Skip to content

Releases: huggingface/transformers

Patch release v5.10.2

04 Jun 18:43
@vasqu vasqu
v5.10.2
0dad7b8

Choose a tag to compare

View all tags
Patch release v5.10.2 Latest
Latest

Patch release v5.10.2

There was a big bug in the model conversion of models related to clip, this affected models like sam3 and others. Please make sure to update 🙏

Full Changelog: v5.10.1...v5.10.2

Contributors

zucchini-nlp
Assets 2
Loading

Release v5.10.1

03 Jun 15:37
@ArthurZucker ArthurZucker
v5.10.1
90c3ae5

Choose a tag to compare

View all tags
Release v5.10.1

Release v5.10.1

v5.10.0 was yanked as we publish on a corrupted branch. Sorry everyone, this happens when we rush a release!!!

New Model additions

Gemma4 unified+ Gemma4 MTP

image

Gemma 4 12B Unified is an encoder-free multimodal model with pretrained and instruction-tuned variants. Unlike standard Gemma 4, which uses dedicated encoder towers, Gemma 4 12B Unified projects raw inputs directly into the language model's embedding space through lightweight linear pipelines. This results in a simpler architecture while maintaining strong multimodal performance.

Key differences from standard Gemma 4:

  • No Vision Tower: Raw pixel patches are projected directly into LM space via a Dense + LayerNorm pipeline with factorized 2D positional embeddings, replacing the vision encoder.
  • No Audio Tower: Raw 16 kHz waveform samples are chunked into fixed-length frames and projected through a simple RMSNorm → Linear pipeline, replacing the mel spectrogram + Conformer encoder.
  • Shared Multimodal Pipeline: Both vision and audio use the same Gemma4UnifiedMultimodalEmbedder (RMSNorm → Linear) for the final projection to text hidden space.

You can find the original Gemma 4 12B Unified checkpoints under the Gemma 4 release.

Sapiens2

Sapiens2 is a family of high-resolution vision transformers pretrained on ~1 billion curated human images, designed for human-centric computer vision tasks including pose estimation, body-part segmentation, surface normal estimation, and pointmap estimation. The models scale from 0.4B to 5B parameters and train at native 1K resolution, with hierarchical 4K variants for extended spatial reasoning. Sapiens2 achieves substantial improvements over its predecessor with +4 mAP in pose estimation, +24.3 mIoU in body-part segmentation, and 45.6% error reduction in normal estimation.

Links: Documentation | Paper

DeepSeek-OCR-2

DeepSeek-OCR-2 is an OCR-specialized vision-language model built on a distinctive architecture that combines a SAM ViT-B vision encoder with a Qwen2 hybrid attention encoder, connected through an MLP projector to a DeepSeek-V2 Mixture-of-Experts (MoE) language model. The model features a hybrid attention mechanism that applies bidirectional attention over image tokens and causal attention over query tokens, enabling efficient and accurate document understanding. It supports both plain OCR tasks and grounding capabilities with coordinate-aware output for document conversion to markdown format.

Links: Documentation

Mellum

Mellum is a code-focused Mixture-of-Experts language model developed by JetBrains. It is derived from the Qwen3-MoE architecture with per-layer-type RoPE and interleaved sliding window attention. The model has 12B total parameters with 2.5B active parameters per token, using 64 routed experts with 8 activated per token across 28 layers.

Links: Documentation

Breaking changes

The Gemma4 vision pooler now casts inputs to float32 before scaling to prevent float16 overflow (inf saturation) with large checkpoints, which may cause minor numerical differences in outputs for users running Gemma-4 vision models in float16.

Audio Language Models (ALMs) now have a dedicated base model class without a language modeling head, aligning them with the design of Vision Language Models (VLMs); users relying on the previous model class structure should update their code to use the new base model class where appropriate.

