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 | 3,451,662,789 | I_kwDOCUB6oc7NvDHF | 41,141 | https://github.com/huggingface/transformers/issues/41141 | https://api.github.com/repos/huggingface/transformers/issues/41141 | Need a concise example of Tensor Parallelism (TP) training using Trainer/SFTTrainer. | ### Feature request
I have checked the code and there are few places which talk about TP. I saw from_pretrained method for model contains tp_plan and device_mesh. I also checked that the TrainingArgument can take parallelism_config which defines the TP/CP plan along with FSDP. However, I am not able to successfully st... | closed | completed | false | 10 | [
"Documentation",
"Feature request",
"Tensor Parallel"
] | [] | 2025-09-25T03:01:02Z | 2026-06-22T12:23:53Z | 2026-06-22T12:23:53Z | NONE | null | 20260622T180028Z | 2026-06-22T18:00:28Z | meet-minimalist | 26,968,850 | MDQ6VXNlcjI2OTY4ODUw | User | false |
huggingface/transformers | 4,718,934,156 | I_kwDOCUB6oc8AAAABGUU4jA | 46,821 | https://github.com/huggingface/transformers/issues/46821 | https://api.github.com/repos/huggingface/transformers/issues/46821 | [TimesFM 2.5]:`window_size` argument raises `AttributeError` | ### System Info
- `transformers` version: 5.13.0.dev0
- Platform: macOS-26.5.1-arm64-arm-64bit
- Python version: 3.11.3
- Huggingface_hub version: 1.20.1
- Safetensors version: 0.8.0
- Accelerate version: 1.14.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.12... | closed | completed | false | 2 | [
"bug"
] | [
"kashif"
] | 2026-06-22T17:14:08Z | 2026-07-16T13:33:09Z | 2026-07-16T13:33:09Z | NONE | null | 20260717T000052Z | 2026-07-17T00:00:52Z | kanenorman | 51,185,594 | MDQ6VXNlcjUxMTg1NTk0 | User | false |
huggingface/transformers | 4,719,212,671 | I_kwDOCUB6oc8AAAABGUl4fw | 46,822 | https://github.com/huggingface/transformers/issues/46822 | https://api.github.com/repos/huggingface/transformers/issues/46822 | ZeRO-3 zero.Init does not partition composite minimax_m3_vl language submodule -> OOM on multi-GPU load | **Env:** transformers 5.12.1, deepspeed 0.18.9, torch 2.12.0+cu130, peft 0.19.1, bitsandbytes 0.49.2, accelerate 1.14.0; 8x B200 (178GB) / 2TB RAM; model MiniMaxAI/MiniMax-M3 (428B sparse MoE VL, minimax_m3_vl).
Loading `AutoModelForImageTextToText.from_pretrained("MiniMaxAI/MiniMax-M3", torch_dtype=bf16)` under DeepS... | open | null | false | 4 | [
"bug"
] | [] | 2026-06-22T17:52:42Z | 2026-06-23T18:38:19Z | null | NONE | null | 20260624T000023Z | 2026-06-24T00:00:23Z | trevorgordon981 | 193,960,977 | U_kgDOC4-cEQ | User | false |
huggingface/transformers | 4,719,212,867 | I_kwDOCUB6oc8AAAABGUl5Qw | 46,823 | https://github.com/huggingface/transformers/issues/46823 | https://api.github.com/repos/huggingface/transformers/issues/46823 | 4-bit load_in_4bit + device_map='auto' + explicit max_memory still dispatches modules to CPU on a model that fits -> ValueError | **Env:** transformers 5.12.1, deepspeed 0.18.9, torch 2.12.0+cu130, peft 0.19.1, bitsandbytes 0.49.2, accelerate 1.14.0; 8x B200 (178GB) / 2TB RAM; model MiniMaxAI/MiniMax-M3 (428B sparse MoE VL, minimax_m3_vl).
Loading MiniMax-M3 in 4-bit (nf4, double-quant) with `device_map="auto"` and `max_memory={i: "165GiB" for i... | closed | completed | false | 3 | [] | [] | 2026-06-22T17:52:44Z | 2026-07-09T14:10:04Z | 2026-07-09T08:59:27Z | NONE | null | 20260709T180026Z | 2026-07-09T18:00:26Z | trevorgordon981 | 193,960,977 | U_kgDOC4-cEQ | User | false |
huggingface/transformers | 4,722,666,998 | I_kwDOCUB6oc8AAAABGX4t9g | 46,829 | https://github.com/huggingface/transformers/issues/46829 | https://api.github.com/repos/huggingface/transformers/issues/46829 | [Bug] VideoPrismForVideoClassification returns last_hidden_state tensor as hidden_states in ImageClassifierOutput | ### System Info
- `transformers` version: 5.13.0.dev0 (main branch, local dev install)
- Platform: Linux (WSL2, kernel 6.6.114.1-microsoft-standard-WSL2)
- Python version: 3.14.4
- PyTorch version: 2.12.1+
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- [x] My own modified scri... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-06-23T04:40:35Z | 2026-06-23T12:39:42Z | 2026-06-23T12:39:42Z | CONTRIBUTOR | null | 20260623T180016Z | 2026-06-23T18:00:16Z | sharmax-vikas | 175,734,208 | U_kgDOCnl9wA | User | false |
huggingface/transformers | 4,723,272,401 | I_kwDOCUB6oc8AAAABGYdq0Q | 46,832 | https://github.com/huggingface/transformers/issues/46832 | https://api.github.com/repos/huggingface/transformers/issues/46832 | [RT-DETR] TypeError in RTDetrModel.forward when num_feature_levels > len(backbone outputs) | ### System Info
transformers version: 5.12.1
Platform: Windows-10-10.0.26200-SP0
Python version: 3.11.9
PyTorch version: 2.12.1+cpu
CUDA available: False
Note: transformers env crashed for me on 5.12.1 (NameError: CompletionCreateParamsStreaming), so system info was collected manually.
RTDetrForObjectDetection rais... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-06-23T06:34:53Z | 2026-06-29T12:22:50Z | 2026-06-29T12:22:50Z | CONTRIBUTOR | null | 20260629T180016Z | 2026-06-29T18:00:16Z | c1prk | 69,437,031 | MDQ6VXNlcjY5NDM3MDMx | User | false |
huggingface/transformers | 4,726,318,405 | I_kwDOCUB6oc8AAAABGbXlRQ | 46,846 | https://github.com/huggingface/transformers/issues/46846 | https://api.github.com/repos/huggingface/transformers/issues/46846 | [Qwen3.5] Missing `linear_attn` entries in `base_model_tp_plan` causes OOM and shape error at TP>1 | ### System Info
(Manually generated, working on an experimental framework)
transformers version: 5.12.1
Python version: 3.12.13
Huggingface_hub version: 1.20.1
Safetensors version: 0.8.0
Accelerate version: 1.14.0
Accelerate config: not found
DeepSpeed version: not installed
PyTorch version (accelerator?): 2.12.1 (NA)... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-06-23T13:29:27Z | 2026-07-03T13:59:02Z | 2026-07-03T13:59:02Z | MEMBER | null | 20260703T180022Z | 2026-07-03T18:00:22Z | tengomucho | 6,949,769 | MDQ6VXNlcjY5NDk3Njk= | User | false |
huggingface/transformers | 4,728,139,667 | I_kwDOCUB6oc8AAAABGdGvkw | 46,851 | https://github.com/huggingface/transformers/issues/46851 | https://api.github.com/repos/huggingface/transformers/issues/46851 | Qwen3.5 Does not work with variable length inputs using position_ids | ### System Info
The current implementation of Qwen3.5 linear attention assumes that user passes cu_seqlen arguments explicitly as opposed to just position_ids like older models. While the problem can be circumvented by passing these inputs, it becomes a problem in downstream packages (such as Verl) that assume the sam... | closed | completed | false | 8 | [
"bug"
] | [] | 2026-06-23T17:34:46Z | 2026-06-24T16:20:46Z | 2026-06-24T16:20:46Z | NONE | null | 20260624T180019Z | 2026-06-24T18:00:19Z | kolehma8 | 12,157,931 | MDQ6VXNlcjEyMTU3OTMx | User | false |
huggingface/transformers | 4,732,309,929 | I_kwDOCUB6oc8AAAABGhFRqQ | 46,858 | https://github.com/huggingface/transformers/issues/46858 | https://api.github.com/repos/huggingface/transformers/issues/46858 | _prefill: cache.get_seq_length() returns device tensor, breaks slice arithmetic for static cache + torch.compile on non-CUDA backends | ### System Info
- transformers version: 5.5.3
- PyTorch version: 2.10.0
- Python version: 3.10.12
- OS: Linux (Ubuntu 22.04)
- Device: CPU (using `torch.compile(backend="inductor")`)
- Last working version: transformers 5.2.0
- First broken version: transformers 5.5.3 (where _prefill was introduced)
##... | closed | completed | false | 6 | [
"bug",
"Code agent slop"
] | [] | 2026-06-24T06:33:39Z | 2026-06-25T12:42:46Z | 2026-06-25T12:42:34Z | NONE | null | 20260625T180018Z | 2026-06-25T18:00:18Z | silicon-dispatch-100 | 26,684,651 | MDQ6VXNlcjI2Njg0NjUx | User | false |
huggingface/transformers | 4,736,244,948 | I_kwDOCUB6oc8AAAABGk1c1A | 46,868 | https://github.com/huggingface/transformers/issues/46868 | https://api.github.com/repos/huggingface/transformers/issues/46868 | Sequence Parallelism Disabled for Evaluation | Hi, I am training a model with sequence parllelism as per these [docs](qwen 3.6 27b). I am run evals evals every 50 training steps, and encountered an OOM during my first eval step:
```
2026-06-24T08:21:11.619Z | 2%\|▏ \| 50/2064 [1:12:45<49:41:22, 88.82s/it]
-- | --
| 2026-06-24T08:21:11.619Z | {'loss': '0.6005', ... | open | null | false | 2 | [] | [] | 2026-06-24T15:34:17Z | 2026-07-11T20:56:20Z | null | NONE | null | 20260712T000022Z | 2026-07-12T00:00:22Z | amanichopra | 42,814,002 | MDQ6VXNlcjQyODE0MDAy | User | false |
huggingface/transformers | 4,736,609,563 | I_kwDOCUB6oc8AAAABGlLtGw | 46,869 | https://github.com/huggingface/transformers/issues/46869 | https://api.github.com/repos/huggingface/transformers/issues/46869 | YOLO model failing on multi GPU device map auto | ### System Info
Yolo model fails:
```python
(Desktop) theo@arthur-brahms:~$ /home/theo/Desktop/.venv/bin/python /home/theo/Desktop/reachy/test-vision.py
Loading weights: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████| 212/212 [00:00<00:00, 2564.02i... | open | null | false | 2 | [
"bug"
] | [] | 2026-06-24T16:25:08Z | 2026-06-26T15:42:41Z | null | NONE | null | 20260626T180014Z | 2026-06-26T18:00:14Z | Carotterap | 295,787,356 | U_kgDOEaFbXA | User | false |
huggingface/transformers | 1,669,574,064 | I_kwDOCUB6oc5jg62w | 22,786 | https://github.com/huggingface/transformers/issues/22786 | https://api.github.com/repos/huggingface/transformers/issues/22786 | Implement a decode method in transformers.BasicTokenizer | ### Feature request
Transformers has provided a nice BasicTokenizer for basic tokenizing when we don't need BPE tokenizers. For data processing (like data format converting), it is better to offer a decode method for basic use.
