- What's New
- Introduction
- Models
- Features
- Results
- Getting Started (Documentation)
- Train, Validation, Inference Scripts
- Awesome PyTorch Resources
- Licenses
- Citing
âť—Updates after Oct 10, 2022 are available in version >= 0.9âť—
- Many changes since the last 0.6.x stable releases. They were previewed in 0.8.x dev releases but not everyone transitioned.
timm.models.layersmoved totimm.layers:from timm.models.layers import namewill still work via deprecation mapping (but please transition totimm.layers).import timm.models.layers.moduleorfrom timm.models.layers.module import nameneeds to be changed now.
- Builder, helper, non-model modules in
timm.modelshave a_prefix added, ietimm.models.helpers->timm.models._helpers, there are temporary deprecation mapping files but those will be removed. - All models now support
architecture.pretrained_tagnaming (exresnet50.rsb_a1).- The pretrained_tag is the specific weight variant (different head) for the architecture.
- Using only
architecturedefaults to the first weights in the default_cfgs for that model architecture. - In adding pretrained tags, many model names that existed to differentiate were renamed to use the tag (ex:
vit_base_patch16_224_in21k->vit_base_patch16_224.augreg_in21k). There are deprecation mappings for these.
- A number of models had their checkpoints remaped to match architecture changes needed to better support
features_only=True, there arecheckpoint_filter_fnmethods in any model module that was remapped. These can be passed totimm.models.load_checkpoint(..., filter_fn=timm.models.swin_transformer_v2.checkpoint_filter_fn)to remap your existing checkpoint. - The Hugging Face Hub (https://huggingface.co/timm) is now the primary source for
timmweights. Model cards include link to papers, original source, license. - Previous 0.6.x can be cloned from 0.6.x branch or installed via pip with version.
- MobileNetV4 models and initial set of
timmtrained weights added:
| model | top1 | top1_err | top5 | top5_err | param_count | img_size |
|---|---|---|---|---|---|---|
| mobilenetv4_hybrid_large.e600_r384_in1k | 84.266 | 15.734 | 96.936 | 3.064 | 37.76 | 448 |
| mobilenetv4_hybrid_large.e600_r384_in1k | 83.800 | 16.200 | 96.770 | 3.230 | 37.76 | 384 |
| mobilenetv4_conv_large.e600_r384_in1k | 83.392 | 16.608 | 96.622 | 3.378 | 32.59 | 448 |
| mobilenetv4_conv_large.e600_r384_in1k | 82.952 | 17.048 | 96.266 | 3.734 | 32.59 | 384 |
| mobilenetv4_conv_large.e500_r256_in1k | 82.674 | 17.326 | 96.31 | 3.69 | 32.59 | 320 |
| mobilenetv4_conv_large.e500_r256_in1k | 81.862 | 18.138 | 95.69 | 4.31 | 32.59 | 256 |
| mobilenetv4_hybrid_medium.e500_r224_in1k | 81.276 | 18.724 | 95.742 | 4.258 | 11.07 | 256 |
| mobilenetv4_conv_medium.e500_r256_in1k | 80.858 | 19.142 | 95.768 | 4.232 | 9.72 | 320 |
| mobilenetv4_hybrid_medium.e500_r224_in1k | 80.442 | 19.558 | 95.38 | 4.62 | 11.07 | 224 |
| mobilenetv4_conv_blur_medium.e500_r224_in1k | 80.142 | 19.858 | 95.298 | 4.702 | 9.72 | 256 |
| mobilenetv4_conv_medium.e500_r256_in1k | 79.928 | 20.072 | 95.184 | 4.816 | 9.72 | 256 |
| mobilenetv4_conv_medium.e500_r224_in1k | 79.808 | 20.192 | 95.186 | 4.814 | 9.72 | 256 |
| mobilenetv4_conv_blur_medium.e500_r224_in1k | 79.438 | 20.562 | 94.932 | 5.068 | 9.72 | 224 |
| mobilenetv4_conv_medium.e500_r224_in1k | 79.094 | 20.906 | 94.77 | 5.23 | 9.72 | 224 |
| mobilenetv4_conv_small.e2400_r224_in1k | 74.616 | 25.384 | 92.072 | 7.928 | 3.77 | 256 |
| mobilenetv4_conv_small.e1200_r224_in1k | 74.292 | 25.708 | 92.116 | 7.884 | 3.77 | 256 |
| mobilenetv4_conv_small.e2400_r224_in1k | 73.756 | 26.244 | 91.422 | 8.578 | 3.77 | 224 |
| mobilenetv4_conv_small.e1200_r224_in1k | 73.454 | 26.546 | 91.34 | 8.66 | 3.77 | 224 |
- Apple MobileCLIP (https://arxiv.org/pdf/2311.17049, FastViT and ViT-B) image tower model support & weights added (part of OpenCLIP support).
