false | 2 | [
"bug"
] | [] | 2026-03-16T14:28:04Z | 2026-04-11T00:52:38Z | null | CONTRIBUTOR | null | 20260411T144729Z | 2026-04-11T14:47:29Z | woct0rdho | 23,053,399 | MDQ6VXNlcjIzMDUzMzk5 | User | false |
huggingface/transformers | 4,085,509,339 | I_kwDOCUB6oc7zg-zb | 44,779 | https://github.com/huggingface/transformers/issues/44779 | https://api.github.com/repos/huggingface/transformers/issues/44779 | Deepseek tokenizer produces incorrect results as of v5 (works in v4) | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux-6.6.113+-x86_64-with-glibc2.35
- Python version: 3.12.12
- Huggingface_hub version: 1.6.0
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.10... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-03-17T00:30:56Z | 2026-03-18T21:35:14Z | 2026-03-18T21:35:14Z | MEMBER | null | 20260325T173244Z | 2026-03-25T17:32:44Z | xenova | 26,504,141 | MDQ6VXNlcjI2NTA0MTQx | User | false |
huggingface/transformers | 4,087,676,990 | I_kwDOCUB6oc7zpQA- | 44,792 | https://github.com/huggingface/transformers/issues/44792 | https://api.github.com/repos/huggingface/transformers/issues/44792 | Failed test case `test_model_generate_images` for janus model | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Linux-5.4.292-1.el8.elrepo.x86_64-x86_64-with-glibc2.35
- Python version: 3.10.12
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.5.3
- Accelerate version: 1.12.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch ve... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-17T10:40:18Z | 2026-04-17T08:32:33Z | 2026-04-17T08:32:33Z | CONTRIBUTOR | null | 20260417T180542Z | 2026-04-17T18:05:42Z | kaixuanliu | 13,268,042 | MDQ6VXNlcjEzMjY4MDQy | User | false |
huggingface/transformers | 4,092,031,687 | I_kwDOCUB6oc7z53LH | 44,805 | https://github.com/huggingface/transformers/issues/44805 | https://api.github.com/repos/huggingface/transformers/issues/44805 | IndexError: The shape of the mask [...] at index 0 does not match the shape of the indexed tensor [...] at index 0 | ### System Info
```
- `transformers` version: 5.3.0
- Platform: Linux-5.15.0-164-generic-x86_64-with-glibc2.35
- Python version: 3.12.13
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: 0.16.4
- PyTorch version (accelerat... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-03-18T00:58:28Z | 2026-03-18T12:20:33Z | 2026-03-18T12:20:33Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | akowalsk | 49,454,006 | MDQ6VXNlcjQ5NDU0MDA2 | User | false |
huggingface/transformers | 4,093,210,823 | I_kwDOCUB6oc7z-XDH | 44,810 | https://github.com/huggingface/transformers/issues/44810 | https://api.github.com/repos/huggingface/transformers/issues/44810 | Showcase / question: a board-proven offline language runtime on ESP32-C3, and whether some future language capability may move beyond general model definitions | Hi Transformers folks,
I wanted to share ... | closed | completed | false | 2 | [] | [] | 2026-03-18T07:09:16Z | 2026-03-18T16:01:42Z | 2026-03-18T15:47:59Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | Alpha-Guardian | 203,005,953 | U_kgDODBmgAQ | User | false |
huggingface/transformers | 4,093,576,102 | I_kwDOCUB6oc7z_wOm | 44,811 | https://github.com/huggingface/transformers/issues/44811 | https://api.github.com/repos/huggingface/transformers/issues/44811 | Whisper processor.batch_decode() function ignoring skip_special_tokens params | ### System Info
- `transformers` version: 4.57.6
- Platform: Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.35
- Python version: 3.10.13
- Huggingface_hub version: 0.36.2
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorc... | closed | completed | false | 5 | [
"bug"
] | [] | 2026-03-18T08:30:17Z | 2026-03-20T10:59:59Z | 2026-03-20T10:59:59Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | cfasana | 143,723,410 | U_kgDOCJELkg | User | false |
huggingface/transformers | 4,094,388,173 | I_kwDOCUB6oc70C2fN | 44,821 | https://github.com/huggingface/transformers/issues/44821 | https://api.github.com/repos/huggingface/transformers/issues/44821 | Unable to load `AutoImageProcessor` from URL | ### System Info