### Motivation
When doing data format converting in some data processing problems, we usu... | closed | completed | false | 3 | [
"Feature request"
] | [] | 2023-04-15T21:11:07Z | 2026-06-25T06:44:30Z | 2026-06-25T06:44:30Z | CONTRIBUTOR | null | 20260625T120018Z | 2026-06-25T12:00:18Z | jiangwangyi | 39,762,734 | MDQ6VXNlcjM5NzYyNzM0 | User | false |
huggingface/transformers | 4,742,221,542 | I_kwDOCUB6oc8AAAABGqiO5g | 46,877 | https://github.com/huggingface/transformers/issues/46877 | https://api.github.com/repos/huggingface/transformers/issues/46877 | auto_docstring prints [ERROR] doc-lint diagnostics to stdout at import time | ### System Info
- `transformers` version: 5.13.0.dev0
- Platform: Windows-11-10.0.26200-SP0
- Python version: 3.13.11
- PyTorch version (GPU?): 2.12.0+cu132 (CUDA 13.2, GPU available: yes)
- GPU: NVIDIA RTX 2000 Ada Generation Laptop GPU
- Accelerate version: 1.14.0
- Safetensors version: 0.8.0
### Who can help?
@st... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-06-25T08:49:25Z | 2026-06-25T14:56:10Z | 2026-06-25T14:56:10Z | NONE | null | 20260625T180018Z | 2026-06-25T18:00:18Z | CalamitousFelicitousness | 63,501,758 | MDQ6VXNlcjYzNTAxNzU4 | User | false |
huggingface/transformers | 4,746,440,848 | I_kwDOCUB6oc8AAAABGujwkA | 46,894 | https://github.com/huggingface/transformers/issues/46894 | https://api.github.com/repos/huggingface/transformers/issues/46894 | CohereASR training-loss double-shift bug (same pattern as Moonshine #46784) | ### System Info
- `transformers` version: 5.13.0.dev0 (main branch at commit dc5a497c18)
- - Platform: Windows
- - - Python version: 3.x
- - - - Using GPU: N/A (code-level bug, not runtime)
### Who can help?
@eustlb @ebezzam @vasqu
### Information
- [x] The official example scripts
- [ ] My own modified scripts
#... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-06-25T18:29:52Z | 2026-07-21T14:38:19Z | 2026-07-21T14:38:19Z | CONTRIBUTOR | null | 20260721T180022Z | 2026-07-21T18:00:22Z | sharmax-vikas | 175,734,208 | U_kgDOCnl9wA | User | false |
huggingface/transformers | 4,746,789,216 | I_kwDOCUB6oc8AAAABGu5BYA | 46,897 | https://github.com/huggingface/transformers/issues/46897 | https://api.github.com/repos/huggingface/transformers/issues/46897 | Florence2 training-loss double-shift bug (same pattern as Moonshine #46784) | ### System Info
- `transformers` version: 5.13.0.dev0 (main branch at commit dc5a497c18)
- - Platform: Windows
- - - Python version: 3.x
- - - - Using GPU: N/A (code-level bug, not runtime)
### Who can help?
@zucchini-nlp @yonigozlan
### Information
- [x] The official example scripts
- [ ] My own modified scripts
... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-06-25T19:24:39Z | 2026-06-29T10:51:17Z | 2026-06-29T10:51:17Z | CONTRIBUTOR | null | 20260629T120018Z | 2026-06-29T12:00:18Z | sharmax-vikas | 175,734,208 | U_kgDOCnl9wA | User | false |
huggingface/transformers | 4,746,887,182 | I_kwDOCUB6oc8AAAABGu_ADg | 46,899 | https://github.com/huggingface/transformers/issues/46899 | https://api.github.com/repos/huggingface/transformers/issues/46899 | [Gemma4] Audio (and vision) processing broken with 4‑bit quantization due to unsafe dtype cast in multimodal embedders | ### System Info
- `transformers` version: 5.12.1
- `bitsandbytes` version: 0.49.2
- Platform: Linux, Python 3.13, CUDA 12.x
- GPU: NVIDIA RTX 5090
- Quantization config: `BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16, bnb_4bit_use_... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-06-25T19:40:51Z | 2026-06-26T13:09:35Z | 2026-06-26T13:09:35Z | NONE | null | 20260626T180014Z | 2026-06-26T18:00:14Z | dukei | 1,279,262 | MDQ6VXNlcjEyNzkyNjI= | User | false |
huggingface/transformers | 4,748,781,069 | I_kwDOCUB6oc8AAAABGwymDQ | 46,901 | https://github.com/huggingface/transformers/issues/46901 | https://api.github.com/repos/huggingface/transformers/issues/46901 | PPFormulaNet training-loss double-shift: trains against labels[..., 1:] instead of labels | ### System Info
- `transformers` version: 5.13.0.dev0 (main, commit `c96378c413`)
- Platform: macOS, Python 3.11 (code-level bug, no specific hardware required)
### Who can help?
@yonigozlan @zucchini-nlp (PPFormulaNet / Florence2 area)
### Description
`PPFormulaNetForConditionalGeneration.forward` builds `decoder... | open | null | false | 0 | [] | [] | 2026-06-26T02:23:31Z | 2026-06-26T02:23:31Z | null | NONE | null | 20260626T060016Z | 2026-06-26T06:00:16Z | OmkumarSolanki | 144,753,825 | U_kgDOCKDEoQ | User | false |
huggingface/transformers | 4,750,295,040 | I_kwDOCUB6oc8AAAABGyPAAA | 46,906 | https://github.com/huggingface/transformers/issues/46906 | https://api.github.com/repos/huggingface/transformers/issues/46906 | [serge] integration failure triage - 2026-06-26 | <!-- serge-triage-run:integration-failure-triage:2026-06-26 -->
Automated **integration-failure triage** for the daily CI window `2026-06-20 → 2026-06-26`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 1 | [] | [] | 2026-06-26T07:52:15Z | 2026-07-07T05:53:50Z | 2026-07-07T05:53:50Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,750,425,129 | I_kwDOCUB6oc8AAAABGyW8KQ | 46,909 | https://github.com/huggingface/transformers/issues/46909 | https://api.github.com/repos/huggingface/transformers/issues/46909 | broken documentation example for DIA | ### System Info
Documentation for DIA is a bit weird:
https://huggingface.co/docs/transformers/model_doc/dia#generation-with-text
- torch device does not exist
- to device map auto does not exist
```python
from transformers import AutoProcessor, DiaForConditionalGeneration
model_checkpoint = "nari-labs/Dia-1.6B-0... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-06-26T08:12:30Z | 2026-06-26T18:22:10Z | 2026-06-26T18:01:27Z | NONE | null | 20260627T000016Z | 2026-06-27T00:00:16Z | Carotterap | 295,787,356 | U_kgDOEaFbXA | User | false |
huggingface/transformers | 2,919,858,677 | I_kwDOCUB6oc6uCYH1 | 36,718 | https://github.com/huggingface/transformers/issues/36718 | https://api.github.com/repos/huggingface/transformers/issues/36718 | Add RoMa keypoint matcher | ### Model description
RoMa is a state of the art deep keypoint detector and matcher. It falls in the "fully dense" matcher category, contrary to semi-dense matchers like LoFTR or Sparse matchers like Superpoint+SuperGlue.
When you want the best performance when you do not have many limitation from speed or memory, ... | open | null | false | 16 | [
"New model",
"Vision"
] | [] | 2025-03-14T10:42:45Z | 2026-07-14T23:02:20Z | null | NONE | null | 20260715T000048Z | 2026-07-15T00:00:48Z | tcourat | 82,943,048 | MDQ6VXNlcjgyOTQzMDQ4 | User | false |
huggingface/transformers | 4,757,712,146 | I_kwDOCUB6oc8AAAABG5TtEg | 46,931 | https://github.com/huggingface/transformers/issues/46931 | https://api.github.com/repos/huggingface/transformers/issues/46931 | ตัวจัดการบัญชีสำหรับ L. CHOMPHUPHUEN.io | <!--
Note: Please search to see if an issue already exists for the language you are trying to translate.
-->
Hi!
Let's bring the documentation to all the <languageName>-speaking community 🌐 (currently 0 out of 267 complete)
Who would want to translate? Please follow the 🤗 [TRANSLATING guide](https://github.com/hug... | closed | completed | false | 1 | [
"WIP"
] | [] | 2026-06-27T09:22:49Z | 2026-06-27T15:05:39Z | 2026-06-27T15:05:39Z | NONE | null | 20260627T180026Z | 2026-06-27T18:00:26Z | imgoingbacktheonlything239-ops | 269,549,381 | U_kgDOEBD_RQ | User | false |
huggingface/transformers | 4,758,250,685 | I_kwDOCUB6oc8AAAABG50kvQ | 46,932 | https://github.com/huggingface/transformers/issues/46932 | https://api.github.com/repos/huggingface/transformers/issues/46932 | [Gemma4] Fix dtype casting for quantized vision/audio embedders (same as #46904) | ### System Info
- `transformers` version: 4.52.0.dev0 (main branch)
- - Platform: Linux
- - - Python version: 3.10
- - - - PyTorch version: 2.x with bitsandbytes
### Who can help?
@zucchini-nlp @SunMarc
### Information
- [ ] The official example scripts
- [x] My own modified scripts
### Tasks
- [ ] An officially... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-06-27T12:12:13Z | 2026-07-03T12:21:54Z | 2026-07-03T12:21:54Z | CONTRIBUTOR | null | 20260703T180022Z | 2026-07-03T18:00:22Z | sharmax-vikas | 175,734,208 | U_kgDOCnl9wA | User | false |
huggingface/transformers | 2,542,978,567 | I_kwDOCUB6oc6XksYH | 33,664 | https://github.com/huggingface/transformers/issues/33664 | https://api.github.com/repos/huggingface/transformers/issues/33664 | Object detection training/fine-tuning for Owl-vit/Owlv2 | ### Feature request
Currently the Owl-vit models support inference and CLIP-style contrastive pre-training, but don't provide a way to train (or fine-tune) the detection part of the model. According to [the paper](https://arxiv.org/pdf/2205.06230), detection training is similar to Detr.