- ViTamin (https://arxiv.org/abs/2404.02132) CLIP image tower model & weights added (part of OpenCLIP support).
- OpenAI CLIP Modified ResNet image tower modelling & weight support (via ByobNet). Refactor AttentionPool2d.
- Support loading PaliGemma jax weights into SigLIP ViT models with average pooling.
- Add Hiera models from Meta (https://github.com/facebookresearch/hiera).
- Add
normalize=flag for transorms, return non-normalized torch.Tensor with original dytpe (forchug) - Version 1.0.3 release
Searching for Better ViT Baselines (For the GPU Poor)weights and vit variants released. Exploring model shapes between Tiny and Base.
- AttentionExtract helper added to extract attention maps from
timmmodels. See example in huggingface#1232 (comment) forward_intermediates()API refined and added to more models including some ConvNets that have other extraction methods.- 1017 of 1047 model architectures support
features_only=Truefeature extraction. Remaining 34 architectures can be supported but based on priority requests. - Remove torch.jit.script annotated functions including old JIT activations. Conflict with dynamo and dynamo does a much better job when used.
- Prepping for a long overdue 1.0 release, things have been stable for a while now.
- Significant feature that's been missing for a while,
features_only=Truesupport for ViT models with flat hidden states or non-std module layouts (so far covering'vit_*', 'twins_*', 'deit*', 'beit*', 'mvitv2*', 'eva*', 'samvit_*', 'flexivit*') - Above feature support achieved through a new
forward_intermediates()API that can be used with a feature wrapping module or direclty.
model = timm.create_model('vit_base_patch16_224')
final_feat, intermediates = model.forward_intermediates(input)
output = model.forward_head(final_feat) # pooling + classifier head
print(final_feat.shape)
torch.Size([2, 197, 768])
for f in intermediates:
print(f.shape)
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
torch.Size([2, 768, 14, 14])
print(output.shape)
torch.Size([2, 1000])model = timm.create_model('eva02_base_patch16_clip_224', pretrained=True, img_size=512, features_only=True, out_indices=(-3, -2,))
output = model(torch.randn(2, 3, 512, 512))
for o in output:
print(o.shape)
torch.Size([2, 768, 32, 32])
torch.Size([2, 768, 32, 32])- TinyCLIP vision tower weights added, thx Thien Tran
- Next-ViT models added. Adapted from https://github.com/bytedance/Next-ViT
- HGNet and PP-HGNetV2 models added. Adapted from https://github.com/PaddlePaddle/PaddleClas by SeeFun
- Removed setup.py, moved to pyproject.toml based build supported by PDM
- Add updated model EMA impl using _for_each for less overhead
- Support device args in train script for non GPU devices
- Other misc fixes and small additions
- Min supported Python version increased to 3.8
- Release 0.9.16
Datasets & transform refactoring
- HuggingFace streaming (iterable) dataset support (
--dataset hfids:org/dataset) - Webdataset wrapper tweaks for improved split info fetching, can auto fetch splits from supported HF hub webdataset
- Tested HF
datasetsand webdataset wrapper streaming from HF hub with recenttimmImageNet uploads to https://huggingface.co/timm - Make input & target column/field keys consistent across datasets and pass via args
- Full monochrome support when using e:g:
--input-size 1 224 224or--in-chans 1, sets PIL image conversion appropriately in dataset - Improved several alternate crop & resize transforms (ResizeKeepRatio, RandomCropOrPad, etc) for use in PixParse document AI project
- Add SimCLR style color jitter prob along with grayscale and gaussian blur options to augmentations and args
- Allow train without validation set (
--val-split '') in train script - Add
--bce-sum(sum over class dim) and--bce-pos-weight(positive weighting) args for training as they're common BCE loss tweaks I was often hard coding
- Added EfficientViT-Large models, thanks SeeFun
- Fix Python 3.7 compat, will be dropping support for it soon
- Other misc fixes
- Release 0.9.12
- Added significant flexibility for Hugging Face Hub based timm models via
model_argsconfig entry.model_argswill be passed as kwargs through to models on creation. - Updated imagenet eval and test set csv files with latest models
vision_transformer.pytyping and doc cleanup by Laureηt- 0.9.11 release
- DFN (Data Filtering Networks) and MetaCLIP ViT weights added
- DINOv2 'register' ViT model weights added (https://huggingface.co/papers/2309.16588, https://huggingface.co/papers/2304.07193)
- Add
quickgeluViT variants for OpenAI, DFN, MetaCLIP weights that use it (less efficient) - Improved typing added to ResNet, MobileNet-v3 thanks to Aryan
- ImageNet-12k fine-tuned (from LAION-2B CLIP)
convnext_xxlarge - 0.9.9 release
- SigLIP image tower weights supported in
vision_transformer.py.- Great potential for fine-tune and downstream feature use.