Parallelization

This release includes numerous bug fixes for model parallelism across multiple models (Gemma4, AltCLIP, ChineseClip, Blip-2, Whisper, Ovis2, Moshi) and parallel execution strategies, including fixes for tensor parallelism (TP), expert parallelism (EP), beam search under model parallel settings, and loss over-counting under TP/EP configurations. The continuous batching manager was also reworked for clearer control flow and improved TP race condition handling, and FSDP initialization via from_pretrained was introduced.

Cache

Fixed a regression in encoder-decoder cache initialization where the decoder config was incorrectly applied to the cross-attention cache, and resolved a RuntimeError caused by buffer size limits when warming up the cache on MPS devices. Additional test infrastructure improvements were made to support read-only cache environments used in CI.

Quantization

Added support for DeepGEMM BF16, mixed FP8/FP4, and MegaMoE quantization via a grouped linear refactor, while fixing two bugs: an FP8 MoE reverse substring issue affecting DSv4 initialization, and a BitsAndBytes 4-bit/8-bit quantization bug that silently dropped chunked tensors from one-to-many weight converters.

Bugfixes and improvements

  • Fix wrong changes produced by style/repo. check bot (#46371) by @ydshieh in [#46371]
  • Fix path traversal when saving Bark voice preset embeddings (#46237) by @LinZiyuu in [#46237]
  • Pass library_name/version to Hub calls via a shared HfApi (#46318) by @Wauplin in [#46318]
  • docs: update ACL Anthology URL in CITATION.cff (#46352) by @irfaan101 in [#46352]
  • [docs] contributing (#45465) by @stevhliu in [#45465]
  • [docs] Romanian translation of contributing.md, modular_transformers.md, multimodal_processing.md, add_vision_processing_components.md, add_audio_processing_components.md, modeling_rules.md, model_output_tracing.md, auto_docstring.md, testing.md, pr_checks.md and add_new_model.md . (#46345) by @filipinescu in [#46345]
  • [docs] xpu continuous batching (#46334) by @stevhliu in [#46334]
  • Fix incorrect attribute mapping relationships in GLM MoE DSA Config (#46338) by @Dovis01 in [#46338]
  • Fix grammar typos in Whisper documentation (#46336) by @calliec-1223 in [#46336]
  • [docs] update num_items_in_batch for causal LMs (#46335) by @stevhliu in [#46335]
  • Update compressed tensors minimum version (#46342) by @SunMarc in [#46342]
  • Fix _is_package_available reporting available without a version (#46125) by @blipbyte in [#46125]
  • remove sec (#46346) by @ydshieh in [#46346]
  • fix: include transitive relative imports when loading from local directory (#46022) by @trducng in [#46022]
  • perf(feature_extraction_sequence): skip re-splitting already-batched numpy arrays in pad() (#46329) by @Anai-Guo in [#46329]
  • [Zamba] Support attn_implementation dispatch (#46317) by @YangKai0616 in [#46317]
  • Fix TestAppRoutes test failures caused by deprecated asyncio.get_event_loop() on Python 3.10+ (#46340) by @ydshieh in [#46340]
  • [Qwen3VL] Fix video token placeholder: use self.video_token instead of hardcoded "<|placeholder|>" (#46296) by @kpal002 in [#46296]
  • chore(linter): fixes for rule 16 (#46023) by @tarekziade in [#46023]
  • [docs] Romanian translation of weightconverter.md, models.md, custom_models.md, monkey_patching.md, fusion_mapping.md, how_to_hack_models.md, model_sharing.md and serialization.md. (#46309) by @filipinescu in [#46309]
  • Normalize CUDA OOM errors when comparing commit failures in check_bad_commit (#46322) by @ydshieh in [#46322]
  • Fix unhandled exception noise from background safetensors conversion thread (#45752) by @dhruv7477 in [#45752]
  • Add Expectations for pipeline token classification tests (#46151) by @ka...
Read more

Contributors

shadeMe, tarekziade, and 53 other contributors
Loading

Release v5.9.0

20 May 14:12
@Cyrilvallez Cyrilvallez
v5.9.0
0a2757d
This commit was signed with the committer’s verified signature.
Cyrilvallez Cyril Vallez
SSH Key Fingerprint: OjK+mdCRyLrQ4vIh2+8FffCzuDZs1WByt+LD1Z6WNw4
Verified
Learn about vigilant mode.