### Motivation
It would be re... | open | null | false | 16 | [
"Good Second Issue",
"Feature request",
"Vision"
] | [] | 2024-09-23T15:16:57Z | 2026-07-01T07:39:38Z | null | CONTRIBUTOR | null | 20260701T120019Z | 2026-07-01T12:00:19Z | catalys1 | 11,340,846 | MDQ6VXNlcjExMzQwODQ2 | User | false |
huggingface/transformers | 4,760,521,357 | I_kwDOCUB6oc8AAAABG7_KjQ | 46,940 | https://github.com/huggingface/transformers/issues/46940 | https://api.github.com/repos/huggingface/transformers/issues/46940 | [serge] integration failure triage - 2026-06-27 | <!-- serge-triage-run:integration-failure-triage:2026-06-27 -->
Automated **integration-failure triage** for the daily CI window `2026-06-21 → 2026-06-27`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-06-28T00:36:29Z | 2026-07-07T05:53:51Z | 2026-07-07T05:53:51Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,764,177,229 | I_kwDOCUB6oc8AAAABG_eTTQ | 46,948 | https://github.com/huggingface/transformers/issues/46948 | https://api.github.com/repos/huggingface/transformers/issues/46948 | [serge] integration failure triage - 2026-06-28 | <!-- serge-triage-run:integration-failure-triage:2026-06-28 -->
Automated **integration-failure triage** for the daily CI window `2026-06-22 → 2026-06-28`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 1 | [] | [] | 2026-06-29T00:38:07Z | 2026-07-07T05:53:51Z | 2026-07-07T05:53:51Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 3,107,599,219 | I_kwDOCUB6oc65OjNz | 38,518 | https://github.com/huggingface/transformers/issues/38518 | https://api.github.com/repos/huggingface/transformers/issues/38518 | Failed to export PyTorch traced graph of Mixtral-8x7B-Instruct-v0.1 due to the PR #32429 | ### System Info
Hi ,
I found the recently transformers 4.52.4 merged this PR https://github.com/huggingface/transformers/pull/32429
and it led me failed to run below code snippet which it could run successfully with 4.51.3.
```python
import transformers
import torch.export as te
import torch
from contextlib import ... | closed | completed | false | 5 | [
"Good Difficult Issue",
"bug"
] | [] | 2025-06-01T17:46:08Z | 2026-06-29T11:59:18Z | 2026-06-29T11:59:18Z | CONTRIBUTOR | null | 20260629T120018Z | 2026-06-29T12:00:18Z | nv-guomingz | 137,257,613 | U_kgDOCC5ijQ | User | false |
huggingface/transformers | 4,769,267,049 | I_kwDOCUB6oc8AAAABHEU9aQ | 46,962 | https://github.com/huggingface/transformers/issues/46962 | https://api.github.com/repos/huggingface/transformers/issues/46962 | [Qwen3] `create_masks_for_generate` return type change after #41251 | ### System Info
`transformers env` throw an error. I catch the system info with pytorch tools.
```
PyTorch version: 2.14.0.dev20260623+xpu
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A
OS: Ubuntu 26.04 LTS (x86_64)
GCC version: (Ubuntu 15.2.0-16ubuntu1) 15.2.0
Clang version: 2... | closed | completed | false | 9 | [
"bug"
] | [] | 2026-06-29T15:03:41Z | 2026-07-06T07:28:40Z | 2026-07-06T07:28:40Z | NONE | null | 20260706T120021Z | 2026-07-06T12:00:21Z | hoshibara | 108,672,484 | U_kgDOBno15A | User | false |
huggingface/transformers | 4,771,468,176 | I_kwDOCUB6oc8AAAABHGbTkA | 46,967 | https://github.com/huggingface/transformers/issues/46967 | https://api.github.com/repos/huggingface/transformers/issues/46967 | Add AstroCLIP model support | ### Feature request
I'd like to add AstroCLIP to Transformers.
AstroCLIP is a CLIP-style dual encoder for astronomy data. It aligns galaxy image
cutouts with galaxy spectra using a DINOv2 image encoder, a SpecFormer spectrum
encoder, cross-attention pooling heads and a symmetric contrastive loss.
Paper: https://hugg... | closed | completed | false | 5 | [] | [] | 2026-06-29T20:45:32Z | 2026-07-02T11:54:55Z | 2026-07-02T11:54:54Z | NONE | null | 20260702T120025Z | 2026-07-02T12:00:25Z | giovannicozzolongo | 79,092,266 | MDQ6VXNlcjc5MDkyMjY2 | User | false |
huggingface/transformers | 4,772,643,784 | I_kwDOCUB6oc8AAAABHHjDyA | 46,969 | https://github.com/huggingface/transformers/issues/46969 | https://api.github.com/repos/huggingface/transformers/issues/46969 | [serge] integration failure triage - 2026-06-29 | <!-- serge-triage-run:integration-failure-triage:2026-06-29 -->
Automated **integration-failure triage** for the daily CI window `2026-06-23 → 2026-06-29`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-06-30T00:36:50Z | 2026-07-07T05:53:52Z | 2026-07-07T05:53:52Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,773,093,103 | I_kwDOCUB6oc8AAAABHH-e7w | 46,972 | https://github.com/huggingface/transformers/issues/46972 | https://api.github.com/repos/huggingface/transformers/issues/46972 | Add SuperCompress prompt compression for efficient inference | ## SuperCompress: Open-Source Prompt Compression
[SuperCompress](https://github.com/arjunkshah/supercompress) is a learned prompt compression library that reduces LLM token usage by ~65% using a ~5K parameter neural policy.
**Why this matters for ML development:**
- Practical ML deployment: a tiny model (~5K params) ... | closed | completed | false | 1 | [] | [] | 2026-06-30T02:17:27Z | 2026-06-30T04:07:18Z | 2026-06-30T04:07:18Z | NONE | null | 20260630T060014Z | 2026-06-30T06:00:14Z | arjunkshah | 180,205,446 | U_kgDOCr23hg | User | false |
huggingface/transformers | 4,777,658,074 | I_kwDOCUB6oc8AAAABHMVG2g | 46,981 | https://github.com/huggingface/transformers/issues/46981 | https://api.github.com/repos/huggingface/transformers/issues/46981 | AutoProcessor.from_pretrained retries HTTP 429 responses multiple times for every config file when using HF_ENDPOINT | ### System Info
(uv run transformers env fails with name « CompletionCreateParamsStreaming » is not defined, but this is a separate issue)
- transformers 5.12.1
- windows 11
### Who can help?
@Cyrilvallez
### Information
- [x] The official example scripts
- [ ] My own modified scripts
### Tasks
- [x] An offi... | closed | completed | false | 5 | [
"bug"
] | [] | 2026-06-30T14:27:18Z | 2026-07-09T02:31:36Z | 2026-07-09T02:31:36Z | NONE | null | 20260709T060021Z | 2026-07-09T06:00:21Z | peacefulotter | 32,218,033 | MDQ6VXNlcjMyMjE4MDMz | User | false |
huggingface/transformers | 4,780,614,010 | I_kwDOCUB6oc8AAAABHPJheg | 46,983 | https://github.com/huggingface/transformers/issues/46983 | https://api.github.com/repos/huggingface/transformers/issues/46983 | Add RWKV7 model support | ### Feature request
### Model description
I would like to propose adding RWKV7 support to Transformers.
RWKV7 is the latest generation of the RWKV recurrent language model family. It is a causal language model with recurrent state updates, but it is not compatible with the existing `rwkv` implementation in Transform... | closed | not_planned | false | 0 | [
"Feature request"
] | [] | 2026-06-30T21:43:21Z | 2026-06-30T21:44:14Z | 2026-06-30T21:44:14Z | CONTRIBUTOR | null | 20260701T000029Z | 2026-07-01T00:00:29Z | chinoll | 22,575,232 | MDQ6VXNlcjIyNTc1MjMy | User | false |
huggingface/transformers | 4,780,626,289 | I_kwDOCUB6oc8AAAABHPKRcQ | 46,984 | https://github.com/huggingface/transformers/issues/46984 | https://api.github.com/repos/huggingface/transformers/issues/46984 | Add RWKV7 model support | ### Model description
I would like to propose adding RWKV7 support to Transformers.
RWKV7 is the latest generation of the RWKV recurrent language model family. It is a causal language model with recurrent state updates, but it is not compatible with the existing `rwkv` implementation in Transformers, which targets ea... | closed | completed | false | 2 | [
"New model"
] | [] | 2026-06-30T21:45:42Z | 2026-07-08T09:11:43Z | 2026-07-08T09:11:43Z | CONTRIBUTOR | null | 20260708T120111Z | 2026-07-08T12:01:11Z | chinoll | 22,575,232 | MDQ6VXNlcjIyNTc1MjMy | User | false |
huggingface/transformers | 4,781,386,510 | I_kwDOCUB6oc8AAAABHP4rDg | 46,985 | https://github.com/huggingface/transformers/issues/46985 | https://api.github.com/repos/huggingface/transformers/issues/46985 | [serge] integration failure triage - 2026-06-30 | <!-- serge-triage-run:integration-failure-triage:2026-06-30 -->
Automated **integration-failure triage** for the daily CI window `2026-06-24 → 2026-06-30`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-07-01T00:38:22Z | 2026-07-07T05:53:52Z | 2026-07-07T05:53:52Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,784,614,119 | I_kwDOCUB6oc8AAAABHS9q5w | 46,996 | https://github.com/huggingface/transformers/issues/46996 | https://api.github.com/repos/huggingface/transformers/issues/46996 | Reporting a vulnerability | @Michellehbn @ArthurZucker
need access to the Huggingface Private Bug Bounty Program
https://hackerone.com/hugging_face
I request the team to send me an invite for the Huggingface Private Bug Bounty Program
My Hackerone username: raz_26
I had already email to the security team but not yet get any response
Hope th... | closed | completed | false | 1 | [] | [] | 2026-07-01T10:19:42Z | 2026-07-01T10:27:55Z | 2026-07-01T10:27:55Z | NONE | null | 20260701T120019Z | 2026-07-01T12:00:19Z | Eti-Hack | 173,596,630 | U_kgDOCljf1g | User | false |
huggingface/transformers | 4,785,639,574 | I_kwDOCUB6oc8AAAABHT8Qlg | 47,000 | https://github.com/huggingface/transformers/issues/47000 | https://api.github.com/repos/huggingface/transformers/issues/47000 | RTDetrLoss crashing with "ValueError: cost matrix is infeasible" | ### System Info
`transformers` version: 5.13.0.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: 0.8.0
- Accelerate version: 1.13.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorc... | open | null | false | 7 | [
"bug"
] | [] | 2026-07-01T12:45:17Z | 2026-07-04T11:05:30Z | null | CONTRIBUTOR | null | 20260704T120019Z | 2026-07-04T12:00:19Z | m-matthias | 16,415,097 | MDQ6VXNlcjE2NDE1MDk3 | User | false |
huggingface/transformers | 4,786,848,896 | I_kwDOCUB6oc8AAAABHVGEgA | 47,004 | https://github.com/huggingface/transformers/issues/47004 | https://api.github.com/repos/huggingface/transformers/issues/47004 | Add LoMa keypoint detector & matcher | ### Model description
LoMa revisits local (sparse) feature matching from a data-driven perspective. In particular, the authors focus on (i) curating a large and diverse training data mixture together with scalable training recipes for both descriptors and matchers, and (ii) increasing training compute along two axes: ... | open | null | false | 6 | [
"New model"
] | [] | 2026-07-01T15:15:57Z | 2026-07-18T22:04:41Z | null | NONE | null | 20260719T000028Z | 2026-07-19T00:00:28Z | tcourat | 82,943,048 | MDQ6VXNlcjgyOTQzMDQ4 | User | false |
huggingface/transformers | 4,785,329,762 | I_kwDOCUB6oc8AAAABHTpWYg | 46,998 | https://github.com/huggingface/transformers/issues/46998 | https://api.github.com/repos/huggingface/transformers/issues/46998 | [serge] integration failure triage - 2026-07-01 | <!-- serge-triage-run:integration-failure-triage:2026-07-01 -->
Automated **integration-failure triage** for the daily CI window `2026-06-25 → 2026-07-01`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-07-01T12:03:11Z | 2026-07-07T05:53:52Z | 2026-07-07T05:53:52Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,787,791,381 | I_kwDOCUB6oc8AAAABHV_mFQ | 47,007 | https://github.com/huggingface/transformers/issues/47007 | https://api.github.com/repos/huggingface/transformers/issues/47007 | Building AI-Automated Vulnerability Discovery Systems - Lessons from Using Advanced LLMs | Hi Hugging Face team,
Transformers has been foundational to the AI revolution. I want to share how we're leveraging advanced language models for an unexpected application: **automated cybersecurity vulnerability discovery**.