- Experimental 'register' support in vit models as per Vision Transformers Need Registers
- Updated RepViT with new weight release. Thanks wangao
- Add patch resizing support (on pretrained weight load) to Swin models
- 0.9.8 release pending
- TinyViT added by SeeFun
- Fix EfficientViT (MIT) to use torch.autocast so it works back to PT 1.10
- 0.9.7 release
- Add dynamic img size support to models in
vision_transformer.py,vision_transformer_hybrid.py,deit.py, andeva.pyw/o breaking backward compat.- Add
dynamic_img_size=Trueto args at model creation time to allow changing the grid size (interpolate abs and/or ROPE pos embed each forward pass). - Add
dynamic_img_pad=Trueto allow image sizes that aren't divisible by patch size (pad bottom right to patch size each forward pass). - Enabling either dynamic mode will break FX tracing unless PatchEmbed module added as leaf.
- Existing method of resizing position embedding by passing different
img_size(interpolate pretrained embed weights once) on creation still works. - Existing method of changing
patch_size(resize pretrained patch_embed weights once) on creation still works. - Example validation cmd
python validate.py /imagenet --model vit_base_patch16_224 --amp --amp-dtype bfloat16 --img-size 255 --crop-pct 1.0 --model-kwargs dynamic_img_size=True dyamic_img_pad=True
- Add
- Many new models since last release
- FastViT - https://arxiv.org/abs/2303.14189
- MobileOne - https://arxiv.org/abs/2206.04040
- InceptionNeXt - https://arxiv.org/abs/2303.16900
- RepGhostNet - https://arxiv.org/abs/2211.06088 (thanks https://github.com/ChengpengChen)
- GhostNetV2 - https://arxiv.org/abs/2211.12905 (thanks https://github.com/yehuitang)
- EfficientViT (MSRA) - https://arxiv.org/abs/2305.07027 (thanks https://github.com/seefun)
- EfficientViT (MIT) - https://arxiv.org/abs/2205.14756 (thanks https://github.com/seefun)
- Add
--reparamarg tobenchmark.py,onnx_export.py, andvalidate.pyto trigger layer reparameterization / fusion for models with any one ofreparameterize(),switch_to_deploy()orfuse()- Including FastViT, MobileOne, RepGhostNet, EfficientViT (MSRA), RepViT, RepVGG, and LeViT
- Preparing 0.9.6 'back to school' release
- Swin, MaxViT, CoAtNet, and BEiT models support resizing of image/window size on creation with adaptation of pretrained weights
- Example validation cmd to test w/ non-square resize
python validate.py /imagenet --model swin_base_patch4_window7_224.ms_in22k_ft_in1k --amp --amp-dtype bfloat16 --input-size 3 256 320 --model-kwargs window_size=8,10 img_size=256,320
- Add GluonCV weights for HRNet w18_small and w18_small_v2. Converted by SeeFun
- Fix
selecsls*model naming regression - Patch and position embedding for ViT/EVA works for bfloat16/float16 weights on load (or activations for on-the-fly resize)
- v0.9.5 release prep
- Added timm trained
seresnextaa201d_32x8d.sw_in12k_ft_in1k_384weights (and.sw_in12kpretrain) with 87.3% top-1 on ImageNet-1k, best ImageNet ResNet family model I'm aware of. - RepViT model and weights (https://arxiv.org/abs/2307.09283) added by wangao
- I-JEPA ViT feature weights (no classifier) added by SeeFun
- SAM-ViT (segment anything) feature weights (no classifier) added by SeeFun
- Add support for alternative feat extraction methods and -ve indices to EfficientNet
- Add NAdamW optimizer
- Misc fixes
timm0.9 released, transition from 0.8.xdev releases
- Hugging Face Hub downloading is now default, 1132 models on https://huggingface.co/timm, 1163 weights in
timm - DINOv2 vit feature backbone weights added thanks to Leng Yue
- FB MAE vit feature backbone weights added
- OpenCLIP DataComp-XL L/14 feat backbone weights added
- MetaFormer (poolformer-v2, caformer, convformer, updated poolformer (v1)) w/ weights added by Fredo Guan
- Experimental
get_intermediate_layersfunction on vit/deit models for grabbing hidden states (inspired by DINO impl). This is WIP and may change significantly... feedback welcome. - Model creation throws error if
pretrained=Trueand no weights exist (instead of continuing with random initialization) - Fix regression with inception / nasnet TF sourced weights with 1001 classes in original classifiers
- bitsandbytes (https://github.com/TimDettmers/bitsandbytes) optimizers added to factory, use
bnbprefix, iebnbadam8bit - Misc cleanup and fixes
- Final testing before switching to a 0.9 and bringing
timmout of pre-release state
- 97% of
timmmodels uploaded to HF Hub and almost all updated to support multi-weight pretrained configs - Minor cleanup and refactoring of another batch of models as multi-weight added. More fused_attn (F.sdpa) and features_only support, and torchscript fixes.