<details><summary>Versions</summary>
<p>
- `transformers` version: 5.3.0
- Platform: Linux-6.8.0-106-generic-x86_64-with-glibc2.39
- Python version: 3.12.3
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed versi... | closed | completed | false | 6 | [
"bug"
] | [] | 2026-03-18T11:08:09Z | 2026-03-23T13:13:38Z | 2026-03-23T13:10:59Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | BSchilperoort | 12,114,825 | MDQ6VXNlcjEyMTE0ODI1 | User | false |
huggingface/transformers | 4,095,309,295 | I_kwDOCUB6oc70GXXv | 44,829 | https://github.com/huggingface/transformers/issues/44829 | https://api.github.com/repos/huggingface/transformers/issues/44829 | AutoModelForSequenceClassification with attn_implementation="flash_attention_3" causes degenerate training (loss increases, model predicts all-one-class) | ### System Info
When fine-tuning `Qwen3ForSequenceClassification` (loaded via `AutoModelForSequenceClassification`) with `attn_implementation="flash_attention_3"`, training completely fails: loss increases instead of decreasing, and the model collapses to predicting all examples as one class. Removing `attn_implementa... | open | null | false | 2 | [
"bug"
] | [] | 2026-03-18T13:56:33Z | 2026-04-18T08:11:30Z | null | NONE | null | 20260418T090534Z | 2026-04-18T09:05:34Z | Jantory | 36,835,418 | MDQ6VXNlcjM2ODM1NDE4 | User | false |
huggingface/transformers | 621,683,208 | MDU6SXNzdWU2MjE2ODMyMDg= | 4,483 | https://github.com/huggingface/transformers/issues/4483 | https://api.github.com/repos/huggingface/transformers/issues/4483 | Trying to add support for GPT2 as decoder in EncoderDecoder model | # 🚀 Feature request
Hi,
I am trying to add the option of using GPT2 as the decoder in the EncoderDecoder model, which only support
## Motivation
For a generation problem, it usually better to use GPT2 as the decoder, over BERT.
## Your contribution
I've made the following changes in `modeling_gpt2.py... | closed | completed | false | 32 | [
"Core: Encoder-Decoder",
"Good First Issue"
] | [
"patrickvonplaten"
] | 2020-05-20T11:24:44Z | 2026-03-02T13:42:22Z | 2026-03-02T13:42:22Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | dimi1357 | 22,443,447 | MDQ6VXNlcjIyNDQzNDQ3 | User | false |
huggingface/transformers | 4,098,706,305 | I_kwDOCUB6oc70TUuB | 44,840 | https://github.com/huggingface/transformers/issues/44840 | https://api.github.com/repos/huggingface/transformers/issues/44840 | typo in code block in docs | typo in code block:
https://github.com/huggingface/transformers/blob/16a5b0936dcaae6efbc03ab1a4fd98dc324bfb9e/docs/source/en/weightconverter.md?plain=1#L69 | closed | completed | false | 0 | [] | [] | 2026-03-19T01:46:22Z | 2026-03-19T11:56:52Z | 2026-03-19T11:56:52Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | zhulinchng | 24,189,730 | MDQ6VXNlcjI0MTg5NzMw | User | false |
huggingface/transformers | 4,098,836,913 | I_kwDOCUB6oc70T0mx | 44,841 | https://github.com/huggingface/transformers/issues/44841 | https://api.github.com/repos/huggingface/transformers/issues/44841 | Processor fails for mistralai/Voxtral-Mini-3B-2507 | ### System Info
I am trying to run inference using `mistralai/Voxtral-Mini-3B-2507` on an audio (`np.ndarray`). On loading the processor using `processor = transformers.AutoProcessor.from_pretrained(MODEL, trust_remote_code=True)`, I am getting the following error:
```
/usr/local/lib/python3.12/dist-packages/huggingfa... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-19T02:26:39Z | 2026-03-24T10:07:36Z | 2026-03-24T10:07:36Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | BhavyaShah1234 | 62,424,200 | MDQ6VXNlcjYyNDI0MjAw | User | false |
huggingface/transformers | 4,099,433,723 | I_kwDOCUB6oc70WGT7 | 44,843 | https://github.com/huggingface/transformers/issues/44843 | https://api.github.com/repos/huggingface/transformers/issues/44843 | AutoTokenizer.from_pretrained calls model_info() unconditionally in _patch_mistral_regex, breaks HF_HUB_OFFLINE mode | ### System Info
- `transformers` version: 4.57.3
- `huggingface_hub` version: 0.36.2
- Python: 3.12
- OS: Linux (Ubuntu 24.04, inside NVIDIA container)
