Hugging Face Transformers: v4.43.1: Patch release
Hugging Face Transformers v4.43.1 is published, but the available notes are too limited to assess upgrade impact. Review the official source before upgrading.
transformers release archive — Page 3
256 tracked entries. Older history is imported progressively; see source coverage.
Hugging Face Transformers: v4.43.0: Llama 3.1, Chameleon, ZoeDepth, Hiera
Hugging Face Transformers v4.43.0. Source highlight: The Llama 3.1 models are released by Meta and come in three flavours: 8B, 70B, and 405B.
Hugging Face Transformers: Patch release v4.42.4
Hugging Face Transformers v4.42.4. Source highlight: Mostly gemma2 support FA2 softcapping!
Hugging Face Transformers: Patch release v4.42.3
Hugging Face Transformers v4.42.3. Source highlight: Make sure we have attention softcapping for "eager" GEMMA2 model
Hugging Face Transformers: Patch release v4.42.2
Hugging Face Transformers v4.42.2. Source highlight: Thanks to our 2 contributors for their prompt fixing mostly applies for training and FA2!
Hugging Face Transformers: v4.42.1: Patch release
Hugging Face Transformers v4.42.1 is published, but the available notes are too limited to assess upgrade impact. Review the official source before upgrading.
Hugging Face Transformers: v4.42.0: Gemma 2, RTDETR, InstructBLIP, LLAVa Next, New Model Adder
Hugging Face Transformers v4.42.0. Source highlight: The Gemma2 model was proposed in Gemma2: Open Models Based on Gemini Technology and Research by Gemma2 Team, Google.
Hugging Face Transformers: Release v4.41.2
Hugging Face Transformers v4.41.2. Source highlight: Mostly fixing some stuff related to `trust_remote_code=True` and `from_pretrained`
Hugging Face Transformers: Release v4.41.1 Fix PaliGemma finetuning, and some small bugs
Hugging Face Transformers v4.41.1. Source highlight: Fix PaliGemma finetuning:
Hugging Face Transformers: v4.41.0: Phi3, JetMoE, PaliGemma, VideoLlava, Falcon2, FalconVLM & GGUF support
Hugging Face Transformers v4.41.0. Source highlight: The Phi-3 model was proposed in Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone by Microsoft.
Hugging Face Transformers v4.40.2 — Release Notes & Engineering Impact
Hugging Face Transformers v4.40.2. Source highlight: Fix torch fx for LLama model
Hugging Face Transformers: v4.40.1: fix `EosTokenCriteria` for `Llama3` on `mps`
Hugging Face Transformers v4.40.1. Source highlight: Kudos to @pcuenca for the prompt fix in:
Hugging Face Transformers: v4.40.0: Llama 3, Idefics 2, Recurrent Gemma, Jamba, DBRX, OLMo, Qwen2MoE, Grounding Dino
Hugging Face Transformers v4.40.0. Source highlight: Llama 3 is supported in this release through the Llama 2 architecture and some fixes in the `tokenizers` library.
Hugging Face Transformers: Release v4.39.3
Hugging Face Transformers v4.39.3. Source highlight: The `AWQ` issue persisted, and there was a regression reported with beam search and input embeddings.
Hugging Face Transformers: Patch release v4.39.2
Hugging Face Transformers v4.39.2. Source highlight: Series of fixes for backwards compatibility (AutoAWQ and other quantization libraries, imports from `trainer_pt_utils`) and functionality (LLaMA tokenizer conversion)
Hugging Face Transformers: Patch release v4.39.1
Hugging Face Transformers v4.39.1. Source highlight: Patch release to fix some breaking changes to LLaVA model, fixes/cleanup for Cohere & Gemma and broken doctest
Hugging Face Transformers: Release v4.39.0
Hugging Face Transformers v4.39.0. Source highlight: 🚨 VRAM consumption 🚨
Hugging Face Transformers v4.38.2 — Release Notes & Engineering Impact
Hugging Face Transformers v4.38.2. Source highlight: Fix backward compatibility issues with Llama and Gemma:
Hugging Face Transformers v4.38.1 — Release Notes & Engineering Impact
Hugging Face Transformers v4.38.1. Source highlight: Fix eager attention in Gemma!