- Gradient accumulation support added to train script and tested (
--grad-accum-steps), thanks Taeksang Kim - More weights on HF Hub (cspnet, cait, volo, xcit, tresnet, hardcorenas, densenet, dpn, vovnet, xception_aligned)
- Added
--head-init-scaleand--head-init-biasto train.py to scale classiifer head and set fixed bias for fine-tune - Remove all InplaceABN (
inplace_abn) use, replaced use in tresnet with standard BatchNorm (modified weights accordingly).
- Add ONNX export script, validate script, helpers that I've had kicking around for along time. Tweak 'same' padding for better export w/ recent ONNX + pytorch.
- Refactor dropout args for vit and vit-like models, separate drop_rate into
drop_rate(classifier dropout),proj_drop_rate(block mlp / out projections),pos_drop_rate(position embedding drop),attn_drop_rate(attention dropout). Also add patch dropout (FLIP) to vit and eva models. - fused F.scaled_dot_product_attention support to more vit models, add env var (TIMM_FUSED_ATTN) to control, and config interface to enable/disable
- Add EVA-CLIP backbones w/ image tower weights, all the way up to 4B param 'enormous' model, and 336x336 OpenAI ViT mode that was missed.
- ALL ResNet models pushed to Hugging Face Hub with multi-weight support
- All past
timmtrained weights added with recipe based tags to differentiate - All ResNet strikes back A1/A2/A3 (seed 0) and R50 example B/C1/C2/D weights available
- Add torchvision v2 recipe weights to existing torchvision originals
- See comparison table in https://huggingface.co/timm/seresnextaa101d_32x8d.sw_in12k_ft_in1k_288#model-comparison
- All past
- New ImageNet-12k + ImageNet-1k fine-tunes available for a few anti-aliased ResNet models
resnetaa50d.sw_in12k_ft_in1k- 81.7 @ 224, 82.6 @ 288resnetaa101d.sw_in12k_ft_in1k- 83.5 @ 224, 84.1 @ 288seresnextaa101d_32x8d.sw_in12k_ft_in1k- 86.0 @ 224, 86.5 @ 288seresnextaa101d_32x8d.sw_in12k_ft_in1k_288- 86.5 @ 288, 86.7 @ 320
- Add first ConvNext-XXLarge CLIP -> IN-1k fine-tune and IN-12k intermediate fine-tunes for convnext-base/large CLIP models.
| model | top1 | top5 | img_size | param_count | gmacs | macts |
|---|---|---|---|---|---|---|
| convnext_xxlarge.clip_laion2b_soup_ft_in1k | 88.612 | 98.704 | 256 | 846.47 | 198.09 | 124.45 |
| convnext_large_mlp.clip_laion2b_soup_ft_in12k_in1k_384 | 88.312 | 98.578 | 384 | 200.13 | 101.11 | 126.74 |
| convnext_large_mlp.clip_laion2b_soup_ft_in12k_in1k_320 | 87.968 | 98.47 | 320 | 200.13 | 70.21 | 88.02 |
| convnext_base.clip_laion2b_augreg_ft_in12k_in1k_384 | 87.138 | 98.212 | 384 | 88.59 | 45.21 | 84.49 |
| convnext_base.clip_laion2b_augreg_ft_in12k_in1k | 86.344 | 97.97 | 256 | 88.59 | 20.09 | 37.55 |
- Add EVA-02 MIM pretrained and fine-tuned weights, push to HF hub and update model cards for all EVA models. First model over 90% top-1 (99% top-5)! Check out the original code & weights at https://github.com/baaivision/EVA for more details on their work blending MIM, CLIP w/ many model, dataset, and train recipe tweaks.