### Who can help?
@ArthurZucker @itazap
### Regression introduced in
PR #42389 (`[Mistral Tokenizers] Fix tokenizer detection`), included in v4.57.2 → v4.57.3.
##... | open | null | false | 5 | [] | [] | 2026-03-19T05:36:56Z | 2026-03-27T10:58:49Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | nv-yna | 248,773,860 | U_kgDODtP85A | User | false |
huggingface/transformers | 4,100,120,851 | I_kwDOCUB6oc70YuET | 44,849 | https://github.com/huggingface/transformers/issues/44849 | https://api.github.com/repos/huggingface/transformers/issues/44849 | Transformers Qwen3.5 had a bug when set output_hidden_states=True | ### System Info
Version: 5.2.0
in qwen3.5
outputs = model_wrapper.generate(**inputs, output_hidden_states=True)
outpus something like this:
```
><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><|image_pad|><... | open | null | false | 5 | [
"bug"
] | [] | 2026-03-19T08:27:35Z | 2026-04-19T08:16:52Z | null | NONE | null | 20260419T090534Z | 2026-04-19T09:05:34Z | lucasjinreal | 21,303,438 | MDQ6VXNlcjIxMzAzNDM4 | User | false |
huggingface/transformers | 4,100,959,949 | I_kwDOCUB6oc70b67N | 44,855 | https://github.com/huggingface/transformers/issues/44855 | https://api.github.com/repos/huggingface/transformers/issues/44855 | IndentationError when importing DebertaV2Model on Python 3.13 - @torch.jit.script fails to parse function with comment between decorator and def | ## Description
Importing `DebertaV2Model` from `transformers` (or any library that depends on it, such as `gliner`) raises an `IndentationError` on Python 3.13. The error originates in `torch.jit.script` when it attempts to re-parse the source of a JIT-scripted function that has a comment placed between the `@torch.ji... | closed | completed | false | 6 | [
"bug"
] | [] | 2026-03-19T11:07:31Z | 2026-03-26T07:49:53Z | 2026-03-25T13:39:56Z | NONE | null | 20260326T102019Z | 2026-03-26T10:20:19Z | MNIKIEMA | 64,019,294 | MDQ6VXNlcjY0MDE5Mjk0 | User | false |
huggingface/transformers | 4,101,508,603 | I_kwDOCUB6oc70eA37 | 44,857 | https://github.com/huggingface/transformers/issues/44857 | https://api.github.com/repos/huggingface/transformers/issues/44857 | LwDetrImageLoss crashes when using float16 AMP and Cuda | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Linux-6.6.87.2-microsoft-standard-WSL2-x86_64-with-glibc2.39
- Python version: 3.12.3
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorc... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-19T12:56:22Z | 2026-03-24T17:02:34Z | 2026-03-24T17:02:34Z | CONTRIBUTOR | null | 20260325T173244Z | 2026-03-25T17:32:44Z | m-matthias | 16,415,097 | MDQ6VXNlcjE2NDE1MDk3 | User | false |
huggingface/transformers | 4,102,371,828 | I_kwDOCUB6oc70hTn0 | 44,861 | https://github.com/huggingface/transformers/issues/44861 | https://api.github.com/repos/huggingface/transformers/issues/44861 | _get_tied_weight_keys crashes with AttributeError when _tied_weights_keys is a list | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux
