Search results

Pass library_name="transformers" to HfApi/Hub functions in push paths so commits attribute correctly

Status: Open.

#45721 In huggingface/transformers;

· davanstrien opened on Apr 30, 2026

PreTrainedTokenizer.convert_ids_to_tokens(skip_special_tokens=True) rebuilds all_special_ids on every iteration of the per-id loop

Status: Open.

#45715 In huggingface/transformers;

· longlee0622 opened on Apr 30, 2026

Six leftover dummy classes in dummy_pt_objects.py fail check_repo.py and leak into dir(transformers) without torch

bug

Status: Open.

#45712 In huggingface/transformers;

· jw9603 opened on Apr 30, 2026

Fix wrong repo link in Dinov2ForImageClassification doc example

Status: Open.

#45710 In huggingface/transformers;

· Milan-Bhimani opened on Apr 30, 2026

T5 silently uses apex.FusedRMSNorm which has a memory leak (NVIDIA/apex#1999)

bug

Status: Open.

#45704 In huggingface/transformers;

· dustnehowl opened on Apr 29, 2026

transformers version changes the tokenization

bug

Status: Open.

#45701 In huggingface/transformers;

· PiRom1 opened on Apr 29, 2026

from_pretrained loads wrong custom module after save_pretrained

bug

Status: Open.

#45698 In huggingface/transformers;

· nurpax opened on Apr 29, 2026

Why the calculation of train_batch_size unrelated to split_batches

Status: Open.

#45693 In huggingface/transformers;

· mklpr opened on Apr 29, 2026

[moe] mps interface has error "histogram_mps" not implemented for 'Int'

bug

Status: Open.

#45685 In huggingface/transformers;

· chenzhe1204 opened on Apr 28, 2026

Gemma 4: Exploding pre-clip gradient norms during LoRA fine-tuning of gemma-4-31B-it

bug

Good Second IssueIssues that are more difficult to do than "Good First" issues - give it a try if you want!Issues that are more difficult to do than "Good First" issues - give it a try if you want!

Status: Open.

#45676 In huggingface/transformers;

· pritmish opened on Apr 28, 2026

[BitsAndBytesConfig] Providing llm_int8_skip_modules clears the default lm_head exclusion, causing AssertionError in 4-bit inference

Status: Open.

#45674 In huggingface/transformers;

· softguy777 opened on Apr 28, 2026

[Gemma4] torch.finfo() TypeError on uint8 weights in audio modules during 4-bit (NF4) inference

Status: Open.

#45672 In huggingface/transformers;

· softguy777 opened on Apr 28, 2026