| model | top1 | top5 | param_count | img_size |
|---|---|---|---|---|
| eva02_large_patch14_448.mim_m38m_ft_in22k_in1k | 90.054 | 99.042 | 305.08 | 448 |
| eva02_large_patch14_448.mim_in22k_ft_in22k_in1k | 89.946 | 99.01 | 305.08 | 448 |
| eva_giant_patch14_560.m30m_ft_in22k_in1k | 89.792 | 98.992 | 1014.45 | 560 |
| eva02_large_patch14_448.mim_in22k_ft_in1k | 89.626 | 98.954 | 305.08 | 448 |
| eva02_large_patch14_448.mim_m38m_ft_in1k | 89.57 | 98.918 | 305.08 | 448 |
| eva_giant_patch14_336.m30m_ft_in22k_in1k | 89.56 | 98.956 | 1013.01 | 336 |
| eva_giant_patch14_336.clip_ft_in1k | 89.466 | 98.82 | 1013.01 | 336 |
| eva_large_patch14_336.in22k_ft_in22k_in1k | 89.214 | 98.854 | 304.53 | 336 |
| eva_giant_patch14_224.clip_ft_in1k | 88.882 | 98.678 | 1012.56 | 224 |
| eva02_base_patch14_448.mim_in22k_ft_in22k_in1k | 88.692 | 98.722 | 87.12 | 448 |
| eva_large_patch14_336.in22k_ft_in1k | 88.652 | 98.722 | 304.53 | 336 |
| eva_large_patch14_196.in22k_ft_in22k_in1k | 88.592 | 98.656 | 304.14 | 196 |
| eva02_base_patch14_448.mim_in22k_ft_in1k | 88.23 | 98.564 | 87.12 | 448 |
| eva_large_patch14_196.in22k_ft_in1k | 87.934 | 98.504 | 304.14 | 196 |
| eva02_small_patch14_336.mim_in22k_ft_in1k | 85.74 | 97.614 | 22.13 | 336 |
| eva02_tiny_patch14_336.mim_in22k_ft_in1k | 80.658 | 95.524 | 5.76 | 336 |
- Multi-weight and HF hub for DeiT and MLP-Mixer based models
- More weights pushed to HF hub along with multi-weight support, including:
regnet.py,rexnet.py,byobnet.py,resnetv2.py,swin_transformer.py,swin_transformer_v2.py,swin_transformer_v2_cr.py - Swin Transformer models support feature extraction (NCHW feat maps for
swinv2_cr_*, and NHWC for all others) and spatial embedding outputs. - FocalNet (from https://github.com/microsoft/FocalNet) models and weights added with significant refactoring, feature extraction, no fixed resolution / sizing constraint
- RegNet weights increased with HF hub push, SWAG, SEER, and torchvision v2 weights. SEER is pretty poor wrt to performance for model size, but possibly useful.
- More ImageNet-12k pretrained and 1k fine-tuned
timmweights:rexnetr_200.sw_in12k_ft_in1k- 82.6 @ 224, 83.2 @ 288rexnetr_300.sw_in12k_ft_in1k- 84.0 @ 224, 84.5 @ 288regnety_120.sw_in12k_ft_in1k- 85.0 @ 224, 85.4 @ 288regnety_160.lion_in12k_ft_in1k- 85.6 @ 224, 86.0 @ 288regnety_160.sw_in12k_ft_in1k- 85.6 @ 224, 86.0 @ 288 (compare to SWAG PT + 1k FT this is same BUT much lower res, blows SEER FT away)
- Model name deprecation + remapping functionality added (a milestone for bringing 0.8.x out of pre-release). Mappings being added...
- Minor bug fixes and improvements.
- Add ConvNeXt-XXLarge CLIP pretrained image tower weights for fine-tune & features (fine-tuning TBD) -- see model card
- Update
convnext_xxlargedefault LayerNorm eps to 1e-5 (for CLIP weights, improved stability) - 0.8.15dev0
- Add 320x320
convnext_large_mlp.clip_laion2b_ft_320andconvnext_lage_mlp.clip_laion2b_ft_soup_320CLIP image tower weights for features & fine-tune - 0.8.13dev0 pypi release for latest changes w/ move to huggingface org
safetensorcheckpoint support added- Add ideas from 'Scaling Vision Transformers to 22 B. Params' (https://arxiv.org/abs/2302.05442) -- qk norm, RmsNorm, parallel block
- Add F.scaled_dot_product_attention support (PyTorch 2.0 only) to
vit_*,vit_relpos*,coatnet/maxxvit(to start) - Lion optimizer (w/ multi-tensor option) added (https://arxiv.org/abs/2302.06675)
- gradient checkpointing works with
features_only=True
PyTorch Image Models (timm) 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.