**Project**: 0day Rubbish
**Repository**: https://github.com/Exploit-Garbage/0day-Rubbish
... | closed | completed | false | 0 | [] | [] | 2026-07-01T17:22:52Z | 2026-07-02T11:35:14Z | 2026-07-02T11:35:14Z | NONE | null | 20260702T120025Z | 2026-07-02T12:00:25Z | Exploit-Garbage | 298,631,731 | U_kgDOEczCMw | User | false |
huggingface/transformers | 3,747,138,333 | I_kwDOCUB6oc7fWMsd | 42,954 | https://github.com/huggingface/transformers/issues/42954 | https://api.github.com/repos/huggingface/transformers/issues/42954 | XiaoMi MiMo | ### Model description
- https://huggingface.co/XiaomiMiMo/MiMo-V2-Flash/tree/main
### Open source status
- [x] The model implementation is available
- [x] The model weights are available
### Provide useful links for the implementation
_No response_ | closed | completed | false | 2 | [
"New model"
] | [] | 2025-12-19T11:15:41Z | 2026-07-01T18:50:10Z | 2026-07-01T18:50:10Z | MEMBER | null | 20260702T000029Z | 2026-07-02T00:00:29Z | ArthurZucker | 48,595,927 | MDQ6VXNlcjQ4NTk1OTI3 | User | false |
huggingface/transformers | 4,789,408,456 | I_kwDOCUB6oc8AAAABHXiSyA | 47,013 | https://github.com/huggingface/transformers/issues/47013 | https://api.github.com/repos/huggingface/transformers/issues/47013 | [xLSTM] Crash in train mode with return_last_states=False: backend returns bare tensor but xLSTMLayer always unpacks (h, state) | ### System Info
- `transformers` version: 5.12.1 (bug also present on current `main`)
- Platform: Windows 11 (also reproducible on Linux)
- Python version: 3.13
- PyTorch version: 2.11.0+cu128
### Who can help?
@vasqu @ArthurZucker
### Information
- [ ] The official example scripts
- [x] My own modified scripts
#... | open | null | false | 3 | [] | [] | 2026-07-01T21:42:33Z | 2026-07-05T05:21:39Z | null | NONE | null | 20260705T060022Z | 2026-07-05T06:00:22Z | Heatdh | 74,280,434 | MDQ6VXNlcjc0MjgwNDM0 | User | false |
huggingface/transformers | 4,790,961,338 | I_kwDOCUB6oc8AAAABHZBEug | 47,015 | https://github.com/huggingface/transformers/issues/47015 | https://api.github.com/repos/huggingface/transformers/issues/47015 | Novel Application of Transformers: AI-Automated Cybersecurity Vulnerability Discovery System | Hi Hugging Face team,
Transformers has revolutionized NLP and beyond. I want to share an unexpected but impactful application: **using transformer-based LLMs for automated cybersecurity vulnerability discovery**.
**Project**: 0day Rubbish
**Repository**: https://github.com/Exploit-Garbage/0day-Rubbish
**What We've... | closed | completed | false | 0 | [
"Code agent slop"
] | [] | 2026-07-02T03:22:07Z | 2026-07-02T11:44:54Z | 2026-07-02T11:44:42Z | NONE | null | 20260702T120025Z | 2026-07-02T12:00:25Z | Exploit-Garbage | 298,631,731 | U_kgDOEczCMw | User | false |
huggingface/transformers | 4,792,845,887 | I_kwDOCUB6oc8AAAABHa0GPw | 47,020 | https://github.com/huggingface/transformers/issues/47020 | https://api.github.com/repos/huggingface/transformers/issues/47020 | XLMRobertaTokenizer.__init__ passes dict to Unigram(vocab=...) expecting a Sequence | ## System Info
```
transformers version: 5.12.1
tokenizers version: 0.22.2
platform: macOS (also reproduced on Linux)
python version: 3.12 / 3.14
```
## Who can help?
@ArthurZucker @itazap
## Information
- [x] My own modified scripts
## Tasks
- [x] My own task or dataset (give details below)
## Reproduction
PR... | closed | completed | false | 4 | [] | [] | 2026-07-02T08:54:20Z | 2026-07-08T10:26:33Z | 2026-07-08T10:26:33Z | CONTRIBUTOR | null | 20260708T120111Z | 2026-07-08T12:01:11Z | saattrupdan | 47,701,536 | MDQ6VXNlcjQ3NzAxNTM2 | User | false |
huggingface/transformers | 4,798,445,417 | I_kwDOCUB6oc8AAAABHgJ3aQ | 47,030 | https://github.com/huggingface/transformers/issues/47030 | https://api.github.com/repos/huggingface/transformers/issues/47030 | FineGrainedFP8 DeepGEMM path returns silently wrong results on SM100 (UE8M0 scale rounding without requantization) | ## System info
- transformers 5.12.1, torch 2.10.0+cu130, kernels 0.12.3
- Hub kernel: `kernels-community/deep-gemm`, pinned version 2 (rev `9590415046fa`, internal `__version__` 2.5.0)
- GPU: NVIDIA B200 (SM100), CUDA 13.0
- Model: `Qwen/Qwen3-4B-FP8` (fine-grained FP8, dynamic activation scheme, fp32 `weight_scale_i... | open | null | false | 1 | [] | [] | 2026-07-02T21:54:50Z | 2026-07-03T01:05:47Z | null | NONE | null | 20260703T060023Z | 2026-07-03T06:00:23Z | keunhong | 297,680 | MDQ6VXNlcjI5NzY4MA== | User | false |
huggingface/transformers | 4,798,543,956 | I_kwDOCUB6oc8AAAABHgP4VA | 47,031 | https://github.com/huggingface/transformers/issues/47031 | https://api.github.com/repos/huggingface/transformers/issues/47031 | Checkpoint conversion silently loads transposed fused-expert weights when the swapped dims are equal (Transpose(check_dims=True), qwen3_vl_moe) |
### System info
transformers 5.12.1 (code-identical on `main` as of 2026-07-02: `core_model_loading.py`
`Transpose.convert`, `conversion_mapping.py` qwen3_vl_moe entries), torch 2.11.0. CPU-only repro.
### Bug
`Transpose(1, 2, check_dims=True)` — used by the qwen3_vl_moe conversion to accept both the legacy
pre-gro... | open | null | false | 2 | [] | [] | 2026-07-02T22:14:35Z | 2026-07-03T13:09:00Z | null | NONE | null | 20260703T180022Z | 2026-07-03T18:00:22Z | pjordanandrsn | 279,818,045 | U_kgDOEK2vPQ | User | false |
huggingface/transformers | 4,798,942,856 | I_kwDOCUB6oc8AAAABHgoOiA | 47,036 | https://github.com/huggingface/transformers/issues/47036 | https://api.github.com/repos/huggingface/transformers/issues/47036 | Community benchmark tip | spam | closed | completed | false | 1 | [] | [] | 2026-07-02T23:48:33Z | 2026-07-03T10:38:26Z | 2026-07-03T04:11:30Z | NONE | null | 20260703T120023Z | 2026-07-03T12:00:23Z | connerlambden | 9,061,871 | MDQ6VXNlcjkwNjE4NzE= | User | false |
huggingface/transformers | 4,799,113,639 | I_kwDOCUB6oc8AAAABHgyppw | 47,037 | https://github.com/huggingface/transformers/issues/47037 | https://api.github.com/repos/huggingface/transformers/issues/47037 | [serge] integration failure triage - 2026-07-02 | <!-- serge-triage-run:integration-failure-triage:2026-07-02 -->
Automated **integration-failure triage** for the daily CI window `2026-06-26 → 2026-07-02`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 5 | [] | [] | 2026-07-03T00:31:21Z | 2026-07-07T05:53:53Z | 2026-07-07T05:53:53Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,802,412,788 | I_kwDOCUB6oc8AAAABHj8A9A | 47,039 | https://github.com/huggingface/transformers/issues/47039 | https://api.github.com/repos/huggingface/transformers/issues/47039 | GenerationConfig serializes ContinuousBatchingConfig as null | ### System Info
Observed on the local development checkout used for PR #47038:
```text
transformers main branch at 9d011e8375ab802f1b9ada2d488af35232f6dd3e
Python 3.12 in a local conda environment on Windows
```
Relevant local validation for the proposed fix has passed:
```text
python -m pytest tests/generation/tes... | open | null | false | 2 | [] | [] | 2026-07-03T09:48:39Z | 2026-07-08T16:32:49Z | null | NONE | null | 20260708T180046Z | 2026-07-08T18:00:46Z | VectorPeak | 73,048,950 | MDQ6VXNlcjczMDQ4OTUw | User | false |
huggingface/transformers | 4,807,034,547 | I_kwDOCUB6oc8AAAABHoWGsw | 47,049 | https://github.com/huggingface/transformers/issues/47049 | https://api.github.com/repos/huggingface/transformers/issues/47049 | [serge] integration failure triage - 2026-07-03 | <!-- serge-triage-run:integration-failure-triage:2026-07-03 -->
Automated **integration-failure triage** for the daily CI window `2026-06-27 → 2026-07-03`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-07-04T00:30:38Z | 2026-07-07T05:53:54Z | 2026-07-07T05:53:53Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,807,434,362 | I_kwDOCUB6oc8AAAABHougeg | 47,050 | https://github.com/huggingface/transformers/issues/47050 | https://api.github.com/repos/huggingface/transformers/issues/47050 | CanineModel fp16 ONNX export fails with dtype mismatch in attention (expected scalar type Float but found Half) | ### System Info
- `transformers` version: 5.12.1
- Platform: Linux-6.18.18-trim-x86_64-with-glibc2.41
- Python version: 3.12.13
- Huggingface_hub version: 1.22.0
- Safetensors version: 0.8.0
- Accelerate version: 1.14.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator... | closed | completed | false | 3 | [
"bug"
] | [] | 2026-07-04T03:12:04Z | 2026-07-06T13:43:47Z | 2026-07-06T13:43:47Z | NONE | null | 20260706T180017Z | 2026-07-06T18:00:17Z | sgpublic | 37,202,870 | MDQ6VXNlcjM3MjAyODcw | User | false |
huggingface/transformers | 4,808,113,319 | I_kwDOCUB6oc8AAAABHpX8pw | 47,053 | https://github.com/huggingface/transformers/issues/47053 | https://api.github.com/repos/huggingface/transformers/issues/47053 | [serge] integration failure triage - 2026-07-04 | <!-- serge-triage-run:integration-failure-triage:2026-07-04 -->
Automated **integration-failure triage** for the daily CI window `2026-06-28 → 2026-07-04`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-07-04T07:33:00Z | 2026-07-07T05:53:54Z | 2026-07-07T05:53:54Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,808,340,460 | I_kwDOCUB6oc8AAAABHplz7A | 47,059 | https://github.com/huggingface/transformers/issues/47059 | https://api.github.com/repos/huggingface/transformers/issues/47059 | tests/cli fail with huggingface_hub 1.22.0: 'HFCliTyperGroup' object has no attribute '_add_completion': | ### System Info
transformers: main (5.14.0.dev0)
huggingface_hub: 1.22.0
typer: 0.26.8 / click: 8.4.2 (also reproduces with typer 0.24.1)
Python 3.10 (CI docker image) — also reproduced on Python 3.11 / Windows
Note: `transformers env` itself crashes with this bug, so output is unavailable.
