Need help with pytorch-image-models?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

13.8K Stars 2.2K Forks Apache License 2.0 1.1K Commits 51 Opened issues


PyTorch image models, scripts, pretrained weights -- ResNet, ResNeXT, EfficientNet, EfficientNetV2, NFNet, Vision Transformer, MixNet, MobileNet-V3/V2, RegNet, DPN, CSPNet, and more

Services available


Need anything else?

Contributors list

PyTorch Image Models


A big thank you to my GitHub Sponsors for their support!

In addition to the sponsors at the link above, I've received hardware and/or cloud resources from * Nvidia ( * TFRC (

I'm fortunate to be able to dedicate significant time and money of my own supporting this and other open source projects. However, as the projects increase in scope, outside support is needed to continue with the current trajectory of hardware, infrastructure, and electricty costs.

What's New

Oct 3, 2021

  • ResNet strikes back ( weights added, plus any extra training components used.
  • Attention model experiments are in as well (across, along with weights. Details forthcoming.
  • A lot more to add here...

Aug 18, 2021

  • Optimizer bonanza!
    • Add LAMB and LARS optimizers, incl trust ratio clipping options. Tweaked to work properly in PyTorch XLA (tested on TPUs w/
      timm bits
    • Add MADGRAD from FB research w/ a few tweaks (decoupled decay option, step handling that works with PyTorch XLA)
    • Some cleanup on all optimizers and factory. No more
      , a bit more consistency, unit tests for all!
    • SGDP and AdamP still won't work with PyTorch XLA but others should (have yet to test Adabelief, Adafactor, Adahessian myself).
  • EfficientNet-V2 XL TF ported weights added, but they don't validate well in PyTorch (L is better). The pre-processing for the V2 TF training is a bit diff and the fine-tuned 21k -> 1k weights are very sensitive and less robust than the 1k weights.
  • Added PyTorch trained EfficientNet-V2 'Tiny' w/ GlobalContext attn weights. Only .1-.2 top-1 better than the SE so more of a curiosity for those interested.

July 12, 2021

July 5-9, 2021

  • Add
    weights, a custom 'tiny' 13.6M param variant that is a bit better than (non NoisyStudent) B3 models. Both faster and better accuracy (at same or lower res)
    • top-1 82.34 @ 288x288 and 82.54 @ 320x320
  • Add SAM pretrained in1k weight for ViT B/16 (
    ) and B/32 (
    ) models.
  • Add 'Aggregating Nested Transformer' (NesT) w/ weights converted from official Flax impl. Contributed by Alexander Soare.
    • jx_nest_base
      - 83.534,
      - 83.120,
      - 81.426

June 23, 2021

  • Reproduce gMLP model training,
    trained to 79.6 top-1, matching paper. Hparams for this and other recent MLP training here

June 20, 2021

  • Release Vision Transformer 'AugReg' weights from How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers
    • .npz weight loading support added, can load any of the 50K+ weights from the AugReg series
    • See example notebook from official impl for navigating the augreg weights
    • Replaced all default weights w/ best AugReg variant (if possible). All AugReg 21k classifiers work.
    • Highlights:
      (87.1 top-1),
      (86.2 top-1),
      (86.0 top-1)
    • vit_deit_*
      renamed to just
    • Remove my old small model, replace with DeiT compatible small w/ AugReg weights
  • Add 1st training of my
    MLP /w GLU, 78.1 top-1 w/ 25M params.
  • Add weights from official ResMLP release (
  • Add
    weights from my 'lightweight' series. 84.7 top-1 at 384x384.
  • Add distilled BiT 50x1 student and 152x2 Teacher weights from Knowledge distillation: A good teacher is patient and consistent
  • NFNets and ResNetV2-BiT models work w/ Pytorch XLA now
    • weight standardization uses F.batchnorm instead of stdmean (std_mean wasn't lowered)
    • eps values adjusted, will be slight differences but should be quite close
  • Improve test coverage and classifier interface of non-conv (vision transformer and mlp) models
  • Cleanup a few classifier / flatten details for models w/ conv classifiers or early global pool
  • Please report any regressions, this PR touched quite a few models.

