Problemsby GlyphHugging Face users keep filing bugs
43 posts from 36 people25d activeFading, +2 this week
Posts per day
The posts behind it
`AuoModel` should map to base class
### System Info - ### Who can help? @eustlb ### Reproduction In the AutoModel mapping we have two models that map to a generaive (ConditionalGeneration) while it should be the base model Currently trying to test the base models causes errors in training-related tests which are...
Add support for compiled transformers
### Feature request I would like the transformers package to support using it in compiled version. It can not be used compiled since 4.51.3. ### Motivation My motivation is that in our project there's a requirement that the distributed Python environment shall be compiled. ###...
[Bug]: HuggingFaceInferenceAPIEmbedding without model_name fails: InferenceClient.get_recommended_model no longer exists
### Bug Description When `model_name` isn't given, `HuggingFaceInferenceAPIEmbedding.__init__` calls `InferenceClient.get_recommended_model(task=task)` (`llama_index/embeddings/huggingface/base.py`, line 449). That method no longer exists in huggingface_hub, so construction...
Advice on a set up for medical transcription and note formulation
Help needed: Voice cloning for a specific artist (Victoria Beckham)"
ZeroGPU Space stuck at APP_STARTING with “Scheduling failure: unable to schedule”
TimesFM: fp16 overflow in decoder `_forward_transform` (8 tests)
...discrepancy; the eager path itself crashes before any comparison is made. 8 failing CI tests in `timesfm`, all with `RuntimeError: value cannot be converted to type c10::Half without overflow`: -...
How can I move already downloaded models to HF cache without redownloading?
Aria, Inkling and PaddleOCR-VL report `input_modalities == "text"`, and declaring their modalities breaks their generation tests
### System Info transformers main (d88946453f) ### Who can help? @zucchini-nlp ### Reproduction For Aria the declaration sits on `AriaTextPreTrainedModel` instead of `AriaPreTrainedModel`, so it's inverted. Declaring the right modalities, as #49677 does for 8 other VLMs, turns...
[serve] Sampling not enabled for positive temperature
### System Info ### Who can help? @SunMarc ### Reproduction Fix with a test is in #49732 Run the serving script below, wait for `serve` to start, and then run the caller. Serving Script: Caller Script: The current output would be the same for all the 10 calls: ### Expected...
Space restart failing with 503 error (Request ID: Root=1-6ad114aa-77d5b0887e267e8418f585f4 )
Gemma4-12B Training Failed
...but Transformers does not recognize this architecture. This could be because of an issue with the checkpoint, or because your version of Transformers is out of date. my transformers version==5.19.0 ### Who can help? _No response_ ### Reproduction As described ### Expected...
CI failure: ClefIntegrationTest::test_clef_flash assertion error
...appropriate tag(s). ### Reproduction No standalone reproduction script is provided. The failing CI test is `tests/models/clef/test_modeling_clef.py::ClefIntegrationTest::test_clef_flash`. The failure is an AssertionError in `check_decision`: `0.0838903859257698 != 0.0817 within...
How would you diagnose a visual model that fails to learn recurring relationships across multiple timeframes?
Mamba2 / Zamba2: cached decoding does not respect the time-step limits used by sequence scan
### Reproduction Hi, I found an inconsistency between the sequence scan and cached single-token update in Mamba2 and Zamba2. In `Mamba2Mixer.forward()`, the scan path receives `dt_limit=self.time_step_limit` and applies: The cached decoding path passes the unprocessed `dt`,...
The text we compare with the first message changed the cluster match from 21% to 92%. How do we find which change mattered?
Packed sequences leak into each other in gated-DeltaNet models when only position_ids are given
### System Info transformers main (4cc2aa8430), torch 2.13.0+cu130, flash-linear-attention 0.5.2, causal-conv1d 1.7.0, B300. ### Who can help? @vasqu ### Reproduction ### Expected behavior The second document scores the same packed or alone (~0.04 in bf16). Attention layers...