### Who can help?
_No res... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-07-04T08:56:14Z | 2026-07-07T09:21:07Z | 2026-07-07T09:21:07Z | NONE | null | 20260707T120022Z | 2026-07-07T12:00:22Z | Madhav-000-s | 181,208,194 | U_kgDOCs0Egg | User | false |
huggingface/transformers | 4,808,745,614 | I_kwDOCUB6oc8AAAABHp-ijg | 47,065 | https://github.com/huggingface/transformers/issues/47065 | https://api.github.com/repos/huggingface/transformers/issues/47065 | DeformableDetrHungarianMatcher and LwDetrHungarianMatcher crash on non-finite cost matrices ("cost matrix is infeasible") | ### System Info
- `transformers` 5.14.0.dev0 (main)
- PyTorch 2.x
- Reproduces on CPU; originally hit on GPU with AMP
### Who can help?
@guarin @yonigozlan
### Description
Same failure mode as #47000 (RT-DETR), but in two more Hungarian matchers that are still unguarded on `main`. Under AMP/fp16 the sigmoid focal ... | closed | completed | false | 0 | [
"Code agent slop"
] | [] | 2026-07-04T11:21:32Z | 2026-07-06T13:01:45Z | 2026-07-06T13:01:40Z | NONE | null | 20260706T180017Z | 2026-07-06T18:00:17Z | MutugiD | 61,890,059 | MDQ6VXNlcjYxODkwMDU5 | User | false |
huggingface/transformers | 4,808,946,457 | I_kwDOCUB6oc8AAAABHqKzGQ | 47,067 | https://github.com/huggingface/transformers/issues/47067 | https://api.github.com/repos/huggingface/transformers/issues/47067 | [i18n-<languageCode>] Translating docs to <languageName> | Hi! I’m a native Japanese speaker and I’d like to help translate the Japanese documentation.
Can I work on docs/source/en/model_doc/distilbert.md?
I’ll add the Japanese version under docs/source/ja/model_doc/ and update docs/source/ja/_toctree.yml. | closed | completed | false | 2 | [
"WIP"
] | [] | 2026-07-04T12:21:47Z | 2026-07-06T15:04:54Z | 2026-07-06T15:04:54Z | NONE | null | 20260706T180017Z | 2026-07-06T18:00:17Z | VIUK-XV | 216,173,586 | U_kgDODOKMEg | User | false |
huggingface/transformers | 4,809,481,408 | I_kwDOCUB6oc8AAAABHqrcwA | 47,070 | https://github.com/huggingface/transformers/issues/47070 | https://api.github.com/repos/huggingface/transformers/issues/47070 | ASR pipeline: unconditional torchcodec import crashes on broken torchcodec installs | ### System Info
- transformers 5.12.1, torch 2.10.0, torchcodec 0.14.0 (installed as a dependency of pyannote-audio 4.0.7)
- macOS 15.6 (arm64), Python 3.11.12
- Same failure mode applies to any platform where the installed torchcodec cannot be imported.
### Reproduction
`AutomaticSpeechRecognitionPipeline.preproces... | open | null | false | 2 | [] | [] | 2026-07-04T15:08:32Z | 2026-07-14T17:06:07Z | null | NONE | null | 20260714T180038Z | 2026-07-14T18:00:38Z | evnsnclr | 139,897,548 | U_kgDOCFaqzA | User | false |
huggingface/transformers | 4,809,134,969 | I_kwDOCUB6oc8AAAABHqWTeQ | 47,068 | https://github.com/huggingface/transformers/issues/47068 | https://api.github.com/repos/huggingface/transformers/issues/47068 | Potential host synchronization in deepspeed_sp_compute_loss due to Python conditional on CUDA tensor | ### System Info
- Transformers: 4.57.2
- PyTorch: 2.7.1+cu128
- DeepSpeed: 0.17.6
- CUDA: 12.8
- GPU: NVIDIA B300 SXM6 (2 nodes × 8 GPUs, 16 GPUs total)
- Interconnect:
- Intra-node: NVSwitch
- Inter-node: InfiniBand
### Information
While profiling a distributed training workload with Nsight Systems, I observed ... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-07-04T13:24:14Z | 2026-07-10T14:04:16Z | 2026-07-10T14:04:16Z | CONTRIBUTOR | null | 20260710T180030Z | 2026-07-10T18:00:30Z | hdimmfh | 74,033,655 | MDQ6VXNlcjc0MDMzNjU1 | User | false |
huggingface/transformers | 4,812,538,665 | I_kwDOCUB6oc8AAAABHtmDKQ | 47,085 | https://github.com/huggingface/transformers/issues/47085 | https://api.github.com/repos/huggingface/transformers/issues/47085 | Fast (Rust-backed) tokenizers aren't thread-safe anymore | ### System Info
transformers 5.5.4
Python 3.12.3
### Who can help?
text models: @ArthurZucker @Cyrilvallez @vasqu
### 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 ... | open | null | false | 8 | [
"bug"
] | [] | 2026-07-05T10:57:08Z | 2026-07-20T15:40:09Z | null | NONE | null | 20260721T000056Z | 2026-07-21T00:00:56Z | dxqb | 183,307,934 | U_kgDOCu0Ong | User | false |
huggingface/transformers | 4,812,866,664 | I_kwDOCUB6oc8AAAABHt6EaA | 47,086 | https://github.com/huggingface/transformers/issues/47086 | https://api.github.com/repos/huggingface/transformers/issues/47086 | Hybrid models: padding is not masked out of linear-attention/conv state on continued forwards (chunked prefill, cache continuation) | ### System Info
- `transformers` main (5.14.0.dev0, commit b70d02fc)
- torch 2.12.1, CPU, float32 (fla / causal_conv1d not installed, so all numbers below use the torch fallback path; the skipped masking happens before kernel dispatch, so the fast paths receive the same unmasked hidden states)
- Python 3.12
### Who c... | closed | completed | false | 2 | [
"bug"
] | [] | 2026-07-05T12:41:27Z | 2026-07-20T17:00:12Z | 2026-07-20T17:00:12Z | CONTRIBUTOR | null | 20260721T000056Z | 2026-07-21T00:00:56Z | abcgco | 53,853,163 | MDQ6VXNlcjUzODUzMTYz | User | false |
huggingface/transformers | 4,815,054,604 | I_kwDOCUB6oc8AAAABHv_nDA | 47,091 | https://github.com/huggingface/transformers/issues/47091 | https://api.github.com/repos/huggingface/transformers/issues/47091 | [serge] integration failure triage - 2026-07-05 | <!-- serge-triage-run:integration-failure-triage:2026-07-05 -->
Automated **integration-failure triage** for the daily CI window `2026-06-29 → 2026-07-05`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-07-06T00:32:08Z | 2026-07-07T05:53:54Z | 2026-07-07T05:53:54Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,816,035,994 | I_kwDOCUB6oc8AAAABHw7gmg | 47,095 | https://github.com/huggingface/transformers/issues/47095 | https://api.github.com/repos/huggingface/transformers/issues/47095 | [Proposal] HeartFlow - Cognitive Engine for HuggingFace Transformers | ## Proposal: HeartFlow × HuggingFace Transformers
### What is HeartFlow?
HeartFlow is an **AI cognitive engine** (68 modules) that adds judgment to any HuggingFace model:
```
├── Three-Layer Memory # Context persistence
├── Decision Routing # 34 rules → 8 strategies
├── U/D/A/H Field # Cognitive healt... | closed | completed | false | 0 | [
"Code agent slop"
] | [] | 2026-07-06T04:46:55Z | 2026-07-06T12:19:19Z | 2026-07-06T12:17:53Z | NONE | null | 20260706T180017Z | 2026-07-06T18:00:17Z | yun520-1 | 251,003,022 | U_kgDODvYAjg | User | false |
huggingface/transformers | 4,817,505,329 | I_kwDOCUB6oc8AAAABHyVMMQ | 47,102 | https://github.com/huggingface/transformers/issues/47102 | https://api.github.com/repos/huggingface/transformers/issues/47102 | [serge] integration failure triage - 2026-07-06 | <!-- serge-triage-run:integration-failure-triage:2026-07-06 -->
Automated **integration-failure triage** for the daily CI window `2026-06-30 → 2026-07-06`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | completed | false | 0 | [] | [] | 2026-07-06T08:52:33Z | 2026-07-07T05:53:55Z | 2026-07-07T05:53:55Z | CONTRIBUTOR | null | 20260707T060715Z | 2026-07-07T06:07:15Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 3,225,463,136 | I_kwDOCUB6oc7AQKlg | 39,379 | https://github.com/huggingface/transformers/issues/39379 | https://api.github.com/repos/huggingface/transformers/issues/39379 | Handling of full_text_row_masked_out_mask in mllama is incorrect. | ### System Info
N/A
### Who can help?
@ArthurZucker
### 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
In MllamaCross... | open | null | false | 11 | [
"WIP",
"bug"
] | [] | 2025-07-12T15:07:43Z | 2026-07-06T08:53:10Z | null | NONE | null | 20260706T120021Z | 2026-07-06T12:00:21Z | nshepperd | 406,019 | MDQ6VXNlcjQwNjAxOQ== | User | false |
huggingface/transformers | 4,819,776,885 | I_kwDOCUB6oc8AAAABH0f1dQ | 47,110 | https://github.com/huggingface/transformers/issues/47110 | https://api.github.com/repos/huggingface/transformers/issues/47110 | special_tokens, 'list' object has no attribute 'keys' | ### System Info
In `tokenization_utils_base.py` line 1190, there is a `special_tokens.keys()` done. But `special_tokens` is typed as `List[str]`.