- Python version: 3.13
### Who can help?
@Cyrilvallez
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Reproduction
Full finetune of `NVIDIA-Nemotron-3-Nano-4B` crashes at checkpoint save wi... | closed | completed | false | 0 | [] | [] | 2026-03-19T15:13:12Z | 2026-03-20T09:46:56Z | 2026-03-20T09:46:56Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | gh-wf | 111,619,017 | U_kgDOBqcryQ | User | false |
huggingface/transformers | 4,102,382,384 | I_kwDOCUB6oc70hWMw | 44,863 | https://github.com/huggingface/transformers/issues/44863 | https://api.github.com/repos/huggingface/transformers/issues/44863 | NemotronH implementation can't load NemotronH checkpoints! | ### System Info
```console
- `transformers` version: 5.3.0
- Platform: macOS-15.7.3-arm64-arm-64bit
- Python version: 3.12.11
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerato... | open | null | false | 7 | [
"bug"
] | [] | 2026-03-19T15:15:06Z | 2026-03-29T18:49:56Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | elfprince13 | 2,703,145 | MDQ6VXNlcjI3MDMxNDU= | User | false |
huggingface/transformers | 4,104,609,447 | I_kwDOCUB6oc70p16n | 44,868 | https://github.com/huggingface/transformers/issues/44868 | https://api.github.com/repos/huggingface/transformers/issues/44868 | [rag-end2end-retriever] Broken Google Drive link for SQuAD dataset and hyperparameters | ## Description
The README for the `rag-end2end-retriever` research project contains a broken
Google Drive link to the SQuAD training dataset, knowledge base, and
hyperparameters used in the experiments.
**Location:**
https://github.com/huggingface/transformers-research-projects/tree/main/rag-end2end-retriever
**B... | open | null | false | 1 | [] | [] | 2026-03-19T22:42:17Z | 2026-04-19T08:16:50Z | null | NONE | null | 20260419T090534Z | 2026-04-19T09:05:34Z | lmmanriquem | 101,687,883 | U_kgDOBg-iSw | User | false |
huggingface/transformers | 4,105,125,767 | I_kwDOCUB6oc70rz-H | 44,869 | https://github.com/huggingface/transformers/issues/44869 | https://api.github.com/repos/huggingface/transformers/issues/44869 | Whisper word timestamp decode crashes on trailing replacement character at end of decoded token stream | ### System Info
### System Info
- OS: macOS
- `transformers`: `5.3.0.dev0`
- Model: `openai/whisper-medium.en`
### Reproduction
I hit an `IndexError: string index out of range` in Whisper word-timestamp decoding and traced it to `src/transformers/models/whisper/tokenization_whisper.py`.
The failing code path is in... | open | null | false | 3 | [
"bug"
] | [] | 2026-03-20T01:25:49Z | 2026-04-15T13:47:57Z | null | NONE | null | 20260415T224019Z | 2026-04-15T22:40:19Z | chromatic-descension | 98,860,570 | U_kgDOBeR-Gg | User | false |
huggingface/transformers | 4,105,674,824 | I_kwDOCUB6oc70t6BI | 44,871 | https://github.com/huggingface/transformers/issues/44871 | https://api.github.com/repos/huggingface/transformers/issues/44871 | [Gemma-3] Inconsistent eos_token_id configuration: tokenizer has single value (1) but model.config has list [1, 106] | ### System Info
- `transformers` version: 5.3.0
- Platform: Windows-11-10.0.26100-SP0
- Python version: 3.12.11
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.7.1+cu1... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-20T04:29:57Z | 2026-03-21T01:40:09Z | 2026-03-21T01:40:09Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | IvanFan-Van | 98,149,954 | U_kgDOBdmmQg | User | false |
huggingface/transformers | 4,106,938,437 | I_kwDOCUB6oc70yuhF | 44,877 | https://github.com/huggingface/transformers/issues/44877 | https://api.github.com/repos/huggingface/transformers/issues/44877 | Strict config prevents loading `granite_speech` config | ### System Info
- `transformers` version: 5.3.0.dev0
- Platform: Windows-10-10.0.26200-SP0
- Python version: 3.11.13
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.6.2