Choose a tag to compare

View all tags
Release v5.9.0

Release v5.9.0

New Model additions

Cohere2Moe

Command A+ is a Mixture-of-Experts (MoE) language model from Cohere that features a hybrid attention pattern combining sliding window and full attention layers. The model incorporates both shared and routed experts and supports a very large context window for processing extensive text sequences.

Links: Documentation

Parakeet tdt (#44171)

HRM-Text

HRM-Text is an improved autoregressive language-modeling variant of the Hierarchical Reasoning Model (HRM) that uses a hierarchical recurrent forward pass with two transformer stacks - one for slow, abstract planning (H) and one for fast, detailed computation (L) - reused inside a nested recurrence. It features PrefixLM attention where instruction tokens attend bidirectionally while response tokens attend causally, per-head sigmoid output gates, and parameterless RMSNorm. The model is designed as a base language model without instruction tuning or chat templates.

Links: Documentation | Paper

Breaking changes

The text_embeds input for SAM3, EdgeTAM, and SAM3-Lite-Text models now expects full text embeddings instead of just pooler outputs, aligning with other models in the library — users must update their inputs accordingly.

  • 🚨Fix memory leaks caused by lru decorators in vision models (#45922) by @yonigozlan

Audio

Audio support was expanded with the addition of AudioFlamingoNext model checkpoints and improved compilability of audio/vision encoders via standalone pure functions. Additional improvements include better error messaging when loading audio from video files and new documentation for audio/video processors.

Generation

Fixed generation issues including inputs_embeds and per_layer_inputs handling for Gemma4, an AttributeError in RAG's generate() caused by missing config fields, and flaky VLM generation tests by blocking special image tokens during sampling.

Bugfixes and improvements

Significant community contributions

The following contributors have made significant changes to the library over the last release:

  • @lmaksym
  • @eustlb
    • user friendly error when loading audio from video (#45221)
    • [MultimodalLM] add language_model to the get/set_input_embeddings logic (#46029)
  • @remi-or
    • [CB] Remove OpenTelemetry (#45984)
    • [CB] [Major] Add tensor paralellism (#45821)
    • [CB] Hide activation footprint by using the CUDA graph pool (#45911)
  • @abcd1927

Contributors

tarekziade, igordertigor, and 25 other contributors
Loading

Patch release v5.8.1

13 May 03:21
@ArthurZucker ArthurZucker
v5.8.1
cc832f9

Choose a tag to compare

View all tags
Patch release v5.8.1

Patch release v5.8.1

This release is mainly to fix the Deepseek V4 integration!!!

image

Contributors

claude, silencelamb, and 4 other contributors
Loading

Release 5.8.0

05 May 16:52
@vasqu vasqu
v5.8.0
049d2bf

Choose a tag to compare

View all tags
Release 5.8.0

Release v5.8.0

New Model additions

DeepSeek-V4

image

DeepSeek-V4 is the next-generation MoE (Mixture of Experts) language model from DeepSeek that introduces several architectural innovations over DeepSeek-V3. The architecture replaces Multi-head Latent Attention (MLA) with a hybrid local + long-range attention design, swaps residual connections for Manifold-Constrained Hyper-Connections (mHC), and bootstraps the first few MoE layers with a static token-id → expert-id hash table. This implementation covers DeepSeek-V4-Flash, DeepSeek-V4-Pro, and their -Base pretrained variants, which share the same architecture but differ in width, depth, expert count and weights.