- Aggregating Nested Transformers - https://arxiv.org/abs/2105.12723
- BEiT - https://arxiv.org/abs/2106.08254
- Big Transfer ResNetV2 (BiT) - https://arxiv.org/abs/1912.11370
- Bottleneck Transformers - https://arxiv.org/abs/2101.11605
- CaiT (Class-Attention in Image Transformers) - https://arxiv.org/abs/2103.17239
- CoaT (Co-Scale Conv-Attentional Image Transformers) - https://arxiv.org/abs/2104.06399
- CoAtNet (Convolution and Attention) - https://arxiv.org/abs/2106.04803
- ConvNeXt - https://arxiv.org/abs/2201.03545
- ConvNeXt-V2 - http://arxiv.org/abs/2301.00808
- ConViT (Soft Convolutional Inductive Biases Vision Transformers)- https://arxiv.org/abs/2103.10697
- CspNet (Cross-Stage Partial Networks) - https://arxiv.org/abs/1911.11929
- DeiT - https://arxiv.org/abs/2012.12877
- DeiT-III - https://arxiv.org/pdf/2204.07118.pdf
- DenseNet - https://arxiv.org/abs/1608.06993
- DLA - https://arxiv.org/abs/1707.06484
- DPN (Dual-Path Network) - https://arxiv.org/abs/1707.01629
- EdgeNeXt - https://arxiv.org/abs/2206.10589
- EfficientFormer - https://arxiv.org/abs/2206.01191
- EfficientNet (MBConvNet Family)
- EfficientNet NoisyStudent (B0-B7, L2) - https://arxiv.org/abs/1911.04252
- EfficientNet AdvProp (B0-B8) - https://arxiv.org/abs/1911.09665
- EfficientNet (B0-B7) - https://arxiv.org/abs/1905.11946
- EfficientNet-EdgeTPU (S, M, L) - https://ai.googleblog.com/2019/08/efficientnet-edgetpu-creating.html
- EfficientNet V2 - https://arxiv.org/abs/2104.00298
- FBNet-C - https://arxiv.org/abs/1812.03443
- MixNet - https://arxiv.org/abs/1907.09595
- MNASNet B1, A1 (Squeeze-Excite), and Small - https://arxiv.org/abs/1807.11626
- MobileNet-V2 - https://arxiv.org/abs/1801.04381
- Single-Path NAS - https://arxiv.org/abs/1904.02877
- TinyNet - https://arxiv.org/abs/2010.14819
- EfficientViT (MIT) - https://arxiv.org/abs/2205.14756
- EfficientViT (MSRA) - https://arxiv.org/abs/2305.07027
- EVA - https://arxiv.org/abs/2211.07636
- EVA-02 - https://arxiv.org/abs/2303.11331
- FastViT - https://arxiv.org/abs/2303.14189
- FlexiViT - https://arxiv.org/abs/2212.08013
- FocalNet (Focal Modulation Networks) - https://arxiv.org/abs/2203.11926
- GCViT (Global Context Vision Transformer) - https://arxiv.org/abs/2206.09959
- GhostNet - https://arxiv.org/abs/1911.11907
- GhostNet-V2 - https://arxiv.org/abs/2211.12905
- gMLP - https://arxiv.org/abs/2105.08050
- GPU-Efficient Networks - https://arxiv.org/abs/2006.14090
- Halo Nets - https://arxiv.org/abs/2103.12731
- HGNet / HGNet-V2 - TBD
- HRNet - https://arxiv.org/abs/1908.07919
- InceptionNeXt - https://arxiv.org/abs/2303.16900
- Inception-V3 - https://arxiv.org/abs/1512.00567
- Inception-ResNet-V2 and Inception-V4 - https://arxiv.org/abs/1602.07261
- Lambda Networks - https://arxiv.org/abs/2102.08602
- LeViT (Vision Transformer in ConvNet's Clothing) - https://arxiv.org/abs/2104.01136
- MaxViT (Multi-Axis Vision Transformer) - https://arxiv.org/abs/2204.01697
- MetaFormer (PoolFormer-v2, ConvFormer, CAFormer) - https://arxiv.org/abs/2210.13452
- MLP-Mixer - https://arxiv.org/abs/2105.01601
- MobileCLIP - https://arxiv.org/abs/2311.17049
- MobileNet-V3 (MBConvNet w/ Efficient Head) - https://arxiv.org/abs/1905.02244
- FBNet-V3 - https://arxiv.org/abs/2006.02049
- HardCoRe-NAS - https://arxiv.org/abs/2102.11646
- LCNet - https://arxiv.org/abs/2109.15099
- MobileNetV4 - https://arxiv.org/abs/2404.10518
- MobileOne - https://arxiv.org/abs/2206.04040
- MobileViT - https://arxiv.org/abs/2110.02178
- MobileViT-V2 - https://arxiv.org/abs/2206.02680
- MViT-V2 (Improved Multiscale Vision Transformer) - https://arxiv.org/abs/2112.01526
- NASNet-A - https://arxiv.org/abs/1707.07012
- NesT - https://arxiv.org/abs/2105.12723
- Next-ViT - https://arxiv.org/abs/2207.05501
- NFNet-F - https://arxiv.org/abs/2102.06171
- NF-RegNet / NF-ResNet - https://arxiv.org/abs/2101.08692
- PNasNet - https://arxiv.org/abs/1712.00559
- PoolFormer (MetaFormer) - https://arxiv.org/abs/2111.11418
- Pooling-based Vision Transformer (PiT) - https://arxiv.org/abs/2103.16302
- PVT-V2 (Improved Pyramid Vision Transformer) - https://arxiv.org/abs/2106.13797
- RegNet - https://arxiv.org/abs/2003.13678
- RegNetZ - https://arxiv.org/abs/2103.06877