Trainer disables find_unused_parameters under any gradient checkpointing, not just reentrant
### System Info transformers main (4cc2aa8430), torch 2.13.0+cu130, 2x B300. ### Who can help? @SunMarc ### Reproduction When `ddp_find_unused_parameters` is unset, `Trainer` uses `find_unused = not gradient_checkpointing`...
LIES
ArtCraft FREE is a LIE. You can't do anything as long as you don't BUY CREDITS
get_uniform_frame_indices can return too many or out-of-range frame indices
### System Info transformers: 5.19.0.dev0 (main @ cb33194) numpy: 2.5.3 Python: 3.12.3 OS: Linux ### Who can help? @zucchini-nlp ### Information - [ ] The official example scripts - [x] My own modified scripts ### Tasks - [ ] An officially supported task in the `examples` folder...
Model Request: Index Translate family
Dear Team, Thank you for your work. Are there plans to support Index-Translate models: [https://huggingface.co/collections/IndexTeam/index-translate](https://huggingface.co/collections/IndexTeam/index-translate) They seems to be quite popular on HF as translation models Thank...
[(Alaska Pet FAQ) How Do I Add a Pet to My Alaska Flight]
1-844-499-9533 How Do I Activate AccountEdge?
Fix : 1-844-499-9533 How Do I Download AccountEdge?
Fix 1-844-499-9533 How Do I Install AccountEdge?
[(Delta Pet FAQ) How Can I Add a Pet to a Delta Flight]""""""{}
[(Delta Pet FAQ) How Do I Add a Pet to My Delta Flight]}{}{}
How Should We Govern AI Agents in One Platform?
Why Trust Wallet Not Working📞 𝟷 (𝟿𝟸𝟿) 𝟻𝟿O-𝟿OO𝟼
Trainer counts fp16 steps skipped by GradScaler as completed update steps (`global_step`, `max_steps`, `on_optimizer_step`)
### System Info transformers 5.17.0 and main @ 96331a9; accelerate 1.15.0 and main @ 421af8b; torch 2.14.0; Python 3.13; Tesla T4, CUDA 13.0, `fp16=True`. Also reproduces on CPU with a `GradScaler` installed on the Accelerator (on CPU, `fp16=True` creates no scaler). ### Who can...
ExponentialDecayLengthPenalty creates NaN EOS scores when combined with minimum-length constraints
### System Info - Transformers: `5.19.0.dev0`, main commit `d6c1e71bd717bf092f8293f0c3c9bd4a5ac5401a`; also reproduced on stable `5.18.0`. - Python: `3.12.14` - PyTorch: `2.14.1+cpu` - Accelerate: `1.15.0` - Platform: Linux x86-64, CPU, single process. - Tiny models constructed...
Next Release?
Hello - Quick question. Any idea when the next release will be? I'm trying to plan my own release and I'd really like to wait until https://github.com/huggingface/transformers/pull/48505 is deployed. If it's not known, that's totally fine but figured I'd ask.
SAM3 Video reconditioning causes mask to grow onto the edges and make the bounding boxes balloon
### System Info - `transformers` version: 5.8.1 - Platform: Linux-6.14.0-1017-oem-x86_64-with-glibc2.39 - Python version: 3.11.15 - Huggingface_hub version: 1.15.0 - Safetensors version: 0.7.0 - Accelerate version: not installed - Accelerate config: not found - DeepSpeed...
How should we test whether structured agent debate improves evidence review?
Unable to re-submit the authorship claim
Issues with multiple spaces stuck with "No @spaces.GPU function detected during startup"
Nothing destroys the process group, so every distributed Trainer run warns at exit
### System Info transformers 5.18.0.dev0, accelerate 1.15.0, torch 2.13.0+cu129. ### Who can help? @SunMarc ### Reproduction Any distributed `Trainer` script that exits after training prints, once per rank: The Trainer sets distributed training up through accelerate and nothing...
