v4.33.2 is a patch release. Key changes include:

• Fix pad to multiple of;
• fix _resize_token_embeddings will set lm head size to 0 when enabled deepspeed zero3.

v4.33.1 is a patch release. Key changes include:

• Falcon port #24523 by @Rocketknight1;
• Falcon: Add RoPE scaling by @gante in #25878.

v4.32.1 is a patch release. Key changes include:

• Put IDEFICS in the right section of the doc;
• removing unnecesssary extra parameter.

v4.32.0 is a minor release. Key changes include:

• new model: IDEFICS via HuggingFaceM4 by @stas00 in #24796;
• [MPT] Add MosaicML's MPT model to transformers by @ArthurZucker & @younesbelkada in #24629.

v4.31.0 is a minor release. Key changes include:

• Add support for Llama 2 by @ArthurZucker in #24891;
• Add Musicgen by @sanchit-gandhi in #24109.

v4.30.2 is a patch release. Key changes include:

• Fix push to hubby @NielsRogge in #24187;
• Fix how we detect the TF package by @Rocketknight1 in #24255.

v4.30.1 is a patch release. Key changes include:

• Fix bnb config json serialization in #24137 by @younesbelkada;
• Correctly build models and import call_context for older TF versions in #24138 by @Rocketknight1.

v4.30.0 is a minor release. Key changes include:

• Top 100 by @LysandreJik in #22912;
• Add LlamaIndex to awesome-transformers.md by @ravi03071991 in #23484.

v4.29.2 is a patch release: fixes the package so non-python files (like cuda kernels) are properly included.

v4.29.1 is a patch release. Key changes include:

• Revert "search buffers for dtype" in #23308 by @sgugger;
• Fix image segmentation tool test in #23306 by @sgugger.

v4.29.0 is a minor release. Key changes include:

• Transformers Agents by @LysandreJik @patrickvonplaten and @sgugger in #23214;
• Add Segment Anything Model (SAM) by @ArthurZucker in #22654.

v4.28.1 is a patch release: revert the change on deta by @ydshieh in #22750

v4.28.0 is a minor release. Key changes include:

• LLaMA Implementation by @zphang in #21955;
• Add Pix2Struct by @younesbelkada in #21400.

v4.27.4 is a patch release: revert "error (also in original) model, scaling only q matrix not qk.t dot product (qk.t/sqrt(dim_per_head)) (#21627) in #22444 by @sgugger

v4.27.3 is a patch release: enforce max_memory for device_map strategies by @sgugger in #22311

v4.27.2 is a patch release: fix balanced and auto device_map by @sgugger in #22271

v4.27.1 is a patch release. Key changes include:

• Revert 22152 MaskedImageCompletionOutput changes;
• Regression pipeline device.

v4.27.0 is a minor release. Key changes include:

• Add BridgeTower model by @abhiwand in #20775;
• Add loss for BridgeTowerForMaskedLM and BridgeTowerForImageAndTextRetrieval by @abhiwand in #21684.

v4.26.1 is a patch release. Key changes include:

• ESM openfold_utils type hints by @ringohoffman in #20544;
• Add cPython files in build by @sgugger in #21372.

v4.26.0 is a minor release. Key changes include:

• Generate: use GenerationConfig as the basis for .generate() parametrization by @gante in #20388;
• Generate: TF uses GenerationConfig as the basis for .generate() parametrization by @gante in #20994.

v4.33.2 is a patch release. Key changes include:

• Fix pad to multiple of;
• fix _resize_token_embeddings will set lm head size to 0 when enabled deepspeed zero3.

v4.33.1 is a patch release. Key changes include:

• Falcon port #24523 by @Rocketknight1;
• Falcon: Add RoPE scaling by @gante in #25878.

v4.32.1 is a patch release. Key changes include:

• Put IDEFICS in the right section of the doc;
• removing unnecesssary extra parameter.

v4.32.0 is a minor release. Key changes include:

• new model: IDEFICS via HuggingFaceM4 by @stas00 in #24796;
• [MPT] Add MosaicML's MPT model to transformers by @ArthurZucker & @younesbelkada in #24629.

v4.31.0 is a minor release. Key changes include:

• Add support for Llama 2 by @ArthurZucker in #24891;
• Add Musicgen by @sanchit-gandhi in #24109.

v4.30.2 is a patch release. Key changes include:

• Fix push to hubby @NielsRogge in #24187;
• Fix how we detect the TF package by @Rocketknight1 in #24255.

v4.30.1 is a patch release. Key changes include:

• Fix bnb config json serialization in #24137 by @younesbelkada;
• Correctly build models and import call_context for older TF versions in #24138 by @Rocketknight1.

v4.30.0 is a minor release. Key changes include:

• Top 100 by @LysandreJik in #22912;
• Add LlamaIndex to awesome-transformers.md by @ravi03071991 in #23484.

v4.29.2 is a patch release: fixes the package so non-python files (like cuda kernels) are properly included.

v4.29.1 is a patch release. Key changes include:

• Revert "search buffers for dtype" in #23308 by @sgugger;
• Fix image segmentation tool test in #23306 by @sgugger.

v4.29.0 is a minor release. Key changes include:

• Transformers Agents by @LysandreJik @patrickvonplaten and @sgugger in #23214;
• Add Segment Anything Model (SAM) by @ArthurZucker in #22654.

v4.28.1 is a patch release: revert the change on deta by @ydshieh in #22750

v4.28.0 is a minor release. Key changes include:

• LLaMA Implementation by @zphang in #21955;
• Add Pix2Struct by @younesbelkada in #21400.

v4.27.4 is a patch release: revert "error (also in original) model, scaling only q matrix not qk.t dot product (qk.t/sqrt(dim_per_head)) (#21627) in #22444 by @sgugger

v4.27.3 is a patch release: enforce max_memory for device_map strategies by @sgugger in #22311

v4.27.2 is a patch release: fix balanced and auto device_map by @sgugger in #22271

v4.27.1 is a patch release. Key changes include:

• Revert 22152 MaskedImageCompletionOutput changes;
• Regression pipeline device.

v4.27.0 is a minor release. Key changes include:

• Add BridgeTower model by @abhiwand in #20775;
• Add loss for BridgeTowerForMaskedLM and BridgeTowerForImageAndTextRetrieval by @abhiwand in #21684.

v4.26.1 is a patch release. Key changes include:

• ESM openfold_utils type hints by @ringohoffman in #20544;
• Add cPython files in build by @sgugger in #21372.

v4.26.0 is a minor release. Key changes include:

• Generate: use GenerationConfig as the basis for .generate() parametrization by @gante in #20388;
• Generate: TF uses GenerationConfig as the basis for .generate() parametrization by @gante in #20994.