Hugging Face Transformers: v4.38: Gemma, Depth Anything, Stable LM; Static Cache, HF Quantizer, AQLM
Hugging Face Transformers v4.38.0. Source highlight: Gemma is a new opensource Language Model series from Google AI that comes with a 2B and 7B variant. The release comes with the pre-trained and instruction fine-tuned versions and you can use them via `AutoModelForCausalLM`, `GemmaForCausalLM` or `pipeline` interface!
Hugging Face Transformers: Patch release v4.37.2
Hugging Face Transformers v4.37.2. Source highlight: Protecting the imports for SigLIP's tokenizer if sentencepiece isn't installed
Hugging Face Transformers: Patch release: v4.37.1
Hugging Face Transformers v4.37.1. Source highlight: A patch release to resolve import errors from removed custom types in generation utils
Hugging Face Transformers: v4.37 Qwen2, Phi-2, SigLIP, ViP-LLaVA, Fast2SpeechConformer, 4-bit serialization, Whisper longform generation
Hugging Face Transformers v4.37.0. Source highlight: Qwen2 is the new model series of large language models from the Qwen team. Previously, the Qwen series was released, including Qwen-72B, Qwen-1.8B, Qwen-VL, Qwen-Audio, etc.
Hugging Face Transformers: Patch release: v4.36.2
Hugging Face Transformers v4.36.2. Source highlight: Patch release to resolve some critical issues relating to the recent cache refactor, flash attention refactor and training in the multi-gpu and multi-node settings:
Hugging Face Transformers: Patch release: v4.36.1
Hugging Face Transformers v4.36.1. Source highlight: A patch release for critical torch issues mostly:
Hugging Face Transformers: v4.36: Mixtral, Llava/BakLlava, SeamlessM4T v2, AMD ROCm, F.sdpa wide-spread support
Hugging Face Transformers v4.36.0. Source highlight: Mixtral is the new open-source model from Mistral AI announced by the blogpost Mixtral of Experts. The model has been proven to have comparable capabilities to Chat-GPT according to the benchmark results shared on the release blogpost.
Hugging Face Transformers: Patch release: v4.35.2
Hugging Face Transformers v4.35.2. Source highlight: A patch release was made for the following commit:
Hugging Face Transformers: Patch release: v4.35.1
Hugging Face Transformers v4.35.1. Source highlight: A patch release was made for the following three commits:
Hugging Face Transformers: Safetensors serialization by default, DistilWhisper, Fuyu, Kosmos-2, SeamlessM4T, Owl-v2
Hugging Face Transformers v4.35.0. Source highlight: Distil-Whisper is a distilled version of Whisper that is 6 times faster, 49% smaller, and performs within 1% word error rate (WER) on out-of-distribution data. It was proposed in the paper Robust Knowledge Distillation via Large-Scale Pseudo Labelling.
Hugging Face Transformers: Patch release: v4.34.1
Hugging Face Transformers v4.34.1. Source highlight: A patch release was made for the following three commits:
Hugging Face Transformers: v4.34: Mistral, Persimmon, Prompt templating, Flash Attention 2, Tokenizer refactor
Hugging Face Transformers v4.34.0. Source highlight: Mistral-7B-v0.1 is a decoder-based LM with the following architectural choices:
Hugging Face Transformers: Patch release: v4.33.3
Hugging Face Transformers v4.33.3. Source highlight: A patch release was made for the following three commits:
Hugging Face Transformers: Patch release: v4.33.2
Hugging Face Transformers v4.33.2. Source highlight: A patch release was done for these two commits:
Hugging Face Transformers: Falcon, Code Llama, ViTDet, DINO v2, VITS
Hugging Face Transformers v4.33.1. Source highlight: Falcon is a class of causal decoder-only models built by TII. The largest Falcon checkpoints have been trained on >=1T tokens of text, with a particular emphasis on the RefinedWeb corpus. They are made available under the Apache 2.0 license.