Can't join discord and the course links are broken
How do i make and add my api key connected to something
Kernels for partial RoPE
### Feature request Would be cool to add kernel for partial rotations like below ### Motivation @vasqu 😄 (https://github.com/huggingface/transformers/pull/48594#discussion_r3959426411) ### Your contribution tell me if you need anything
Loading lugh/ClinicalT5-base
I load my model the following way:. from transformers import AutoTokenizer, AutoModelForSeq2SeqLM model_id = "luqh/ClinicalT5-base" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForSeq2SeqLM.from_pretrained(model_id, from_flax=True) However I get the...
[Bug] `HfArgumentParser` silently corrupts `tuple` and `set` typed dataclass fields from CLI
...with `tuple` or `set` types (e.g. `tuple[int, int]` or `set[str]`), `HfArgumentParser` fails to parse them correctly, resulting in silent data corruption or unparsed argument errors: ### Root Cause Analysis In `src/transformers/hf_argparser.py` (lines 220-234),...
Nested per-layer `rope_parameters` is misparsed when `layer_types` omits one of its keys
### System Info - `transformers`: `main` @ `83d46aa2a2` (reported version `5.16.0.dev0`). The relevant code in `modeling_rope_utils.py` is identical at the `v5.16.1` tag, so the released version is affected too. - `torch`: 2.8.0+cu128 - Python: 3.12.9 - Platform:...
Companies and products named
About this problem
Evidence
43 posts from 36 people in 5 places, about 4 a week over 70 days. Mostly on Hugging Face Forums, huggingface/transformers, Stack Overflow [huggingface-transformers]. Tools named alongside: GitHub, NVIDIA, PyTorch, Tesla.
Frustration Frustration 0 of 3· Seen on GitHub issues, Community forums, Stack Overflow
Summary, model-written
Hugging Face users experience failed GPU pool startups, broken Discord and course links, undetected GPU space functions, API key setup confusion, and authorship claim re-submission errors.
How it was grouped
Posts about Hugging Face that state a pain but match no known issue yet. First post Aug 28, 2026, latest Nov 5, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
Other Hugging Face problems
Hugging Face upgrades break working setups
14 posts from 13 people on GitHub issues, Reddit, Community forums.
Running Hugging Face locally hits memory and speed limits
17 posts from 15 people on GitHub issues, Reddit.
Fine-tuning Hugging Face is costly and unpredictable
9 posts from 9 people on GitHub issues, Reddit.
Trouble with Hugging Face fail
Top post: "All paid GPU pools fail to start: MOUNT_ATTR_IDMAP error at container init (Exit code 128)", one of 8 posts from 5 people on GitHub issues, Community forums.
Trouble with Hugging Face generate
Top post: "prefill_chunk_size drops image inputs, VLMs generate without the image", one of 6 posts from 5 people on GitHub issues, Community forums.
Trouble with Hugging Face token
Top post: "<|image|> token has special=False in Rust tokenizer backend, should it be True?", one of 4 posts from 4 people on GitHub issues.
Trouble with Hugging Face logits
Top post: "`DogeCDMoE` router `.view(2, bsz * seq_len, -1)` mixes tokens: with `is_moe=True`, logits depend on future", one of 4 posts from 4 people on GitHub issues.
Trouble with Hugging Face docs
Top post: "Translating docs to <languageName>", one of 4 posts from 3 people on GitHub issues.
Trouble with Hugging Face torch
Top post: "import transformers spends 1-3 s in an eager importlib.metadata.packages_distributions() when torch >= 2.14", one of 4 posts from 4 people on GitHub issues.
Trouble with Hugging Face load
Top post: "gemma 4 can not load audio file and produce Audio features and audio tokens do not match, tokens: 31,", one of 3 posts from 3 people on GitHub issues.
Trouble with Hugging Face chat
Top post: "Chat parsing: `json` parser's `string_delims` doesn't handle escaped delimiters", one of 3 posts from 3 people on GitHub issues, Community forums.
Hugging Face runs out of context on real work
3 posts from 3 people on GitHub issues.
History