- RepVGG - https://arxiv.org/abs/2101.03697
- RepGhostNet - https://arxiv.org/abs/2211.06088
- RepViT - https://arxiv.org/abs/2307.09283
- ResMLP - https://arxiv.org/abs/2105.03404
- ResNet/ResNeXt
- ResNet (v1b/v1.5) - https://arxiv.org/abs/1512.03385
- ResNeXt - https://arxiv.org/abs/1611.05431
- 'Bag of Tricks' / Gluon C, D, E, S variations - https://arxiv.org/abs/1812.01187
- Weakly-supervised (WSL) Instagram pretrained / ImageNet tuned ResNeXt101 - https://arxiv.org/abs/1805.00932
- Semi-supervised (SSL) / Semi-weakly Supervised (SWSL) ResNet/ResNeXts - https://arxiv.org/abs/1905.00546
- ECA-Net (ECAResNet) - https://arxiv.org/abs/1910.03151v4
- Squeeze-and-Excitation Networks (SEResNet) - https://arxiv.org/abs/1709.01507
- ResNet-RS - https://arxiv.org/abs/2103.07579
- Res2Net - https://arxiv.org/abs/1904.01169
- ResNeSt - https://arxiv.org/abs/2004.08955
- ReXNet - https://arxiv.org/abs/2007.00992
- SelecSLS - https://arxiv.org/abs/1907.00837
- Selective Kernel Networks - https://arxiv.org/abs/1903.06586
- Sequencer2D - https://arxiv.org/abs/2205.01972
- Swin S3 (AutoFormerV2) - https://arxiv.org/abs/2111.14725
- Swin Transformer - https://arxiv.org/abs/2103.14030
- Swin Transformer V2 - https://arxiv.org/abs/2111.09883
- Transformer-iN-Transformer (TNT) - https://arxiv.org/abs/2103.00112
- TResNet - https://arxiv.org/abs/2003.13630
- Twins (Spatial Attention in Vision Transformers) - https://arxiv.org/pdf/2104.13840.pdf
- Visformer - https://arxiv.org/abs/2104.12533
- Vision Transformer - https://arxiv.org/abs/2010.11929
- ViTamin - https://arxiv.org/abs/2404.02132
- VOLO (Vision Outlooker) - https://arxiv.org/abs/2106.13112
- VovNet V2 and V1 - https://arxiv.org/abs/1911.06667
- Xception - https://arxiv.org/abs/1610.02357
- Xception (Modified Aligned, Gluon) - https://arxiv.org/abs/1802.02611
- Xception (Modified Aligned, TF) - https://arxiv.org/abs/1802.02611
- XCiT (Cross-Covariance Image Transformers) - https://arxiv.org/abs/2106.09681
Included optimizers available via create_optimizer / create_optimizer_v2 factory methods:
adabeliefan implementation of AdaBelief adapted from https://github.com/juntang-zhuang/Adabelief-Optimizer - https://arxiv.org/abs/2010.07468adafactoradapted from FAIRSeq impl - https://arxiv.org/abs/1804.04235adahessianby David Samuel - https://arxiv.org/abs/2006.00719adampandsgdpby Naver ClovAI - https://arxiv.org/abs/2006.08217adanan implementation of Adan adapted from https://github.com/sail-sg/Adan - https://arxiv.org/abs/2208.06677lamban implementation of Lamb and LambC (w/ trust-clipping) cleaned up and modified to support use with XLA - https://arxiv.org/abs/1904.00962larsan implementation of LARS and LARC (w/ trust-clipping) - https://arxiv.org/abs/1708.03888lionand implementation of Lion adapted from https://github.com/google/automl/tree/master/lion - https://arxiv.org/abs/2302.06675lookaheadadapted from impl by Liam - https://arxiv.org/abs/1907.08610madgrad- and implementation of MADGRAD adapted from https://github.com/facebookresearch/madgrad - https://arxiv.org/abs/2101.11075nadaman implementation of Adam w/ Nesterov momentumnadamwan impementation of AdamW (Adam w/ decoupled weight-decay) w/ Nesterov momentum. A simplified impl based on https://github.com/mlcommons/algorithmic-efficiencynovogradby Masashi Kimura - https://arxiv.org/abs/1905.11286radamby Liyuan Liu - https://arxiv.org/abs/1908.03265rmsprop_tfadapted from PyTorch RMSProp by myself. Reproduces much improved Tensorflow RMSProp behavioursgdwand implementation of SGD w/ decoupled weight-decayfused<name>optimizers by name with NVIDIA Apex installedbits<name>optimizers by name with BitsAndBytes installed
- Random Erasing from Zhun Zhong - https://arxiv.org/abs/1708.04896)
- Mixup - https://arxiv.org/abs/1710.09412
- CutMix - https://arxiv.org/abs/1905.04899
- AutoAugment (https: CE77 //arxiv.org/abs/1805.09501) and RandAugment (https://arxiv.org/abs/1909.13719) ImageNet configurations modeled after impl for EfficientNet training (https://github.com/tensorflow/tpu/blob/master/models/official/efficientnet/autoaugment.py)
- AugMix w/ JSD loss, JSD w/ clean + augmented mixing support works with AutoAugment and RandAugment as well - https://arxiv.org/abs/1912.02781