And when using it, I have `special_tokens` being equal to `['<s>NOTUSED', '</s>NOTUSED', '<unk>NOTUSED']` this lead to the error `'list' object has no attribute 'keys'`
##... | open | null | false | 3 | [
"bug"
] | [] | 2026-07-06T13:50:19Z | 2026-07-15T17:38:07Z | null | NONE | null | 20260715T180032Z | 2026-07-15T18:00:32Z | Ratinax | 72,433,771 | MDQ6VXNlcjcyNDMzNzcx | User | false |
huggingface/transformers | 3,283,629,239 | I_kwDOCUB6oc7DuDS3 | 39,842 | https://github.com/huggingface/transformers/issues/39842 | https://api.github.com/repos/huggingface/transformers/issues/39842 | Expected behavior of `compute_result` is hard to expect and inconsistent | In trainer there exists a parameter `compute_result` given to `compute_metrics` when `batch_eval_metrics` is given to True.
https://github.com/huggingface/transformers/blob/1e0665a191f73f6b002209c3dfcda478baac6bac/src/transformers/trainer.py#L370-L375
I think there are several problems for `compute_result`,
1. User c... | closed | completed | false | 4 | [] | [] | 2025-08-01T11:43:28Z | 2026-07-06T14:51:55Z | 2025-10-04T08:02:41Z | CONTRIBUTOR | null | 20260706T180017Z | 2026-07-06T18:00:17Z | MilkClouds | 26,109,705 | MDQ6VXNlcjI2MTA5NzA1 | User | false |
huggingface/transformers | 4,822,621,942 | I_kwDOCUB6oc8AAAABH3Ne9g | 47,116 | https://github.com/huggingface/transformers/issues/47116 | https://api.github.com/repos/huggingface/transformers/issues/47116 | Streaming Pipeline Regression in Transformers v5 | ### System Info
Transformers 5+
### Who can help?
@Rocketknight1 @ArthurZucker
### 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... | closed | completed | false | 4 | [
"bug"
] | [] | 2026-07-06T19:56:06Z | 2026-07-09T12:00:58Z | 2026-07-09T12:00:58Z | CONTRIBUTOR | null | 20260709T180026Z | 2026-07-09T18:00:26Z | davidmezzetti | 561,939 | MDQ6VXNlcjU2MTkzOQ== | User | false |
huggingface/transformers | 4,825,824,362 | I_kwDOCUB6oc8AAAABH6Q8ag | 47,120 | https://github.com/huggingface/transformers/issues/47120 | https://api.github.com/repos/huggingface/transformers/issues/47120 | [serge] integration failure triage - 2026-07-07 | <!-- serge-triage-run:integration-failure-triage:2026-07-07 -->
Automated **integration-failure triage** for the daily CI window `2026-07-01 → 2026-07-07`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-07T06:19:36Z | 2026-07-09T00:31:28Z | 2026-07-09T00:31:28Z | CONTRIBUTOR | null | 20260709T060021Z | 2026-07-09T06:00:21Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,828,489,779 | I_kwDOCUB6oc8AAAABH8zoMw | 47,138 | https://github.com/huggingface/transformers/issues/47138 | https://api.github.com/repos/huggingface/transformers/issues/47138 | transformers chat routes management endpoints under /v1 when base_url includes /v1 | ### System Info
- `transformers` source checkout: current `main`
- Affected command: `transformers chat`
- Related local serve endpoints:
- OpenAI-compatible inference API: `/v1/...`
- management endpoints: `/health`, `/load_model`
### Who can help?
_No response_
### Information
- [ ] The official example scr... | open | null | false | 1 | [] | [] | 2026-07-07T12:27:38Z | 2026-07-13T12:25:26Z | null | NONE | null | 20260713T180023Z | 2026-07-13T18:00:23Z | VectorPeak | 73,048,950 | MDQ6VXNlcjczMDQ4OTUw | User | false |
huggingface/transformers | 4,828,492,886 | I_kwDOCUB6oc8AAAABH8z0Vg | 47,139 | https://github.com/huggingface/transformers/issues/47139 | https://api.github.com/repos/huggingface/transformers/issues/47139 | transformers chat routes management endpoints under /v1 when base_url includes /v1 | ### System Info
- `transformers` source checkout: current `main`
- Affected command: `transformers chat`
- Related local serve endpoints:
- OpenAI-compatible inference API: `/v1/...`
- management endpoints: `/health`, `/load_model`
### Who can help?
_No response_
### Information
- [ ] The official example scr... | open | null | false | 0 | [] | [] | 2026-07-07T12:28:03Z | 2026-07-07T12:28:03Z | null | NONE | null | 20260707T180018Z | 2026-07-07T18:00:18Z | VectorPeak | 73,048,950 | MDQ6VXNlcjczMDQ4OTUw | User | false |
huggingface/transformers | 4,831,008,795 | I_kwDOCUB6oc8AAAABH_NYGw | 47,145 | https://github.com/huggingface/transformers/issues/47145 | https://api.github.com/repos/huggingface/transformers/issues/47145 | Silent corruption: FP8 compressed-tensors "frozen" checkpoints load without applying weight_scale | ### System Info
```text
- `transformers` version: 5.13.0
- Platform: Linux-6.8.0-111-generic-x86_64-with-glibc2.35
- Python version: 3.12.13
- Huggingface_hub version: 1.22.0
- Safetensors version: 0.8.0
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch v... | open | null | false | 3 | [
"bug"
] | [] | 2026-07-07T18:00:05Z | 2026-07-08T20:05:15Z | null | NONE | null | 20260709T000028Z | 2026-07-09T00:00:28Z | kvcache670 | 264,492,208 | U_kgDOD8PUsA | User | false |
huggingface/transformers | 4,835,941,878 | I_kwDOCUB6oc8AAAABID6d9g | 47,169 | https://github.com/huggingface/transformers/issues/47169 | https://api.github.com/repos/huggingface/transformers/issues/47169 | BarkModel.enable_cpu_offload() strands submodules on CPU → device mismatch at generate() (transformers 4.57.6 + accelerate 1.13.0) | ### System Info
- `transformers` version: 4.57.6
- `accelerate` version: 1.13.0
- `torch` version: 2.12.1 (CUDA 13.0), CUDA available: yes
- Platform: Linux, Python 3.11
- GPU: single CUDA device
### Who can help?
Text-to-speech / Bark maintainers. (Feel free to reassign — the root cause is a `BarkModel.enable_cpu_o... | open | null | false | 2 | [] | [] | 2026-07-08T08:59:48Z | 2026-07-08T13:42:01Z | null | NONE | null | 20260708T180046Z | 2026-07-08T18:00:46Z | psyiode | 271,846,782 | U_kgDOEDQNfg | User | false |
huggingface/transformers | 4,838,751,224 | I_kwDOCUB6oc8AAAABIGl7-A | 47,176 | https://github.com/huggingface/transformers/issues/47176 | https://api.github.com/repos/huggingface/transformers/issues/47176 | Command allowlist bypass via $(...) command substitution under ALLOW_SHELL_OPERATORS=true | ### System Info
Hi, I've reported this on 8 June 2026 to huntr as guided in the security.md policy but have not gotten any response since: https://huntr.com/bounties/589e980e-cb15-4087-b1b3-d7d6ad2c0a15
can i share details here?
### Who can help?
_No response_
### Information
- [ ] The official example scripts
- ... | open | null | false | 0 | [
"bug"
] | [] | 2026-07-08T15:13:46Z | 2026-07-08T15:13:46Z | null | NONE | null | 20260708T180046Z | 2026-07-08T18:00:46Z | geo-chen | 2,404,584 | MDQ6VXNlcjI0MDQ1ODQ= | User | false |
huggingface/transformers | 4,838,767,325 | I_kwDOCUB6oc8AAAABIGm63Q | 47,177 | https://github.com/huggingface/transformers/issues/47177 | https://api.github.com/repos/huggingface/transformers/issues/47177 | Tar-slip / zip-slip arbitrary file write → RCE ... | ### System Info
Hi, I've reported this on 8 June 2026 to huntr as guided in the security.md policy but have not gotten any response since: https://huntr.com/bounties/ae9336f0-5fc7-4c87-a1d1-67ead0ec6f38
can i share details here?
### Who can help?
_No response_
### Information
- [ ] The official example scripts
-... | open | null | false | 3 | [
"bug"
] | [] | 2026-07-08T15:15:49Z | 2026-07-10T14:47:46Z | null | NONE | null | 20260710T180030Z | 2026-07-10T18:00:30Z | geo-chen | 2,404,584 | MDQ6VXNlcjI0MDQ1ODQ= | User | false |
huggingface/transformers | 4,842,126,535 | I_kwDOCUB6oc8AAAABIJz8xw | 47,182 | https://github.com/huggingface/transformers/issues/47182 | https://api.github.com/repos/huggingface/transformers/issues/47182 | [serge] integration failure triage - 2026-07-08 | <!-- serge-triage-run:integration-failure-triage:2026-07-08 -->
Automated **integration-failure triage** for the daily CI window `2026-07-02 → 2026-07-08`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-09T00:31:09Z | 2026-07-10T00:32:21Z | 2026-07-10T00:32:21Z | CONTRIBUTOR | null | 20260710T060026Z | 2026-07-10T06:00:26Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,845,253,967 | I_kwDOCUB6oc8AAAABIMy1Tw | 47,205 | https://github.com/huggingface/transformers/issues/47205 | https://api.github.com/repos/huggingface/transformers/issues/47205 | Standardize test image fixture loading for COCO images | ### Problem
Several tests still download COCO fixture images directly with `httpx.get(...)` or `requests.get(...)`. This bypasses `utils/fetch_hub_objects_for_ci.py` and the existing `url_to_local_path(...)` fallback behavior, so CI prefetching can succeed while local or partial test runs still hit `http://images.coco... | closed | completed | false | 3 | [] | [] | 2026-07-09T10:04:45Z | 2026-07-17T08:20:16Z | 2026-07-17T08:20:16Z | MEMBER | null | 20260717T120048Z | 2026-07-17T12:00:48Z | tarekziade | 250,019 | MDQ6VXNlcjI1MDAxOQ== | User | false |
huggingface/transformers | 4,848,085,101 | I_kwDOCUB6oc8AAAABIPfobQ | 47,217 | https://github.com/huggingface/transformers/issues/47217 | https://api.github.com/repos/huggingface/transformers/issues/47217 | VL models fail when their image placeholder is in user text. | ### System Info
- `transformers` version: 5.13.0
- Platform: macOS-26.3.1-arm64-arm-64bit-Mach-O
- Python version: 3.14.3
- Huggingface_hub version: 1.23.0
- Safetensors version: 0.8.0
- Accelerate version: not installed
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?)... | open | null | false | 9 | [
"bug"
] | [] | 2026-07-09T16:33:56Z | 2026-07-17T09:38:43Z | null | NONE | null | 20260717T120048Z | 2026-07-17T12:00:48Z | samuelstevens | 26,638,161 | MDQ6VXNlcjI2NjM4MTYx | User | false |
huggingface/transformers | 4,849,963,475 | I_kwDOCUB6oc8AAAABIRSR0w | 47,225 | https://github.com/huggingface/transformers/issues/47225 | https://api.github.com/repos/huggingface/transformers/issues/47225 | orthogonal_ weight init crashes with `geqrf_cuda not implemented for BFloat16` under v5 bf16 on-device init (RWKV and any QR-based init) | ### System Info
`transformers env` could not be run in the CI artifact context; closest captured environment:
- transformers 5.x (v5 on-device init path; `torch_dtype is deprecated` warning present)
- torch 2.14.0a0 (nightly)
- Python 3.12
- CUDA 13.x
- GPU: NVIDIA H100 (cc 9.0); also reproduced on B200 and GB200 — in... | closed | completed | false | 0 | [] | [] | 2026-07-09T21:17:18Z | 2026-07-17T14:33:52Z | 2026-07-17T14:33:52Z | CONTRIBUTOR | null | 20260717T180032Z | 2026-07-17T18:00:32Z | janbernloehr | 1,050,099 | MDQ6VXNlcjEwNTAwOTk= | User | false |
huggingface/transformers | 4,850,945,827 | I_kwDOCUB6oc8AAAABISOPIw | 47,227 | https://github.com/huggingface/transformers/issues/47227 | https://api.github.com/repos/huggingface/transformers/issues/47227 | [serge] integration failure triage - 2026-07-09 | <!-- serge-triage-run:integration-failure-triage:2026-07-09 -->
Automated **integration-failure triage** for the daily CI window `2026-07-03 → 2026-07-09`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-10T00:32:03Z | 2026-07-11T00:26:17Z | 2026-07-11T00:26:17Z | CONTRIBUTOR | null | 20260711T060022Z | 2026-07-11T06:00:22Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,851,062,522 | I_kwDOCUB6oc8AAAABISVW-g | 47,228 | https://github.com/huggingface/transformers/issues/47228 | https://api.github.com/repos/huggingface/transformers/issues/47228 | Sam3 bf16 compilation regression (v5.8.0 -> v5.9.0) | ### System Info
- transformers: 5.9.0 (also fails on main; works on 5.8.0)