June 8, 2021

  • Add first ResMLP weights, trained in PyTorch XLA on TPU-VM w/ my XLA branch. 24 block variant, 79.2 top-1.
  • Add ResNet51-Q model w/ pretrained weights at 82.36 top-1.
    • NFNet inspired block layout with quad layer stem and no maxpool
    • Same param count (35.7M) and throughput as ResNetRS-50 but +1.5 top-1 @ 224x224 and +2.5 top-1 at 288x288

May 25, 2021

  • Add LeViT, Visformer, ConViT (PR by Aman Arora), Twins (PR by paper authors) transformer models
  • Add ResMLP and gMLP MLP vision models to the existing MLP Mixer impl
  • Fix a number of torchscript issues with various vision transformer models
  • Cleanup inputsize/imgsize override handling and improve testing / test coverage for all vision transformer and MLP models
  • More flexible pos embedding resize (non-square) for ViT and TnT. Thanks Alexander Soare
  • Add
    model and weights (started training before official code). 84.8 top-1, 53M params.

May 14, 2021

  • Add EfficientNet-V2 official model defs w/ ported weights from official Tensorflow/Keras impl.
    • 1k trained variants:
    • 21k trained variants:
    • 21k pretrained -> 1k fine-tuned:
    • v2 models w/ v1 scaling:
    • Rename my prev V2 guess
    • Some blank
      models in-place for future native PyTorch training

May 5, 2021

  • Add MLP-Mixer models and port pretrained weights from Google JAX impl
  • Add CaiT models and pretrained weights from FB
  • Add ResNet-RS models and weights from TF. Thanks Aman Arora
  • Add CoaT models and weights. Thanks Mohammed Rizin
  • Add new ImageNet-21k weights & finetuned weights for TResNet, MobileNet-V3, ViT models. Thanks mrT
  • Add GhostNet models and weights. Thanks Kai Han
  • Update ByoaNet attention modules
    • Improve SA module inits
    • Hack together experimental stand-alone Swin based attn module and
    • Consistent '26t' model defs for experiments.
  • Add improved Efficientnet-V2S (prelim model def) weights. 83.8 top-1.
  • WandB logging support

April 13, 2021

  • Add Swin Transformer models and weights from

April 12, 2021

  • Add ECA-NFNet-L1 (slimmed down F1 w/ SiLU, 41M params) trained with this code. 84% top-1 @ 320x320. Trained at 256x256.
  • Add EfficientNet-V2S model (unverified model definition) weights. 83.3 top-1 @ 288x288. Only trained single res 224. Working on progressive training.
  • Add ByoaNet model definition (Bring-your-own-attention) w/ SelfAttention block and corresponding SA/SA-like modules and model defs
    • Lambda Networks -
    • Bottleneck Transformers -
    • Halo Nets -
  • Adabelief optimizer contributed by Juntang Zhuang

April 1, 2021

  • Add snazzy
    script for bulk
    model benchmarking of train and/or inference
  • Add Pooling-based Vision Transformer (PiT) models (from
    • Merged distilled variant into main for torchscript compatibility
    • Some
      cleanup/style tweaks and weights have hub download support
  • Cleanup Vision Transformer (ViT) models
    • Merge distilled (DeiT) model into main so that torchscript can work
    • Support updated weight init (defaults to old still) that closer matches original JAX impl (possibly better training from scratch)
    • Separate hybrid model defs into different file and add several new model defs to fiddle with, support patch_size != 1 for hybrids
    • Fix fine-tuning numclass changes (PiT and ViT) and posembed resizing (Vit) with distilled variants
    • nn.Sequential for block stack (does not break downstream compat)
  • TnT (Transformer-in-Transformer) models contributed by author (from
  • Add RegNetY-160 weights from DeiT teacher model
  • Add new NFNet-L0 w/ SE attn (rename
    ) weights 82.75 top-1 @ 288x288
  • Some fixes/improvements for TFDS dataset wrapper

