Search results

RoFormer attention implementation does not use attention interface

Status: Open.

#46144 In huggingface/transformers;

· ir2718 opened on May 21, 2026

**kwargs not passed through methods of RoFormer models

Status: Open.

#46143 In huggingface/transformers;

· ir2718 opened on May 21, 2026

Discussion: optional RankSEG-style decoding for Transformers semantic segmentation post-processing

Status: Open.

#46139 In huggingface/transformers;

· Leev1s opened on May 21, 2026

Add TIPSv2 (Advancing Vision-Language Pretraining with Enhanced Patch-Text Alignment) by Google DeepMind

New model

Status: Open.

#46133 In huggingface/transformers;

· farrosalferro opened on May 21, 2026

[deepseek_v4] conversion_mapping doesn't cover mtp.* paths — MTP keys silently random-init even after _keys_to_ignore is empty

Status: Open.

#46129 In huggingface/transformers;

· pasta-paul opened on May 20, 2026

MaskGenerationPipeline: is_last never True on final partial batch, silently dropping results

bug

Status: Open.

#46123 In huggingface/transformers;

· J3r3myPerera opened on May 20, 2026

convert_rope_params_to_dict raises TypeError when ignore_keys_at_rope_validation is a JSON-loaded list

bug

Status: Open.

#46121 In huggingface/transformers;

· Charly21r opened on May 20, 2026

Path Traversal in Sharded Checkpoint Loader via Unsanitized weight_map Entries in *.index.json

bug

Status: Open.

#46097 In huggingface/transformers;

· karnakarreddi opened on May 20, 2026

[deepseekv4]Does Transformers provide a weight conversion script to convert the Hugging Face weights into a format that can be read by Transformers from_pretrained?

Feature requestRequest for a new featureRequest for a new feature

Status: Open.

#46095 In huggingface/transformers;

· young-creator opened on May 20, 2026

[Inquiry] What is the current status and roadmap for supporting packed sequences (packing) on Qwen3.5/hybrid linear attention models?

Status: Open.

#46093 In huggingface/transformers;

· liangxuZhang opened on May 20, 2026

LlamaConfig rejects explicit head_dim when hidden_size is not divisible by num_attention_heads

Status: Open.

#46082 In huggingface/transformers;

· Anri-Lombard opened on May 19, 2026

Gemma4 use_bidirectional_attention="all" still builds causal attention masks

bug

Status: Open.

#46077 In huggingface/transformers;

· oliverholworthy opened on May 19, 2026