1 Pull request merged by 1 person

• 🌐 [i18n-KO] Translated qwen2_vl.md to Korean

  #36750 merged Mar 30, 2025

10 Pull requests opened by 9 people

• Add Fast Image Processor for MobileNetV1

  #37111 opened Mar 30, 2025

• Add Fast Image Processors for mobileViT

  #37112 opened Mar 30, 2025

• Add Fast Mobilenet-V2 Processor

  #37113 opened Mar 30, 2025

• RWKV: fix mask warning typo

  #37114 opened Mar 30, 2025

• chore: Update model doc for code_llama

  #37115 opened Mar 30, 2025

• Add FastImageProcessor for EfficientNet and update test suite

  #37119 opened Mar 31, 2025

• enable test_assisted_decoding_in_different_gpu UT on XPU

  #37120 opened Mar 31, 2025

• fix XPU UT error case brough by RNG difference btw XPU and CUDA

  #37121 opened Mar 31, 2025

• Correctly drop tokens in SwitchTransformer

  #37123 opened Mar 31, 2025

• Make canine model exportable by removing unncessary complicated logic

  #37124 opened Mar 31, 2025

5 Issues closed by 2 people

• rework `test_multi_gpu_data_parallel_forward`

  #31087 closed Mar 30, 2025

• model.gradient_checkpointing_enable() makes loss.requires_grad be False

  #35826 closed Mar 30, 2025

• ValueError: Unrecognized image processor in Qwen/Qwen2.5-VL-3B-Instruct.

  #36193 closed Mar 30, 2025

• past_key_value(s) name inconsistency causing problems

  #36290 closed Mar 30, 2025

• [Bugs] RuntimeError: No CUDA GPUs are available in transformers v4.48.0 or above when running Ray RLHF example

  #36295 closed Mar 30, 2025

5 Issues opened by 5 people

• Possible to move HybridCache from GPU to CPU?

  #37125 opened Mar 31, 2025

• Bug in Phi4 processor

  #37122 opened Mar 31, 2025

• FastAPI with LLM inference does not release accumulated VRAM

  #37118 opened Mar 31, 2025

• Paligemma connector details

  #37117 opened Mar 30, 2025

• modernBERT Duplicate Template Name

  #37116 opened Mar 30, 2025

27 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• 🌐 [i18n-KO] Translated `roberta.md` to Korean

  #37069 commented on Mar 30, 2025 • 9 new comments

• Add Qwen2.5-Omni

  #36752 commented on Mar 30, 2025 • 3 new comments

• Add Fast Image Processor for Donut

  #37081 commented on Mar 30, 2025 • 1 new comment

• Fix torchao usage

  #37034 commented on Mar 31, 2025 • 1 new comment

• Fix Gemma3 embedding scaling

  #37109 commented on Mar 31, 2025 • 0 new comments

• feat: updated model card for qwen_2.5_vl

  #37099 commented on Mar 31, 2025 • 0 new comments

• [WIP] Add Ovis2 model and processor implementation

  #37088 commented on Mar 30, 2025 • 0 new comments

• [draft] random tests order

  #37082 commented on Mar 30, 2025 • 0 new comments

• Fix 4090/ada not detected as having FP8 support

  #37067 commented on Mar 31, 2025 • 0 new comments

• Update model card for electra

  #37063 commented on Mar 30, 2025 • 0 new comments

• Add EfficientNet Image PreProcessor

  #37055 commented on Mar 30, 2025 • 0 new comments

• [tests] remove cuda-only test marker

  #37032 commented on Mar 31, 2025 • 0 new comments

• add gpt2 test on XPU

  #37028 commented on Mar 31, 2025 • 0 new comments

• fix and enhance pipeline_webserver.md

  #36992 commented on Mar 30, 2025 • 0 new comments

• Improve typing in TrainingArgument

  #36944 commented on Mar 30, 2025 • 0 new comments

• fix whisper re-compile

  #36712 commented on Mar 31, 2025 • 0 new comments

• prune LM Head for USD

  #36695 commented on Mar 30, 2025 • 0 new comments

• enable tp on CPU

  #36299 commented on Mar 31, 2025 • 0 new comments

• Improvements in attention_forward functions

  #36218 commented on Mar 30, 2025 • 0 new comments

• Add MLCD model

  #36182 commented on Mar 31, 2025 • 0 new comments

• [MLU] Fix FA2 check error, remove deepspeed-mlu deps.

  #36159 commented on Mar 31, 2025 • 0 new comments

• SwitchTransformer: Initialization of tensor to collect expert results is incorrect for dropped tokens (from ML POV)

  #37017 commented on Mar 31, 2025 • 0 new comments

• LLaVa_mistral models are unrecognized

  #37087 commented on Mar 30, 2025 • 0 new comments

• Add NeoBERT

  #37015 commented on Mar 30, 2025 • 0 new comments

• ViTPose tutorial fails

  #36454 commented on Mar 30, 2025 • 0 new comments

• [Contributions Welcome] Add Fast Image Processors

  #36978 commented on Mar 30, 2025 • 0 new comments

• Symbolic trance with past_key_values input is not supported yet for the qwen2.

  #36959 commented on Mar 30, 2025 • 0 new comments