March 17, 2021

  • Add new ECA-NFNet-L0 (rename
    ) weights trained by myself.
    • 82.6 top-1 @ 288x288, 82.8 @ 320x320, trained at 224x224
    • Uses SiLU activation, approx 2x faster than
      and 50% faster than
      w/ 1/3 param count
  • Integrate Hugging Face model hub into timm createmodel and defaultcfg handling for pretrained weight and config sharing (more on this soon!)
  • Merge HardCoRe NAS models contributed by
  • Merge PyTorch trained EfficientNet-EL and pruned ES/EL variants contributed by DeGirum

March 7, 2021

  • First 0.4.x PyPi release w/ NFNets (& related), ByoB (GPU-Efficient, RepVGG, etc).
  • Change feature extraction for pre-activation nets (NFNets, ResNetV2) to return features before activation.
  • Tested with PyTorch 1.8 release. Updated CI to use 1.8.
  • Benchmarked several arch on RTX 3090, Titan RTX, and V100 across 1.7.1, 1.8, NGC 20.12, and 21.02. Some interesting performance variations to take note of

Feb 18, 2021

  • Add pretrained weights and model variants for NFNet-F* models from DeepMind Haiku impl.
    • Models are prefixed with
      . They require SAME padding conv, skipinit enabled, and activation gains applied in act fn.
    • These models are big, expect to run out of GPU memory. With the GELU activiation + other options, they are roughly 1/2 the inference speed of my SiLU PyTorch optimized
    • Original model results are based on pre-processing that is not the same as all other models so you'll see different results in the results csv (once updated).
    • Matching the original pre-processing as closely as possible I get these results:
    • dm_nfnet_f6
      - 86.352
    • dm_nfnet_f5
      - 86.100
    • dm_nfnet_f4
      - 85.834
    • dm_nfnet_f3
      - 85.676
    • dm_nfnet_f2
      - 85.178
    • dm_nfnet_f1
      - 84.696
    • dm_nfnet_f0
      - 83.464

Feb 16, 2021

  • Add Adaptive Gradient Clipping (AGC) as per Integrated w/ PyTorch gradient clipping via mode arg that defaults to prev 'norm' mode. For backward arg compat, clip-grad arg must be specified to enable when using
    • AGC w/ default clipping factor
      --clip-grad .01 --clip-mode agc
    • PyTorch global norm of 1.0 (old behaviour, always norm),
      --clip-grad 1.0
    • PyTorch value clipping of 10,
      --clip-grad 10. --clip-mode value
    • AGC performance is definitely sensitive to the clipping factor. More experimentation needed to determine good values for smaller batch sizes and optimizers besides those in paper. So far I've found .001-.005 is necessary for stable RMSProp training w/ NFNet/NF-ResNet.

Feb 12, 2021

  • Update Normalization-Free nets to include new NFNet-F ( model defs

Feb 10, 2021

  • First Normalization-Free model training experiments done,
    • nf_resnet50 - 80.68 top-1 @ 288x288, 80.31 @ 256x256
    • nfregnetb1 - 79.30 @ 288x288, 78.75 @ 256x256
  • More model archs, incl a flexible ByobNet backbone ('Bring-your-own-blocks')
    • GPU-Efficient-Networks (, impl in
    • RepVGG (, impl in
    • classic VGG (from torchvision, impl in
  • Refinements to normalizer layer arg handling and normalizer+act layer handling in some models
  • Default AMP mode changed to native PyTorch AMP instead of APEX. Issues not being fixed with APEX. Native works with
    model training, APEX does not.
  • Fix a few bugs introduced since last pypi release