Links: Documentation | Paper

Gemma 4 Assistant

image

Gemma 4 Assistant is a small, text-only model that enables speculative decoding for Gemma 4 models using the Multi-Token Prediction (MTP) method and associated candidate generator. The model shares the same Gemma4TextModel backbone as other Gemma 4 models but uses KV sharing throughout the entire model, allowing it to reuse the KV cache populated by the target model and skip the pre-fill phase entirely. This architecture includes cross-attention to make the most of the target model's context, allowing the assistant to accurately predict more drafted tokens per drafting round.

Links: Documentation

GraniteSpeechPlus

image

Granite Speech Plus is a variant of Granite Speech that enhances the projector by consuming the concatenation of the encoder's final hidden states with an arbitrary subset of its intermediate hidden states along the feature dimension. It is a multimodal speech-to-text model that can transcribe audio, provide speaker annotation and word level timestamps by responding to text prompts. The model inherits the same architecture components as Granite Speech including the speech encoder, query transformer projector, language model, and optional LoRA adapter.

Links: Documentation

Granite4Vision

Granite Vision 4.1 is a vision-language model from IBM Research designed for enterprise-grade document data extraction. It specializes in chart extraction (Chart2CSV, Chart2Summary, Chart2Code), table extraction (JSON, HTML, OTSL), and semantic key-value pair extraction. The model builds on LLaVA-NeXT with architectural innovations including SigLIP2 Vision Encoder, Window Q-Former Projectors, and DeepStack Feature Injection with 8 vision-to-LLM injection points.

Links: Documentation

EXAONE-4.5

image

EXAONE 4.5 is the first open-weight vision language model developed by LG AI Research, integrating a dedicated visual encoder into the existing EXAONE 4.0 framework to expand multimodal capabilities. The model features 33 billion parameters in total, including 1.2 billion parameters from the vision encoder, and achieves competitive performance in general benchmarks while outperforming similar-sized models in document understanding and Korean contextual reasoning. It builds on EXAONE 4.0 with key enhancements including an expanded vocabulary of 153,600 tokens, support for up to 256K token context windows, and a Multi-Token Prediction (MTP) mechanism.

Links: Documentation | Paper | Blog Post

PP-FormulaNet

PP-FormulaNet-L and PP-FormulaNet_plus-L are lightweight models designed for table structure recognition, focusing on accurately recognizing table structures in documents and natural scenes. The models are part of the SLANet series and can be used for image-to-text tasks, specifically for detecting and processing mathematical formulas and table structures from images.

Links: Documentation

Breaking changes

Apex integration has been removed from the library (including RMSNorm usage in T5 and related models), so users relying on Apex for mixed precision or fused ops should migrate to PyTorch's native equivalents instead.

Tokenization

Fixed tokenizer mapping issues for DeepSeek R1 distilled (Qwen2) and DeepSeek OCR models, and resolved a significant performance regression in PreTrainedTokenizer.convert_ids_to_tokens where skip_special_tokens=True was rebuilding the special token set on every iteration, resulting in a ~300x speedup for that code path.

Bugfixes and improvements

Significant community contributions

The following contributors have made significant changes to the library over the last release:

Contributors

tarekziade, nurpax, and 29 other contributors
Loading

Release v5.7.0

28 Apr 18:32
@vasqu vasqu
v5.7.0
6ffbb07

Choose a tag to compare

View all tags
Release v5.7.0

Release v5.7.0

New Model additions

Laguna

image

Laguna is Poolside's mixture-of-experts language model family that extends standard SwiGLU MoE transformers with two key innovations. It features per-layer head counts allowing different decoder layers to have different query-head counts while sharing the same KV cache shape, and implements a sigmoid MoE router with auxiliary-loss-free load balancing that uses element-wise sigmoid of gate logits plus learned per-expert bias for router scoring.