- torch: 2.13.0+cu130
- Python: 3.10.12
- Platform: Ubuntu 22.04.3 LTS, kernel 6.8.0-134-generic, x86_64
- GPU: NVIDIA A100-PCIE-40GB
- CUDA: 13.2
### Who can help?
@yonigozlan @molbap
### Information
- [ ] The official example... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-07-10T00:56:26Z | 2026-07-14T13:28:56Z | 2026-07-14T13:28:56Z | NONE | null | 20260714T180038Z | 2026-07-14T18:00:38Z | ianxmason | 25,320,046 | MDQ6VXNlcjI1MzIwMDQ2 | User | false |
huggingface/transformers | 4,855,232,588 | I_kwDOCUB6oc8AAAABIWT4TA | 47,246 | https://github.com/huggingface/transformers/issues/47246 | https://api.github.com/repos/huggingface/transformers/issues/47246 | Incorrect inter-chunk recurrence in `NemotronHMamba2Mixer.torch_forward` (slow path). the Mamba2 fix from #35154 was never propagated to `nemotron_h` | ### System Info
- `transformers` versions affected: **5.6.0 through 5.13.0, and current `main`** (the `nemotron_h` model was refactored into the library after ~5.3.0).
- Platform: any (bug is in pure-PyTorch code; independent of OS/hardware).
- Only the **slow / no-kernel path** is affected (`use_mamba_kernels=False`,... | open | null | false | 2 | [
"bug"
] | [] | 2026-07-10T13:02:21Z | 2026-07-10T15:11:50Z | null | NONE | null | 20260710T180030Z | 2026-07-10T18:00:30Z | Jeronymous | 22,522,728 | MDQ6VXNlcjIyNTIyNzI4 | User | false |
huggingface/transformers | 4,858,314,849 | I_kwDOCUB6oc8AAAABIZQAYQ | 47,254 | https://github.com/huggingface/transformers/issues/47254 | https://api.github.com/repos/huggingface/transformers/issues/47254 | `CheckpointError` with PEFT + DeepSpeed ZeRO-3 + gradient checkpointing | ### System Info
- Platform: Linux-5.15.0-1048-aws-x86_64-with-glibc2.31
- Python version: 3.13.13
- PyTorch version: 2.11.0+cu128
- Transformers version: 5.13.0.dev0
- Accelerate version: 1.15.0.dev0
- DeepSpeed version: 0.19.2
- PEFT version: 0.19.1
- GPUs: 2x NVIDIA H100 80GB
### Who can help?
@SunMarc
### Inf... | open | null | false | 2 | [
"bug"
] | [] | 2026-07-10T20:09:31Z | 2026-07-10T21:29:51Z | null | MEMBER | null | 20260711T000031Z | 2026-07-11T00:00:31Z | qgallouedec | 45,557,362 | MDQ6VXNlcjQ1NTU3MzYy | User | false |
huggingface/transformers | 4,859,622,158 | I_kwDOCUB6oc8AAAABIafzDg | 47,256 | https://github.com/huggingface/transformers/issues/47256 | https://api.github.com/repos/huggingface/transformers/issues/47256 | [serge] integration failure triage - 2026-07-10 | <!-- serge-triage-run:integration-failure-triage:2026-07-10 -->
Automated **integration-failure triage** for the daily CI window `2026-07-04 → 2026-07-10`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-11T00:25:59Z | 2026-07-12T00:27:29Z | 2026-07-12T00:27:29Z | CONTRIBUTOR | null | 20260712T060026Z | 2026-07-12T06:00:26Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,864,752,726 | I_kwDOCUB6oc8AAAABIfY8Vg | 47,265 | https://github.com/huggingface/transformers/issues/47265 | https://api.github.com/repos/huggingface/transformers/issues/47265 | [serge] integration failure triage - 2026-07-11 | <!-- serge-triage-run:integration-failure-triage:2026-07-11 -->
Automated **integration-failure triage** for the daily CI window `2026-07-05 → 2026-07-11`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-12T00:27:13Z | 2026-07-13T00:28:33Z | 2026-07-13T00:28:33Z | CONTRIBUTOR | null | 20260713T060023Z | 2026-07-13T06:00:23Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 3,620,846,452 | I_kwDOCUB6oc7X0bt0 | 42,179 | https://github.com/huggingface/transformers/issues/42179 | https://api.github.com/repos/huggingface/transformers/issues/42179 | Add TileLang Kernel Support | ### Feature request
I would like to propose adding support for TileLang kernel in the transformers library. TileLang is a modular approach for writing attention kernels that could provide flexibility and performance benefits.
github link: https://github.com/tile-ai/tilelang
- Add TileLang as an optional attention back... | open | null | false | 1 | [
"Feature request"
] | [] | 2025-11-13T11:38:33Z | 2026-07-12T02:52:24Z | null | NONE | null | 20260712T060026Z | 2026-07-12T06:00:26Z | crownz248 | 58,587,382 | MDQ6VXNlcjU4NTg3Mzgy | User | false |
huggingface/transformers | 1,409,796,214 | I_kwDOCUB6oc5UB8h2 | 19,631 | https://github.com/huggingface/transformers/issues/19631 | https://api.github.com/repos/huggingface/transformers/issues/19631 | Add EDSR and MDSR | ### Model description
EDSR (Enhanced Deep Residual Networks for Single Image Super-Resolution) is for image super resolution, here's the [paper](https://arxiv.org/abs/1707.02921).
### Open source status
- [x] The model implementation is available
- [x] The model weights are available
### Provide useful lin... | closed | not_planned | false | 10 | [
"New model"
] | [] | 2022-10-14T20:02:21Z | 2026-07-12T06:55:34Z | 2026-07-12T06:55:34Z | CONTRIBUTOR | null | 20260712T120026Z | 2026-07-12T12:00:26Z | IMvision12 | 88,665,786 | MDQ6VXNlcjg4NjY1Nzg2 | User | false |
huggingface/transformers | 2,475,937,184 | I_kwDOCUB6oc6Tk82g | 32,902 | https://github.com/huggingface/transformers/issues/32902 | https://api.github.com/repos/huggingface/transformers/issues/32902 | StoppingCriteria for Repetition | ### Feature request
similar to repetition_penalty for generation config, but as a stopping criteria.
### Motivation
(small?) models tend to generated endless loops of the same few tokens, or a combination where they only increase like a single digit. (could not find any similar FRs)
I run into this quite a lot... | open | null | false | 4 | [
"Feature request"
] | [] | 2024-08-20T15:26:26Z | 2026-07-12T08:44:07Z | null | NONE | null | 20260712T120026Z | 2026-07-12T12:00:26Z | Vipitis | 23,384,483 | MDQ6VXNlcjIzMzg0NDgz | User | false |
huggingface/transformers | 4,866,167,181 | I_kwDOCUB6oc8AAAABIgvRjQ | 47,272 | https://github.com/huggingface/transformers/issues/47272 | https://api.github.com/repos/huggingface/transformers/issues/47272 | Avoid repeated Cohere ASR encoder projection during cached generation | **AI-assisted coordination disclosure:** This proposal was developed with AI assistance after three days of implementation, testing, benchmarking, and human review in our [Cohere Arabic batch-inference project](https://github.com/AliOsm/cohere-transcribe-arabic-batch-inference). I am opening this issue first to follow ... | open | null | false | 2 | [] | [] | 2026-07-12T09:16:01Z | 2026-07-18T02:49:35Z | null | CONTRIBUTOR | null | 20260718T060036Z | 2026-07-18T06:00:36Z | AliOsm | 7,662,492 | MDQ6VXNlcjc2NjI0OTI= | User | false |
huggingface/transformers | 4,866,712,285 | I_kwDOCUB6oc8AAAABIhQi3Q | 47,273 | https://github.com/huggingface/transformers/issues/47273 | https://api.github.com/repos/huggingface/transformers/issues/47273 | Cohere ASR chunk reassembly can crash on empty chunks and accept inconsistent metadata | **AI-generated implementation disclosure:** The proposed implementation was developed with AI assistance after three days of implementation, testing, benchmarking, and human review in our [production Cohere Arabic batch-inference project](https://github.com/AliOsm/cohere-transcribe-arabic-batch-inference). Any eventual... | open | null | false | 0 | [] | [] | 2026-07-12T10:43:10Z | 2026-07-12T10:43:10Z | null | CONTRIBUTOR | null | 20260712T120026Z | 2026-07-12T12:00:26Z | AliOsm | 7,662,492 | MDQ6VXNlcjc2NjI0OTI= | User | false |
huggingface/transformers | 4,866,733,597 | I_kwDOCUB6oc8AAAABIhR2HQ | 47,274 | https://github.com/huggingface/transformers/issues/47274 | https://api.github.com/repos/huggingface/transformers/issues/47274 | stable diffusion in tensorflow | I would like to share my idea from my perspective
Translation of all code in hugging face to tensorflow and Implementing Stable Diffusion in TensorFlow—primarily via the KerasCV Stable Diffusion API—provides industry-leading execution speeds, production-ready serving architectures, and seamless mobile deployment capa... | closed | completed | false | 3 | [] | [] | 2026-07-12T10:51:00Z | 2026-07-14T12:18:23Z | 2026-07-13T12:31:58Z | NONE | null | 20260714T180038Z | 2026-07-14T18:00:38Z | deborshi-web | 194,237,561 | U_kgDOC5PUeQ | User | false |
huggingface/transformers | 4,866,956,979 | I_kwDOCUB6oc8AAAABIhfesw | 47,276 | https://github.com/huggingface/transformers/issues/47276 | https://api.github.com/repos/huggingface/transformers/issues/47276 | Misleading synced_gpus warning message in continuous batching | ### System Info
- `transformers` version: 5.14.0.dev0
- Platform: Windows-11-10.0.26200-SP0
- Python version: 3.12.7
- Huggingface_hub version: 1.22.0
- Safetensors version: 0.8.0
- Accelerate version: 1.14.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerator?): 2.12.1+cp... | closed | completed | false | 0 | [
"bug"
] | [] | 2026-07-12T12:07:12Z | 2026-07-14T07:09:25Z | 2026-07-14T07:09:25Z | CONTRIBUTOR | null | 20260714T120113Z | 2026-07-14T12:01:13Z | Partha-Shankar | 152,790,727 | U_kgDOCRtmxw | User | false |
huggingface/transformers | 4,867,719,175 | I_kwDOCUB6oc8AAAABIiOABw | 47,277 | https://github.com/huggingface/transformers/issues/47277 | https://api.github.com/repos/huggingface/transformers/issues/47277 | Bug: `pthread_cond_clockwait` error in Termux/Android because of `esaxx-rs` | ### System Info
Symptoms:
```
Compiling console v0.16.1
warning: esaxx-rs@0.1.10: In file included from src/esaxx.cpp:621:
warning: esaxx-rs@0.1.10: In file included from src/esa.hxx:30:
warning: esaxx-rs@0.1.10: In file included from /data/data/com.termux/files/usr/bin/../../usr/include/c++/v1/vector:363:... | closed | completed | false | 1 | [
"bug"
] | [] | 2026-07-12T15:58:35Z | 2026-07-12T16:08:17Z | 2026-07-12T16:08:17Z | NONE | null | 20260712T180021Z | 2026-07-12T18:00:21Z | Manamama | 78,492,008 | MDQ6VXNlcjc4NDkyMDA4 | User | false |
huggingface/transformers | 4,869,400,743 | I_kwDOCUB6oc8AAAABIj0opw | 47,279 | https://github.com/huggingface/transformers/issues/47279 | https://api.github.com/repos/huggingface/transformers/issues/47279 | [serge] integration failure triage - 2026-07-12 | <!-- serge-triage-run:integration-failure-triage:2026-07-12 -->
Automated **integration-failure triage** for the daily CI window `2026-07-06 → 2026-07-12`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-13T00:28:17Z | 2026-07-14T00:24:51Z | 2026-07-14T00:24:51Z | CONTRIBUTOR | null | 20260714T060029Z | 2026-07-14T06:00:29Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,872,951,156 | I_kwDOCUB6oc8AAAABInNVdA | 47,299 | https://github.com/huggingface/transformers/issues/47299 | https://api.github.com/repos/huggingface/transformers/issues/47299 | [i18n-<supremeA AI Model ai.th AI >] Translating docs to <Copilot> | <!--
Note: Please search to see if an issue already exists for the language you are trying to translate.