- SplitBachNorm - allows splitting batch norm layers between clean and augmented (auxiliary batch norm) data
- DropPath aka "Stochastic Depth" - https://arxiv.org/abs/1603.09382
- DropBlock - https://arxiv.org/abs/1810.12890
- Blur Pooling - https://arxiv.org/abs/1904.11486
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 -
get_classifierandreset_classifier - doing a forward pass on just the features -
forward_features(see documentation) - these makes it easy to write consistent network wrappers that work with any of the models
- accessing/changing the classifier -
- 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_indicescreation arg specifies which feature maps to return, these indices are 0 based and generally correspond to theC(i + 1)feature level.output_stridecreation 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
.feature_infomember
- 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 (https://github.com/cypw/DPNs)
- Learning rate schedulers
- Ideas adopted from
- AllenNLP schedulers
- FAIRseq lr_scheduler
- SGDR: Stochastic Gradient Descent with Warm Restarts (https://arxiv.org/abs/1608.03983)
- Schedulers include
step,cosinew/ restarts,tanhw/ restarts,plateau
- Ideas adopted from
- Space-to-Depth by mrT23 (https://arxiv.org/abs/1801.04590) -- original paper?
- Adaptive Gradient Clipping (https://arxiv.org/abs/2102.06171, https://github.com/deepmind/deepmind-research/tree/master/nfnets)
- An extensive selection of channel and/or spatial attention modules:
- Bottleneck Transformer - https://arxiv.org/abs/2101.11605
- CBAM - https://arxiv.org/abs/1807.06521
- Effective Squeeze-Excitation (ESE) - https://arxiv.org/abs/1911.06667
- Efficient Channel Attention (ECA) - https://arxiv.org/abs/1910.03151
- Gather-Excite (GE) - https://arxiv.org/abs/1810.12348
- Global Context (GC) - https://arxiv.org/abs/1904.11492
- Halo - https://arxiv.org/abs/2103.12731
- Involution - https://arxiv.org/abs/2103.06255
- Lambda Layer - https://arxiv.org/abs/2102.08602
- Non-Local (NL) - https://arxiv.org/abs/1711.07971
- Squeeze-and-Excitation (SE) - https://arxiv.org/abs/1709.01507
- Selective Kernel (SK) - (https://arxiv.org/abs/1903.06586
- Split (SPLAT) - https://arxiv.org/abs/2004.08955
- Shifted Window (SWIN) - https://arxiv.org/abs/2103.14030
Model validation results can be found in the results tables
The official documentation can be found at https://huggingface.co/docs/hub/timm. Documentation contributions are welcome.
Getting Started with PyTorch Image Models (timm): A Practitioner’s Guide by Chris Hughes is an extensive blog post covering many aspects of timm in detail.
timmdocs is an alternate set of documentation for timm. A big thanks to Aman Arora for his efforts creating timmdocs.
paperswithcode is a good resource for browsing the models within timm.
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.
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.
- Detectron2 - https://github.com/facebookresearch/detectron2
- Segmentation Models (Semantic) - https://github.com/qubvel/segmentation_models.pytorch
- EfficientDet (Obj Det, Semantic soon) - https://github.com/rwightman/efficientdet-pytorch
- Albumentations - https://github.com/albumentations-team/albumentations
- Kornia - https://github.com/kornia/kornia
- RepDistiller - https://github.com/HobbitLong/RepDistiller
- torchdistill - https://github.com/yoshitomo-matsubara/torchdistill
- PyTorch Metric Learning - https://github.com/KevinMusgrave/pytorch-metric-learning
- fastai - https://github.com/fastai/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.
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 (https://image-net.org/download). 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.
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, https://github.com/facebookresearch/semi-supervised-ImageNet1K-models, https://github.com/facebookresearch/WSL-Images). 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.
@misc{rw2019timm,
author = {Ross Wightman},
title = {PyTorch Image Models},
year = {2019},
publisher = {GitHub},
journal = {GitHub repository},
doi = {10.5281/zenodo.4414861},
howpublished = {\url{https://github.com/rwightman/pytorch-image-models}}
}