State-of-the-art Machine Learning for PyTorch, TensorFlow, and JAX.
π€ Transformers provides APIs and tools to easily download and train state-of-the-art pretrained models. Using pretrained models can reduce your compute costs, carbon footprint, and save you the time and resources required to train a model from scratch. These models support common tasks in different modalities, such as:
π Natural Language Processing: text classification, named entity recognition, question answering, language modeling, summarization, translation, multiple choice, and text generation.
πΌοΈ Computer Vision: image classification, object detection, and segmentation.
π£οΈ Audio: automatic speech recognition and audio classification.
π Multimodal: table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
π€ Transformers support framework interoperability between PyTorch, TensorFlow, and JAX. This provides the flexibility to use a different framework at each stage of a modelβs life; train a model in three lines of code in one framework, and load it for inference in another. Models can also be exported to a format like ONNX and TorchScript for deployment in production environments.
Join the growing community on the Hub, forum, or Discord today!
The documentation is organized into five sections:
GET STARTED provides a quick tour of the library and installation instructions to get up and running.
TUTORIALS are a great place to start if youβre a beginner. This section will help you gain the basic skills you need to start using the library.
HOW-TO GUIDES show you how to achieve a specific goal, like finetuning a pretrained model for language modeling or how to write and share a custom model.
CONCEPTUAL GUIDES offers more discussion and explanation of the underlying concepts and ideas behind models, tasks, and the design philosophy of π€ Transformers.
API describes all classes and functions:
The table below represents the current support in the library for each of those models, whether they have a Python tokenizer (called βslowβ). A βfastβ tokenizer backed by the π€ Tokenizers library, whether they have support in Jax (via Flax), PyTorch, and/or TensorFlow.
Model | PyTorch support | TensorFlow support | Flax Support |
---|---|---|---|
ALBERT | β | β | β |
ALIGN | β | β | β |
AltCLIP | β | β | β |
Audio Spectrogram Transformer | β | β | β |
Autoformer | β | β | β |
Bark | β | β | β |
BART | β | β | β |
BARThez | β | β | β |
BARTpho | β | β | β |
BEiT | β | β | β |
BERT | β | β | β |
Bert Generation | β | β | β |
BertJapanese | β | β | β |
BERTweet | β | β | β |
BigBird | β | β | β |
BigBird-Pegasus | β | β | β |
BioGpt | β | β | β |
BiT | β | β | β |
Blenderbot | β | β | β |
BlenderbotSmall | β | β | β |
BLIP | β | β | β |
BLIP-2 | β | β | β |
BLOOM | β | β | β |
BORT | β | β | β |
BridgeTower | β | β | β |
BROS | β | β | β |
ByT5 | β | β | β |
CamemBERT | β | β | β |
CANINE | β | β | β |
Chinese-CLIP | β | β | β |
CLAP | β | β | β |
CLIP | β | β | β |
CLIPSeg | β | β | β |
CLVP | β | β | β |
CodeGen | β | β | β |
CodeLlama | β | β | β |
Cohere | β | β | β |
Conditional DETR | β | β | β |
ConvBERT | β | β | β |
ConvNeXT | β | β | β |
ConvNeXTV2 | β | β | β |
CPM | β | β | β |
CPM-Ant | β | β | β |
CTRL | β | β | β |
CvT | β | β | β |
Data2VecAudio | β | β | β |
Data2VecText | β | β | β |
Data2VecVision | β | β | β |
DBRX | β | β | β |
DeBERTa | β | β | β |
DeBERTa-v2 | β | β | β |
Decision Transformer | β | β | β |
Deformable DETR | β | β | β |
DeiT | β | β | β |
DePlot | β | β | β |
Depth Anything | β | β | β |
DETA | β | β | β |
DETR | β | β | β |
DialoGPT | β | β | β |
DiNAT | β | β | β |
DINOv2 | β | β | β |
DistilBERT | β | β | β |
DiT | β | β | β |
DonutSwin | β | β | β |
DPR | β | β | β |
DPT | β | β | β |
EfficientFormer | β | β | β |
EfficientNet | β | β | β |
ELECTRA | β | β | β |
EnCodec | β | β | β |
Encoder decoder | β | β | β |
ERNIE | β | β | β |
ErnieM | β | β | β |
ESM | β | β | β |
FairSeq Machine-Translation | β | β | β |
Falcon | β | β | β |
FastSpeech2Conformer | β | β | β |
FLAN-T5 | β | β | β |
FLAN-UL2 | β | β | β |
FlauBERT | β | β | β |
FLAVA | β | β | β |
FNet | β | β | β |
FocalNet | β | β | β |
Funnel Transformer | β | β | β |
Fuyu | β | β | β |
Gemma | β | β | β |
GIT | β | β | β |
GLPN | β | β | β |
GPT Neo | β | β | β |
GPT NeoX | β | β | β |