Hugging Face Transformers: Patch release: v4.32.1
Hugging Face Transformers v4.32.1. Source highlight: Patch release including several patches from v4.31.0, listed below:
Hugging Face Transformers: IDEFICS, GPTQ Quantization
Hugging Face Transformers v4.32.0. Source highlight: The IDEFICS model was proposed in OBELICS: An Open Web-Scale Filtered Dataset of Interleaved Image-Text Documents by Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, Victor Sa
Hugging Face Transformers: v4.31.0: Llama v2, MusicGen, Bark, MMS, EnCodec, InstructBLIP, Umt5, MRa, vIvIt
Hugging Face Transformers v4.31.0. Source highlight: Llama 2 was proposed in LLaMA: Open Foundation and Fine-Tuned Chat Models by Hugo Touvron et al. It builds upon the Llama architecture adding Grouped Query Attention for efficient inference.
Hugging Face Transformers: v4.30.2: Patch release
Hugging Face Transformers v4.30.2. Source highlight: Fix push to hubby @NielsRogge in #24187
Hugging Face Transformers: v4.30.1 Patch release
Hugging Face Transformers v4.30.1. Source highlight: Fix bnb config json serialization in #24137
Hugging Face Transformers: v4.30.0: 100k, Agents improvements, Safetensors core dependency, Swiftformer, Autoformer, MobileViTv2, timm-as-a-backbone
Hugging Face Transformers v4.30.0. Source highlight: Transformers has just reached 100k stars on GitHub, and to celebrate we wanted to highlight 100 projects in the vicinity of `transformers` and we have decided to create an awesome-transformers page to do just that.
Hugging Face Transformers: v4.29.2: Patch release
Hugging Face Transformers v4.29.2 is published, but the available notes are too limited to assess upgrade impact. Review the official source before upgrading.
Hugging Face Transformers: V4.29.1: Patch release
Hugging Face Transformers v4.29.1. Source highlight: Reverts a regression in the FSDP integration.
Hugging Face Transformers: v4.29.0: Transformers Agents, SAM, RWKV, FocalNet, OpenLLaMa
Hugging Face Transformers v4.29.0. Source highlight: Transformers Agent is a new API that lets you use the library and Diffusers by prompting an agent (which is a large language model) in natural language. That agent will then output code using a set of predefined tools, leveraging the appropriate (and state-of-the-art) models for the task the user wa
Hugging Face Transformers: v4.28.1: Patch release
Hugging Face Transformers v4.28.1 is published, but the available notes are too limited to assess upgrade impact. Review the official source before upgrading.
Hugging Face Transformers: v4.28.0: LLaMa, Pix2Struct, MatCha, DePlot, MEGA, NLLB-MoE, GPTBigCode
Hugging Face Transformers v4.28.0. Source highlight: The LLaMA model was proposed in LLaMA: Open and Efficient Foundation Language Models. It is a collection of foundation language models ranging from 7B to 65B parameters. You can request access to the weights here then use the conversion script to generate a checkpoint compatible with Hugging Face
Hugging Face Transformers: v4.27.4: Patch release
Hugging Face Transformers v4.27.4. Source highlight: This patch fixes a regression with FlauBERT and XLM models.
Hugging Face Transformers: v4.27.3: Patch release
Hugging Face Transformers v4.27.3 is published, but the available notes are too limited to assess upgrade impact. Review the official source before upgrading.
Hugging Face Transformers: v4.27.2: Patch release
Hugging Face Transformers v4.27.2 is published, but the available notes are too limited to assess upgrade impact. Review the official source before upgrading.
Hugging Face Transformers: v4.27.1: Patch release
Hugging Face Transformers v4.27.1. Source highlight: Patches two unwanted breaking changes that were released in v4.27.0:
Hugging Face Transformers: BridgeTower, Whisper speedup, DETA, SpeechT5, BLIP-2, CLAP, ALIGN, API updates
Hugging Face Transformers v4.27.0. Source highlight: The goal of this model is to build a bridge between each uni-modal encoder and the cross-modal encoder to enable comprehensive and detailed interaction at each layer of the cross-modal encoder thus achieving remarkable performance on various downstream tasks with almost negligible additional perform