Feb 8, 2021

  • Add several ResNet weights with ECA attention. 26t & 50t trained @ 256, test @ 320. 269d train @ 256, fine-tune @320, test @ 352.
    • ecaresnet26t
      - 79.88 top-1 @ 320x320, 79.08 @ 256x256
    • ecaresnet50t
      - 82.35 top-1 @ 320x320, 81.52 @ 256x256
    • ecaresnet269d
      - 84.93 top-1 @ 352x352, 84.87 @ 320x320
  • Remove separate tiered (
    ) vs tierednarrow (
    ) ResNet model defs, all
    changed to
    models removed (`seresnext26t
    32x4d` only model w/ weights that was removed).
  • Support model defaultcfgs with separate train vs test resolution `testinputsize
    and remove extra
    320` suffix ResNet model defs that were just for test.

Jan 30, 2021

  • Add initial "Normalization Free" NF-RegNet-B* and NF-ResNet model definitions based on paper

Jan 25, 2021

  • Add ResNetV2 Big Transfer (BiT) models w/ ImageNet-1k and 21k weights from
  • Add official R50+ViT-B/16 hybrid models + weights from
  • ImageNet-21k ViT weights are added w/ model defs and representation layer (pre logits) support
    • NOTE: ImageNet-21k classifier heads were zero'd in original weights, they are only useful for transfer learning
  • Add model defs and weights for DeiT Vision Transformer models from
  • Refactor dataset classes into ImageDataset/IterableImageDataset + dataset specific parser classes
  • Add Tensorflow-Datasets (TFDS) wrapper to allow use of TFDS image classification sets with train script
    • Ex: /data/tfds --dataset tfds/oxford_iiit_pet --val-split test --model resnet50 -b 256 --amp --num-classes 37 --opt adamw --lr 3e-4 --weight-decay .001 --pretrained -j 2
  • Add improved .tar dataset parser that reads images from .tar, folder of .tar files, or .tar within .tar
    • Run validation on full ImageNet-21k directly from tar w/ BiT model: /data/fall11_whole.tar --model resnetv2_50x1_bitm_in21k --amp
  • Models in this update should be stable w/ possible exception of ViT/BiT, possibility of some regressions with train/val scripts and dataset handling

Jan 3, 2021

  • Add SE-ResNet-152D weights
    • 256x256 val, 0.94 crop top-1 - 83.75
    • 320x320 val, 1.0 crop - 84.36
  • Update results files


PyTorch Image Models (

) is a collection of image models, layers, utilities, optimizers, schedulers, data-loaders / augmentations, and reference training / validation scripts that aim to pull together a wide variety of SOTA models with ability to reproduce ImageNet training results.

The work of many others is present here. I've tried to make sure all source material is acknowledged via links to github, arxiv papers, etc in the README, documentation, and code docstrings. Please let me know if I missed anything.


All model architecture families include variants with pretrained weights. There are specific model variants without any weights, it is NOT a bug. Help training new or better weights is always appreciated. Here are some example training hparams to get you started.

A full version of the list below with source links can be found in the documentation.