GPT NeoX Japanese | β | β | β |
GPT-J | β | β | β |
GPT-Sw3 | β | β | β |
GPTBigCode | β | β | β |
GPTSAN-japanese | β | β | β |
Graphormer | β | β | β |
Grounding DINO | β | β | β |
GroupViT | β | β | β |
HerBERT | β | β | β |
Hubert | β | β | β |
I-BERT | β | β | β |
IDEFICS | β | β | β |
Idefics2 | β | β | β |
ImageGPT | β | β | β |
Informer | β | β | β |
InstructBLIP | β | β | β |
Jamba | β | β | β |
Jukebox | β | β | β |
KOSMOS-2 | β | β | β |
LayoutLM | β | β | β |
LayoutLMv2 | β | β | β |
LayoutLMv3 | β | β | β |
LayoutXLM | β | β | β |
LED | β | β | β |
LeViT | β | β | β |
LiLT | β | β | β |
LLaMA | β | β | β |
Llama2 | β | β | β |
Llama3 | β | β | β |
LLaVa | β | β | β |
LLaVA-NeXT | β | β | β |
Longformer | β | β | β |
LongT5 | β | β | β |
LUKE | β | β | β |
LXMERT | β | β | β |
M-CTC-T | β | β | β |
M2M100 | β | β | β |
MADLAD-400 | β | β | β |
Mamba | β | β | β |
Marian | β | β | β |
MarkupLM | β | β | β |
Mask2Former | β | β | β |
MaskFormer | β | β | β |
MatCha | β | β | β |
mBART | β | β | β |
mBART-50 | β | β | β |
MEGA | β | β | β |
Megatron-BERT | β | β | β |
Megatron-GPT2 | β | β | β |
MGP-STR | β | β | β |
Mistral | β | β | β |
Mixtral | β | β | β |
mLUKE | β | β | β |
MMS | β | β | β |
MobileBERT | β | β | β |
MobileNetV1 | β | β | β |
MobileNetV2 | β | β | β |
MobileViT | β | β | β |
MobileViTV2 | β | β | β |
MPNet | β | β | β |
MPT | β | β | β |
MRA | β | β | β |
MT5 | β | β | β |
MusicGen | β | β | β |
MusicGen Melody | β | β | β |
MVP | β | β | β |
NAT | β | β | β |
Nezha | β | β | β |
NLLB | β | β | β |
NLLB-MOE | β | β | β |
Nougat | β | β | β |
NystrΓΆmformer | β | β | β |
OLMo | β | β | β |
OneFormer | β | β | β |
OpenAI GPT | β | β | β |
OpenAI GPT-2 | β | β | β |
OpenLlama | β | β | β |
OPT | β | β | β |
OWL-ViT | β | β | β |
OWLv2 | β | β | β |
PatchTSMixer | β | β | β |
PatchTST | β | β | β |
Pegasus | β | β | β |
PEGASUS-X | β | β | β |
Perceiver | β | β | β |
Persimmon | β | β | β |
Phi | β | β | β |
Phi3 | β | β | β |
PhoBERT | β | β | β |
Pix2Struct | β | β | β |
PLBart | β | β | β |
PoolFormer | β | β | β |
Pop2Piano | β | β | β |
ProphetNet | β | β | β |
PVT | β | β | β |
PVTv2 | β | β | β |
QDQBert | β | β | β |
Qwen2 | β | β | β |
Qwen2MoE | β | β | β |
RAG | β | β | β |
REALM | β | β | β |
RecurrentGemma | β | β | β |
Reformer | β | β | β |
RegNet | β | β | β |
RemBERT | β | β | β |
ResNet | β | β | β |
RetriBERT | β | β | β |
RoBERTa | β | β | β |
RoBERTa-PreLayerNorm | β | β | β |
RoCBert | β | β | β |
RoFormer | β | β | β |
RWKV | β | β | β |
SAM | β | β | β |
SeamlessM4T | β | β | β |
SeamlessM4Tv2 | β | β | β |
SegFormer | β | β | β |
SegGPT | β | β | β |
SEW | β | β | β |
SEW-D | β | β | β |
SigLIP | β | β | β |
Speech Encoder decoder | β | β | β |
Speech2Text | β | β | β |
SpeechT5 | β | β | β |
Splinter | β | β | β |
SqueezeBERT | β | β | β |
StableLm | β | β | β |
Starcoder2 | β | β | β |
SuperPoint | β | β | β |
SwiftFormer | β | β | β |
Swin Transformer | β | β | β |
Swin Transformer V2 | β | β | β |
Swin2SR | β | β | β |
SwitchTransformers | β | β | β |
T5 | β | β | β |
T5v1.1 | β | β | β |
Table Transformer | β | β | β |
TAPAS | β | β | β |
TAPEX | β | β | β |
Time Series Transformer | β | β | β |
TimeSformer | β | β | β |
Trajectory Transformer | β | β | β |
Transformer-XL | β | β | β |
TrOCR | β | β | β |
TVLT | β | β | β |
TVP | β | β | β |
UDOP | β | β | β |
UL2 | β | β | β |
UMT5 | β | β | β |
UniSpeech | β | β | β |
UniSpeechSat | β | β | β |
UnivNet | β | β | β |
UPerNet | β | β | β |
VAN | β | β | β |
VideoMAE | β | β | β |
ViLT | β | β | β |
VipLlava | β | β | β |
Vision Encoder decoder | β | β | β |
VisionTextDualEncoder | β | β | β |
VisualBERT | β | β | β |
ViT | β | β | β |
ViT Hybrid | β | β | β |
VitDet | β | β | β |
ViTMAE | β | β | β |
ViTMatte | β | β | β |
ViTMSN | β | β | β |
VITS | β | β | β |
ViViT | β | β | β |
Wav2Vec2 | β | β | β |
Wav2Vec2-BERT | β | β | β |
Wav2Vec2-Conformer | β | β | β |
Wav2Vec2Phoneme | β | β | β |
WavLM | β | β | β |
Whisper | β | β | β |
X-CLIP | β | β | β |
X-MOD | β | β | β |
XGLM | β | β | β |
XLM | β | β | β |
XLM-ProphetNet | β | β | β |
XLM-RoBERTa | β | β | β |
XLM-RoBERTa-XL | β | β | β |
XLM-V | β | β | β |
XLNet | β | β | β |
XLS-R | β | β | β |
XLSR-Wav2Vec2 | β | β | β |
YOLOS | β | β | β |
YOSO | β | β | β |