- Accelerate version: 1.11.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.9.0+cu... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-03-20T09:57:02Z | 2026-03-27T09:30:18Z | 2026-03-27T09:30:17Z | MEMBER | null | 20260407T090028Z | 2026-04-07T09:00:28Z | tomaarsen | 37,621,491 | MDQ6VXNlcjM3NjIxNDkx | User | false |
huggingface/transformers | 4,109,855,475 | I_kwDOCUB6oc7092rz | 44,898 | https://github.com/huggingface/transformers/issues/44898 | https://api.github.com/repos/huggingface/transformers/issues/44898 | [BUG] Perceiver image classification (non-default res) fails even with interpolate_pos_encoding=True | ### System Info
* `transformers` version: `5.0.0.dev0`
* Platform: `Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.39`
* Python version: `3.12.3`
* `huggingface_hub` version: `1.3.2`
* `safetensors` version: `0.7.0`
* `accelerate` version: `1.12.0`
* Accelerate config: `not installed`
* DeepSpeed version:... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-03-20T19:58:09Z | 2026-04-18T09:06:59Z | 2026-03-25T11:48:03Z | CONTRIBUTOR | null | 20260418T100536Z | 2026-04-18T10:05:36Z | harshaljanjani | 75,426,551 | MDQ6VXNlcjc1NDI2NTUx | User | false |
huggingface/transformers | 4,111,564,283 | I_kwDOCUB6oc71EX37 | 44,906 | https://github.com/huggingface/transformers/issues/44906 | https://api.github.com/repos/huggingface/transformers/issues/44906 | Remove unnecessary `expand_as` in `get_placeholder_mask` across VLMs | ### Feature request
## Problem
The `get_placeholder_mask` function (and equivalent inline patterns) across ~70 multimodal model files expands a boolean placeholder mask from shape `(B, S, 1)` to `(B, S, H)` via `.expand_as(inputs_embeds)` before passing it to `masked_scatter`. This expansion is unnecessary because `m... | open | null | false | 0 | [
"Feature request"
] | [] | 2026-03-21T06:05:36Z | 2026-03-21T06:05:36Z | null | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | syncdoth | 45,599,998 | MDQ6VXNlcjQ1NTk5OTk4 | User | false |
huggingface/transformers | 4,111,661,658 | I_kwDOCUB6oc71Evpa | 44,908 | https://github.com/huggingface/transformers/issues/44908 | https://api.github.com/repos/huggingface/transformers/issues/44908 | inverse_sqrt scheduler ignores lr_scheduler_kwargs (timescale not passed) | ### System Info
Incomplete arguments passed for schedulers where name is explicitly checked.
https://github.com/huggingface/transformers/blob/v5.3.0/src/transformers/optimization.py#L664
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- [ ] My own modified scripts
### Tasks
- ... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-21T06:53:50Z | 2026-03-24T13:06:18Z | 2026-03-24T13:06:18Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | magarwal0205 | 6,791,598 | MDQ6VXNlcjY3OTE1OTg= | User | false |
huggingface/transformers | 4,112,866,232 | I_kwDOCUB6oc71JVu4 | 44,910 | https://github.com/huggingface/transformers/issues/44910 | https://api.github.com/repos/huggingface/transformers/issues/44910 | [Bug] Flash Attention crashes with illegal memory access on Qwen3.5 due to 3D position_ids being misinterpreted as packed sequence | ### System Info
# [Bug] Flash Attention crashes with `illegal memory access` on Qwen3.5 due to 3D `position_ids` being misinterpreted as packed sequence
We fixed it in https://github.com/ouroborosscr/transformers/tree/fix/qwen35-flash-attn-3d-position-ids
## Description
When using `attn_implementation="flash_attent... | closed | completed | false | 6 | [
"bug"
] | [] | 2026-03-21T15:38:54Z | 2026-03-25T02:03:24Z | 2026-03-24T14:36:36Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | ouroborosscr | 75,082,702 | MDQ6VXNlcjc1MDgyNzAy | User | false |
huggingface/transformers | 4,113,147,717 | I_kwDOCUB6oc71KadF | 44,912 | https://github.com/huggingface/transformers/issues/44912 | https://api.github.com/repos/huggingface/transformers/issues/44912 | git-oss-20b will not properly load with MXFP4 quantization and falls back to bf16 | ### System Info
Hi, probably this is related to
https://github.com/huggingface/transformers/issues/42723
I get:
MXFP4 quantization requires Triton and kernels installed: CUDA requires Triton >= 3.4.0, XPU requires Triton >= 3.5.0, we will default to dequantizing the model to bf16
When executing the following code:
p... | closed | completed | false | 5 | [
"bug"
] | [] | 2026-03-21T17:38:02Z | 2026-03-24T15:33:17Z | 2026-03-24T15:33:17Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | jottbecr | 126,362,918 | U_kgDOB4glJg | User | false |
huggingface/transformers | 4,113,212,978 | I_kwDOCUB6oc71KqYy | 44,913 | https://github.com/huggingface/transformers/issues/44913 | https://api.github.com/repos/huggingface/transformers/issues/44913 | In GPTNeoXConfig, rotary_pct silently reverts to default on reload | ### System Info
- `transformers` version: 5.3.0
- Platform: Linux-6.17.0-19-generic-x86_64-with-glibc2.39
- Python version: 3.12.4
- Huggingface_hub version: 1.7.2
- Safetensors version: 0.4.5
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accel... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-21T17:58:32Z | 2026-03-27T09:32:37Z | 2026-03-27T09:19:47Z | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | ratishsp | 3,006,607 | MDQ6VXNlcjMwMDY2MDc= | User | false |
huggingface/transformers | 4,114,204,538 | I_kwDOCUB6oc71Ocd6 | 44,918 | https://github.com/huggingface/transformers/issues/44918 | https://api.github.com/repos/huggingface/transformers/issues/44918 | Unpacking Qwen3.5 input embeddings fails with trl SFT trainer | ### System Info
My environment:
```
datasets==4.6.1
faiss-cpu==1.13.2
numpy==2.4.2
pyserini==1.5.0
sentence-transformers==5.2.3
torch==2.10.0
torchvision==0.25.0
tqdm==4.67.3
trl==0.29.1
wandb==0.25.1
```
### Who can help?
@zucchini-nlp
### Information
- [ ] The official example scripts
- [x] My own modified scri... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-21T23:43:15Z | 2026-03-23T15:20:11Z | 2026-03-23T15:20:11Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | JakobJBauer | 53,357,351 | MDQ6VXNlcjUzMzU3MzUx | User | false |
huggingface/transformers | 4,116,497,241 | I_kwDOCUB6oc71XMNZ | 44,928 | https://github.com/huggingface/transformers/issues/44928 | https://api.github.com/repos/huggingface/transformers/issues/44928 | [Bug] Catastrophic gradient explosion (NaN) in RLHF with Qwen3.5 due to 3D position_ids forcing SDPA Math fallback and BF16 collapse | ### System Info
Copy-and-paste the text below in your GitHub issue and FILL OUT the two last points.
- `transformers` version: 5.3.0
- Platform: Linux-6.8.0-40-generic-x86_64-with-glibc2.35
- Python version: 3.11.15
- Huggingface_hub version: 1.7.1
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerat... | open | null | false | 3 | [
"bug"
] | [] | 2026-03-22T16:46:05Z | 2026-03-25T13:51:26Z | null | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | ouroborosscr | 75,082,702 | MDQ6VXNlcjc1MDgyNzAy | User | false |
huggingface/transformers | 4,116,586,484 | I_kwDOCUB6oc71Xh_0 | 44,929 | https://github.com/huggingface/transformers/issues/44929 | https://api.github.com/repos/huggingface/transformers/issues/44929 | First-class fine-tuning support for Mamba / Mamba-2 SSMs — architecture is production-ready, but the training path in Transformers isn't | ### Feature request
You can load Mamba models in Transformers — but the moment you try to actually fine-tune one, things fall apart fast. The standard Trainer was built around attention + KV cache assumptions that SSMs simply don't share. Gradient checkpointing breaks in weird ways, DataCollatorForLanguageModeling doe... | open | null | false | 5 | [
"Feature request"
] | [] | 2026-03-22T17:23:58Z | 2026-04-02T14:59:27Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | lochanharishwar | 221,997,305 | U_kgDODTto-Q | User | false |
huggingface/transformers | 4,116,661,565 | I_kwDOCUB6oc71X0U9 | 44,933 | https://github.com/huggingface/transformers/issues/44933 | https://api.github.com/repos/huggingface/transformers/issues/44933 | Nonexistant import from image_utils | ### System Info