-->
Hi!
Let's bring the documentation to all the <languageName>-speaking community 🌐 (currently 0 out of 267 complete)
Who would want to translate? Please follow the 🤗 [TRANSLATING guide](https://github.com/hug... | closed | completed | false | 2 | [
"WIP",
"Code agent slop"
] | [] | 2026-07-13T10:59:33Z | 2026-07-13T13:05:48Z | 2026-07-13T13:05:40Z | NONE | null | 20260713T180023Z | 2026-07-13T18:00:23Z | imgoingbacktheonlything239-ops | 269,549,381 | U_kgDOEBD_RQ | User | false |
huggingface/transformers | 4,872,270,242 | I_kwDOCUB6oc8AAAABImjxog | 47,293 | https://github.com/huggingface/transformers/issues/47293 | https://api.github.com/repos/huggingface/transformers/issues/47293 | pr_build_doc_with_comment.yml is rejected at startup (job 'if:' references a job not in its 'needs:') — 3,086 startup failures, no notification | Hi — I think `pr_build_doc_with_comment.yml` has been failing to start since early April, quietly.
In the `verity_pr_commit` job:
```yaml
if: ${{ needs.get-pr-number.outputs.PR_NUMBER != ''}}
runs-on: ubuntu-22.04
needs: get-pr-info
```
The `if:` refers to `get-pr-number`, but only `get-pr-info` is listed in `needs:... | open | null | false | 2 | [] | [] | 2026-07-13T09:26:59Z | 2026-07-13T11:18:53Z | null | NONE | null | 20260713T120047Z | 2026-07-13T12:00:47Z | kobihikri | 4,434,689 | MDQ6VXNlcjQ0MzQ2ODk= | User | false |
huggingface/transformers | 1,988,956,402 | I_kwDOCUB6oc52jRDy | 27,449 | https://github.com/huggingface/transformers/issues/27449 | https://api.github.com/repos/huggingface/transformers/issues/27449 | Compute logits and past_key_values once for the initial context rather than `num_beams` times? | ### Feature request
It looks like (at least) the following methods:
- sample
- beam search
- beam sample
- group beam search
all immediately expand their input batch size to either `num_beams` or `num_return_sequences`. This means that the initial context, which can be very long, will be run through the model `... | closed | completed | false | 6 | [] | [] | 2023-11-11T13:48:25Z | 2026-07-14T09:05:50Z | 2023-12-20T08:04:00Z | CONTRIBUTOR | null | 20260714T120113Z | 2026-07-14T12:01:13Z | tom-p-reichel | 43,631,024 | MDQ6VXNlcjQzNjMxMDI0 | User | false |
huggingface/transformers | 1,906,420,828 | I_kwDOCUB6oc5xoaxc | 26,316 | https://github.com/huggingface/transformers/issues/26316 | https://api.github.com/repos/huggingface/transformers/issues/26316 | Streaming and assisted decoding strategy for pipeline OR clearer understanding of pipelines to manually apply tasks | ### Feature request
Currently looking to get a number of non-pipeline functions with pipelines
- Decoder strategies
- Assisted decoding
- Streaming output
etc
**Alternately**, if I can adopt the non-pipeline transformer approach with tokenizer act more pipeline'ish, that'd also be good. Specifically, what are t... | closed | completed | false | 1 | [
"Feature request"
] | [] | 2023-09-21T08:34:25Z | 2026-07-14T11:47:54Z | 2026-07-14T11:47:53Z | NONE | null | 20260714T120113Z | 2026-07-14T12:01:13Z | gidzr | 83,053,994 | MDQ6VXNlcjgzMDUzOTk0 | User | false |
huggingface/transformers | 4,878,758,011 | I_kwDOCUB6oc8AAAABIsvwew | 47,309 | https://github.com/huggingface/transformers/issues/47309 | https://api.github.com/repos/huggingface/transformers/issues/47309 | [serge] integration failure triage - 2026-07-13 | <!-- serge-triage-run:integration-failure-triage:2026-07-13 -->
Automated **integration-failure triage** for the daily CI window `2026-07-07 → 2026-07-13`.
This issue was generated by AI-assisted automation. The grouping, summaries, and recommended follow-up can be incomplete or misleading; verify the failures before... | closed | not_planned | false | 1 | [] | [
"tarekziade"
] | 2026-07-14T00:24:35Z | 2026-07-17T00:27:28Z | 2026-07-17T00:27:28Z | CONTRIBUTOR | null | 20260717T060031Z | 2026-07-17T06:00:31Z | github-actions[bot] | 41,898,282 | MDM6Qm90NDE4OTgyODI= | Bot | false |
huggingface/transformers | 4,879,217,455 | I_kwDOCUB6oc8AAAABItLzLw | 47,312 | https://github.com/huggingface/transformers/issues/47312 | https://api.github.com/repos/huggingface/transformers/issues/47312 | Incompatible with kernels 0.16 | ### System Info
When I trying to load a model with `use_kernels=True`, the error happens:
```
Traceback (most recent call last):
File "/home/sdp/transformers/src/transformers/pipelines/base.py", line 256, in load_model
model = model_class.from_pretrained(model, **fp32_kwargs)
^^^^^^^^^^^^^^^^^^^^^^^... | open | null | false | 3 | [
"bug"
] | [] | 2026-07-14T02:12:59Z | 2026-07-19T13:23:17Z | null | CONTRIBUTOR | null | 20260719T180031Z | 2026-07-19T18:00:31Z | jiqing-feng | 107,918,818 | U_kgDOBm614g | User | false |
huggingface/transformers | 4,881,949,707 | I_kwDOCUB6oc8AAAABIvykCw | 47,317 | https://github.com/huggingface/transformers/issues/47317 | https://api.github.com/repos/huggingface/transformers/issues/47317 | PPFormulaNet training-loss bug | ### System Info
PPFormulaNet training-loss bug — ForCausalLMLoss incorrectly shifts labels in encoder-decoder model
The problem: PPFormulaNetForConditionalGeneration.forward() uses self.loss_function() (line 491 in modular_pp_formulanet.py) which maps to ForCausalLMLoss. ForCausalLMLoss internally shifts labels left b... | open | null | false | 0 | [
"bug"
] | [] | 2026-07-14T09:58:02Z | 2026-07-14T09:58:02Z | null | CONTRIBUTOR | null | 20260714T120113Z | 2026-07-14T12:01:13Z | sharmax-vikas | 175,734,208 | U_kgDOCnl9wA | User | false |
huggingface/transformers | 4,884,595,807 | I_kwDOCUB6oc8AAAABIyUEXw | 47,321 | https://github.com/huggingface/transformers/issues/47321 | https://api.github.com/repos/huggingface/transformers/issues/47321 | AMD quark class not updated | ### System Info
- `transformers` version: 5.13.1
- Platform: Linux-6.8.0-1058-gcp-x86_64-with-glibc2.39
- Python version: 3.13.13
- Huggingface_hub version: 1.23.0
- Safetensors version: 0.8.0
- Accelerate version: 1.14.0
- Accelerate config: not found
- DeepSpeed version: not installed
- PyTorch version (accelerat... | open | null | false | 1 | [
"bug"
] | [] | 2026-07-14T15:29:30Z | 2026-07-14T15:39:02Z | null | NONE | null | 20260714T180038Z | 2026-07-14T18:00:38Z | debasisdwivedy | 20,119,757 | MDQ6VXNlcjIwMTE5NzU3 | User | false |
huggingface/transformers | 4,885,165,405 | I_kwDOCUB6oc8AAAABIy21XQ | 47,325 | https://github.com/huggingface/transformers/issues/47325 | https://api.github.com/repos/huggingface/transformers/issues/47325 | `return_assistant_tokens_mask` masks the rest of the sequence when a `{% generation %}` span ends at token index 0 or on whitespace stripped by the pre-tokenizer | ### System Info
- `transformers` main (`fc7cf96`), `tokenizers` 0.22.2, Python 3.11 (macOS). The affected code path is unchanged on `main`, so this reproduces on recent releases as well.
### Reproduction
```python
from transformers import AutoTokenizer
tok = AutoTokenizer.from_pretrained("gpt2")
# The assistant tu... | closed | completed | false | 1 | [] | [] | 2026-07-14T16:42:44Z | 2026-07-17T06:08:19Z | 2026-07-17T06:08:19Z | NONE | null | 20260717T120048Z | 2026-07-17T12:00:48Z | sohumt123 | 208,212,517 | U_kgDODGkSJQ | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.