  • Aggregating Nested Transformers -
  • Big Transfer ResNetV2 (BiT) -
  • Bottleneck Transformers -
  • CaiT (Class-Attention in Image Transformers) -
  • CoaT (Co-Scale Conv-Attentional Image Transformers) -
  • ConViT (Soft Convolutional Inductive Biases Vision Transformers)-
  • CspNet (Cross-Stage Partial Networks) -
  • DeiT (Vision Transformer) -
  • DenseNet -
  • DLA -
  • DPN (Dual-Path Network) -
  • EfficientNet (MBConvNet Family)
    • EfficientNet NoisyStudent (B0-B7, L2) -
    • EfficientNet AdvProp (B0-B8) -
    • EfficientNet (B0-B7) -
    • EfficientNet-EdgeTPU (S, M, L) -
    • EfficientNet V2 -
    • FBNet-C -
    • MixNet -
    • MNASNet B1, A1 (Squeeze-Excite), and Small -
    • MobileNet-V2 -
    • Single-Path NAS -
  • GhostNet -
  • gMLP -
  • GPU-Efficient Networks -
  • Halo Nets -
  • HardCoRe-NAS -
  • HRNet -
  • Inception-V3 -
  • Inception-ResNet-V2 and Inception-V4 -
  • Lambda Networks -
  • LeViT (Vision Transformer in ConvNet's Clothing) -
  • MLP-Mixer -
  • MobileNet-V3 (MBConvNet w/ Efficient Head) -
  • NASNet-A -
  • NFNet-F -
  • NF-RegNet / NF-ResNet -
  • PNasNet -
  • Pooling-based Vision Transformer (PiT) -
  • RegNet -
  • RepVGG -
  • ResMLP -
  • ResNet/ResNeXt
    • ResNet (v1b/v1.5) -
    • ResNeXt -
    • 'Bag of Tricks' / Gluon C, D, E, S variations -
    • Weakly-supervised (WSL) Instagram pretrained / ImageNet tuned ResNeXt101 -
    • Semi-supervised (SSL) / Semi-weakly Supervised (SWSL) ResNet/ResNeXts -
    • ECA-Net (ECAResNet) -
    • Squeeze-and-Excitation Networks (SEResNet) -
    • ResNet-RS -
  • Res2Net -
  • ResNeSt -
  • ReXNet -
  • SelecSLS -
  • Selective Kernel Networks -
  • Swin Transformer -
  • Transformer-iN-Transformer (TNT) -
  • TResNet -
  • Twins (Spatial Attention in Vision Transformers) -
  • Vision Transformer -
  • VovNet V2 and V1 -
  • Xception -
  • Xception (Modified Aligned, Gluon) -
  • Xception (Modified Aligned, TF) -
  • XCiT (Cross-Covariance Image Transformers) -


Several (less common) features that I often utilize in my projects are included. Many of their additions are the reason why I maintain my own set of models, instead of using others' via PIP:

  • All models have a common default configuration interface and API for
    • accessing/changing the classifier -
    • doing a forward pass on just the features -
      (see documentation)
    • these makes it easy to write consistent network wrappers that work with any of the models
  • All models support multi-scale feature map extraction (feature pyramids) via create_model (see documentation)
    • create_model(name, features_only=True, out_indices=..., output_stride=...)
    • out_indices
      creation arg specifies which feature maps to return, these indices are 0 based and generally correspond to the
      C(i + 1)
      feature level.
    • output_stride
      creation arg controls output stride of the network by using dilated convolutions. Most networks are stride 32 by default. Not all networks support this.
    • feature map channel counts, reduction level (stride) can be queried AFTER model creation via the
  • All models have a consistent pretrained weight loader that adapts last linear if necessary, and from 3 to 1 channel input if desired
  • High performance reference training, validation, and inference scripts that work in several process/GPU modes:
    • NVIDIA DDP w/ a single GPU per process, multiple processes with APEX present (AMP mixed-precision optional)
    • PyTorch DistributedDataParallel w/ multi-gpu, single process (AMP disabled as it crashes when enabled)
    • PyTorch w/ single GPU single process (AMP optional)
  • A dynamic global pool implementation that allows selecting from average pooling, max pooling, average + max, or concat([average, max]) at model creation. All global pooling is adaptive average by default and compatible with pretrained weights.
  • A 'Test Time Pool' wrapper that can wrap any of the included models and usually provides improved performance doing inference with input images larger than the training size. Idea adapted from original DPN implementation when I ported (
  • Learning rate schedulers
    • Ideas adopted from
    • Schedulers include
      w/ restarts,
      w/ restarts,
  • Optimizers:
    • rmsprop_tf
      adapted from PyTorch RMSProp by myself. Reproduces much improved Tensorflow RMSProp behaviour.
    • radam
      by Liyuan Liu (
    • novograd
      by Masashi Kimura (
    • lookahead
      adapted from impl by Liam (
    • fused
      optimizers by name with NVIDIA Apex installed
    • adamp
      by Naver ClovAI (
    • adafactor
      adapted from FAIRSeq impl (
    • adahessian
      by David Samuel (
  • Random Erasing from Zhun Zhong (
  • Mixup (
  • CutMix (
  • AutoAugment ( and RandAugment ( ImageNet configurations modeled after impl for EfficientNet training (
  • AugMix w/ JSD loss (, JSD w/ clean + augmented mixing support works with AutoAugment and RandAugment as well
  • SplitBachNorm - allows splitting batch norm layers between clean and augmented (auxiliary batch norm) data
  • DropPath aka "Stochastic Depth" (
  • DropBlock (
  • Blur Pooling (
  • Space-to-Depth by mrT23 ( -- original paper?
  • Adaptive Gradient Clipping (,
  • An extensive selection of channel and/or spatial attention modules:
    • Bottleneck Transformer -
    • CBAM -
    • Effective Squeeze-Excitation (ESE) -
    • Efficient Channel Attention (ECA) -
    • Gather-Excite (GE) -
    • Global Context (GC) -
    • Halo -
    • Involution -
    • Lambda Layer -
    • Non-Local (NL) -
    • Squeeze-and-Excitation (SE) -
    • Selective Kernel (SK) - (
    • Split (SPLAT) -
    • Shifted Window (SWIN) -