I was getting the following error when running the latest version of main
`ImportError: cannot import name 'PILImageResampling' from 'transformers.image_utils' (/Users/josh/Documents/sandbox/.hugging/lib/python3.12/site-packages/transformers/image_utils.py)`
I found where it's being imported in src/t... | open | null | false | 8 | [
"bug"
] | [] | 2026-03-22T17:57:33Z | 2026-03-23T13:47:57Z | null | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | josh-kean | 40,722,578 | MDQ6VXNlcjQwNzIyNTc4 | User | false |
huggingface/transformers | 4,116,794,040 | I_kwDOCUB6oc71YUq4 | 44,936 | https://github.com/huggingface/transformers/issues/44936 | https://api.github.com/repos/huggingface/transformers/issues/44936 | trainer.evaluate() fails after trainer.train() | ### System Info
- `transformers` version: 5.3.0
- Platform: Windows-11-10.0.26200-SP0
- Python version: 3.13.0
- Huggingface_hub version: 1.7.2
- Safetensors version: 0.7.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.10.0+cpu (NA... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-03-22T18:48:49Z | 2026-04-13T13:47:53Z | 2026-04-13T13:47:53Z | NONE | null | 20260414T122001Z | 2026-04-14T12:20:01Z | HenrikEilers | 48,092,550 | MDQ6VXNlcjQ4MDkyNTUw | User | false |
huggingface/transformers | 4,117,554,107 | I_kwDOCUB6oc71bOO7 | 44,937 | https://github.com/huggingface/transformers/issues/44937 | https://api.github.com/repos/huggingface/transformers/issues/44937 | Check out "Google LLC" | <spam> | closed | completed | false | 0 | [] | [] | 2026-03-23T00:54:12Z | 2026-03-23T14:10:26Z | 2026-03-23T14:10:26Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | mija4264-arch38 | 251,540,492 | U_kgDODv40DA | User | false |
huggingface/transformers | 4,118,575,281 | I_kwDOCUB6oc71fHix | 44,938 | https://github.com/huggingface/transformers/issues/44938 | https://api.github.com/repos/huggingface/transformers/issues/44938 | transformers in Python 3.14 failed to load | ### System Info
Python 3.14
### Who can help?
_No response_
### Information
- [x] The official example scripts
- [ ] My own modified scripts
### Tasks
- [x] An officially supported task in the `examples` folder (such as GLUE/SQuAD, ...)
- [ ] My own task or dataset (give details below)
### Reproduction
transfo... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-03-23T06:27:08Z | 2026-03-24T13:10:15Z | 2026-03-24T12:29:27Z | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | resc863 | 50,689,611 | MDQ6VXNlcjUwNjg5NjEx | User | false |
huggingface/transformers | 4,120,161,600 | I_kwDOCUB6oc71lK1A | 44,945 | https://github.com/huggingface/transformers/issues/44945 | https://api.github.com/repos/huggingface/transformers/issues/44945 | Incorrect LLM output when using pipeline parallelism | ### System Info
transformers==4.57.1
Python==3.12.12
Kaggle env
### Who can help?
@CyrilVallez
@3outeille
### 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 ... | open | null | false | 0 | [
"bug"
] | [] | 2026-03-23T11:26:59Z | 2026-03-23T11:26:59Z | null | NONE | null | 20260325T173244Z | 2026-03-25T17:32:44Z | tasinislam21 | 33,097,868 | MDQ6VXNlcjMzMDk3ODY4 | User | false |
huggingface/transformers | 4,122,908,189 | I_kwDOCUB6oc71vpYd | 44,955 | https://github.com/huggingface/transformers/issues/44955 | https://api.github.com/repos/huggingface/transformers/issues/44955 | Orgs should be able to agree to 3rd party use agreements so their members can access 3rd party models | ### Feature request
for CI / CD the only way to get a sane workflow is to make a fake single user that the whole company uses to access the 3rd party resources. This undercuts your enterprise options and the idea that all users should log in as themselves. I know it's probably some legal hangup, but as is you are unde... | open | null | false | 4 | [
"Feature request"
] | [] | 2026-03-23T18:39:58Z | 2026-03-31T03:44:13Z | null | NONE | null | 20260407T090028Z | 2026-04-07T09:00:28Z | asglover | 140,220,574 | U_kgDOCFuYng | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.