Links: Documentation

DEIMv2

image

DEIMv2 (DETR with Improved Matching v2) is a real-time object detection model that extends DEIM with DINOv3 features and spans eight model sizes from X to Atto for diverse deployment scenarios. It uses a Spatial Tuning Adapter (STA) for larger variants to convert DINOv3's single-scale output into multi-scale features, while ultra-lightweight models employ pruned HGNetv2 backbones. The unified design achieves superior performance-cost trade-offs, with DEIMv2-X reaching 57.8 AP with only 50.3M parameters and DEIMv2-S being the first sub-10M model to exceed 50 AP on COCO.

Links: Documentation | Paper

Attention

Several attention-related bugs were fixed across multiple models, including a cross-attention cache type error in T5Gemma2 for long inputs, incorrect cached forward behavior in Qwen3.5's gated-delta-net linear attention, and a crash in GraniteMoeHybrid when no Mamba layers are present. Attention function dispatch was also updated to align with the latest model implementations.

Tokenizers

There was a bug in AutoTokenizer that caused the wrong tokenizer class to be initialized. This caused regressions in models like DeepSeek R1.

Generation

Continuous batching generation received several fixes and improvements, including correcting KV deduplication and memory estimation for long sequences (16K+), and removing misleading warnings about num_return_sequences and other unsupported features that were incorrectly firing even when functionality worked correctly. Documentation for per-request sampling parameters was also added.

Kernels

Improved kernel support by fixing configuration reading and error handling for FP8 checkpoints (e.g., Qwen3.5-35B-A3B-FP8), enabling custom expert kernels registered from the HF Hub to be properly loaded, and resolving an incompatibility that prevented Gemma3n and Gemma4 from using the rotary kernel.

Bugfixes and improvements

Significant community contributions

The following contributors have made significant changes to the library over the last release:

Contributors

kashif, tarekziade, and 33 other contributors
Loading

Patch release v5.6.2

23 Apr 18:36
@vasqu vasqu
v5.6.2
aa935fb

Choose a tag to compare

View all tags
Patch release v5.6.2

Patch release v5.6.2

Qwen 3.5 and 3.6 MoE (text-only) were broken when using with FP8. It should now work again with this 🫡

  • Fix configuration reading and error handling for kernels (#45610) by @hmellor

Full Changelog: v5.6.1...v5.6.2

Contributors

hmellor
Loading

Patch release v5.6.1

23 Apr 08:20
@ArthurZucker ArthurZucker
v5.6.1
45add8a

Choose a tag to compare

View all tags
Patch release v5.6.1

Patch release v5.6.1

Flash attention path was broken! Sorry everyone for this one 🤗

  • Fix AttributeError on s_aux=None in flash_attention_forward (#45589) by @jamesbraza

Contributors

jamesbraza
Loading

Release v5.6.0

22 Apr 15:52
@vasqu vasqu
v5.6.0
3e80155

Choose a tag to compare

View all tags
Release v5.6.0

Release v5.6.0

New Model additions

OpenAI Privacy Filter

OpenAI Privacy Filter is a bidirectional token-classification model for personally identifiable information (PII) detection and masking in text. It is intended for high-throughput data sanitization workflows where teams need a model that they can run on-premises that is fast, context-aware, and tunable. The model labels an input sequence in a single forward pass, then decodes coherent spans with a constrained Viterbi procedure, predicting probability distributions over 8 privacy-related output categories for each input token.

Links: Documentation

QianfanOCR

Qianfan-OCR is a 4B-parameter end-to-end document intelligence model developed by Baidu that performs direct image-to-text conversion without traditional multi-stage OCR pipelines. It supports a broad range of prompt-driven tasks including structured document parsing, table extraction, chart understanding, document question answering, and key information extraction all within one unified model. The model features a unique "Layout-as-Thought" capability that generates structured layout representations before producing final outputs, making it particularly effective for complex documents with mixed element types.

Links: Documentation | Paper

SAM3-LiteText

SAM3-LiteText is a lightweight variant of SAM3 that replaces the heavy SAM3 text encoder (353M parameters) with a compact MobileCLIP-based text encoder optimized through knowledge distillation, while keeping the SAM3 ViT-H image encoder intact. This reduces text encoder parameters by up to 88% while maintaining segmentation performance comparable to the original model. The model enables efficient vision-language segmentation by addressing the redundancy found in text prompting for segmentation tasks.