Model validation results can be found in the documentation and in the results tables

Getting Started (Documentation)

My current documentation for

covers the basics.

timmdocs is quickly becoming a much more comprehensive set of documentation for

. A big thanks to Aman Arora for his efforts creating timmdocs.

paperswithcode is a good resource for browsing the models within


Train, Validation, Inference Scripts

The root folder of the repository contains reference train, validation, and inference scripts that work with the included models and other features of this repository. They are adaptable for other datasets and use cases with a little hacking. See documentation for some basics and training hparams for some train examples that produce SOTA ImageNet results.

Awesome PyTorch Resources

One of the greatest assets of PyTorch is the community and their contributions. A few of my favourite resources that pair well with the models and components here are listed below.

Object Detection, Instance and Semantic Segmentation

  • Detectron2 -
  • Segmentation Models (Semantic) -
  • EfficientDet (Obj Det, Semantic soon) -

Computer Vision / Image Augmentation

  • Albumentations -
  • Kornia -

Knowledge Distillation

  • RepDistiller -
  • torchdistill -

Metric Learning

  • PyTorch Metric Learning -

Training / Frameworks

  • fastai -



The code here is licensed Apache 2.0. I've taken care to make sure any third party code included or adapted has compatible (permissive) licenses such as MIT, BSD, etc. I've made an effort to avoid any GPL / LGPL conflicts. That said, it is your responsibility to ensure you comply with licenses here and conditions of any dependent licenses. Where applicable, I've linked the sources/references for various components in docstrings. If you think I've missed anything please create an issue.

Pretrained Weights

So far all of the pretrained weights available here are pretrained on ImageNet with a select few that have some additional pretraining (see extra note below). ImageNet was released for non-commercial research purposes only ( It's not clear what the implications of that are for the use of pretrained weights from that dataset. Any models I have trained with ImageNet are done for research purposes and one should assume that the original dataset license applies to the weights. It's best to seek legal advice if you intend to use the pretrained weights in a commercial product.

Pretrained on more than ImageNet

Several weights included or references here were pretrained with proprietary datasets that I do not have access to. These include the Facebook WSL, SSL, SWSL ResNe(Xt) and the Google Noisy Student EfficientNet models. The Facebook models have an explicit non-commercial license (CC-BY-NC 4.0,, The Google models do not appear to have any restriction beyond the Apache 2.0 license (and ImageNet concerns). In either case, you should contact Facebook or Google with any questions.



  author = {Ross Wightman},
  title = {PyTorch Image Models},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  doi = {10.5281/zenodo.4414861},
  howpublished = {\url{}}

Latest DOI


We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.