Links: Documentation | Paper

SLANet

SLANet and SLANet_plus are lightweight models designed for table structure recognition, focusing on accurately recognizing table structures in documents and natural scenes. The model improves accuracy and inference speed by adopting a CPU-friendly lightweight backbone network PP-LCNet, a high-low-level feature fusion module CSP-PAN, and a feature decoding module SLA Head that aligns structural and positional information. SLANet was developed by Baidu PaddlePaddle Vision Team as part of their table structure recognition solutions.

Links: Documentation

Breaking changes

The internal rotary_fn is no longer registered as a hidden kernel function, so any code referencing self.rotary_fn(...) within an Attention module will break and must be updated to call the function directly instead.

  • 🚨 [Kernels] Fix kernel function registration (#45420) by @vasqu

Serve

The transformers serve command received several enhancements, including a new /v1/completions endpoint for legacy text completion, multimodal support for audio and video inputs, improved tool-calling via parse_response, proper forwarding of tool_calls/tool_call_id fields, a 400 error on model mismatch when the server is pinned to a specific model, and fixes for the response API. Documentation was also updated to cover new serving options such as --compile and --model-timeout.

Vision

Several vision-related bug fixes were applied in this release, including correcting Qwen2.5-VL temporal RoPE scaling for still images, fixing missing/mismatched image processor backends for Emu3 and BLIP, resolving modular image processor class duplication, and preventing accelerate from incorrectly splitting vision encoders in PeVideo/PeAudioVideo models. Image loading performance was also improved by leveraging torchvision's native decode_image in the torchvision backend, yielding up to ~17% speedup over PIL-based loading.

Parallelization

Fixed several bugs affecting distributed training, including silently wrong results or NaN loss with Expert Parallelism, NaN weights on non-rank-0 FSDP processes, and a resize failure in PP-DocLayoutV3; additionally added support for loading adapters with Tensor Parallelism, added MoE to the Gemma4 TP plan, and published documentation for TP training.

Tokenization

Fixed a docstring typo in streamer classes, resolved a Kimi-K2.5 tokenizer regression and _patch_mistral_regex AttributeError, and patched a streaming generation crash for Qwen3VLProcessor caused by incorrect _tokenizer attribute access. Additional housekeeping included moving the GPT-SW3 instruct tokenizer to an internal testing repo and fixing a global state leak in the tokenizer registry during tests.

Cache

Cache handling was improved for Gemma4 and Gemma3n models by dissociating KV state sharing from the Cache class, ensuring KV states are always shared regardless of whether a Cache is used. Additionally, the image cache for Paddle models was updated to align with the latest API.

Audio

Audio models gained vLLM compatibility through targeted fixes across several model implementations, while reliability improvements were also made including exponential back-off retries for audio file downloads, a crash fix in the text-to-speech pipeline when generation configs contain None values, and corrected test failures for Kyutai Speech-To-Text.

Bugfixes and improvements

Read more

Contributors

earino, kashif, and 68 other contributors
Loading

Patch release v5.5.4

13 Apr 16:58
@ArthurZucker ArthurZucker
v5.5.4
75d3bdc

Choose a tag to compare

View all tags
Patch release v5.5.4

Patch release v5.5.4

This is mostly some fixes that are good to have asap, mostly for tokenizers;
** Fix Kimi-K2.5 tokenizer regression and _patch_mistral_regex Attribute… (#45305) by ArthurZucker

For training:
** Fix #45305 + add regression test GAS (#45349) by florian6973, SunMarc
** Fix IndexError with DeepSpeed ZeRO-3 when kernels rotary is active (#…) by ArthurZucker

And for Qwen2.5-VL :
** Fix Qwen2.5-VL temporal RoPE scaling applied to still images (#45330) by Kash6, zucchini-nlp

Loading