2a7d256fd5
Re-enable mem-efficient/jit activations after torchscript tests
5 years ago
f902bcd54c
Layer refactoring continues, ResNet downsample rewrite for proper dilation in 3x3 and avg_pool cases
...
* select_conv2d -> create_conv2d
* added create_attn to create attention module from string/bool/module
* factor padding helpers into own file, use in both conv2d_same and avg_pool2d_same
* add some more test eca resnet variants
* minor tweaks, naming, comments, consistency
5 years ago
a99ec4e7d1
A bunch more layer reorg, splitting many layers into own files. Improve torchscript compatibility.
5 years ago
13746a33fc
Big move, layer modules and fn to timm/models/layers
5 years ago
f54612f648
Merge branch 'select_kernel' into attention
5 years ago
4defbbbaa8
Fix module name mistake, start layers sub-package
5 years ago
7011cd0902
A little bit of ECA cleanup
5 years ago
46471df7b2
Merge pull request #82 from VRandme/eca
...
ECA-Net Efficient Channel Attention
5 years ago
d0eb59ef46
Remove unused default_init for EfficientNets, experimenting with fanout calc for #84
5 years ago
e6a762346a
Implement Adaptive Kernel selection
...
When channel size is given,
calculate adaptive kernel size according to original paper.
Otherwise use the given kernel size(k_size), which defaults to 3
5 years ago
13e8da2b46
SelectKernel split_input works best when input channels split like grouped conv, but output is full width. Disable zero_init for SK nets, seems a bad combo.
5 years ago
6db087a1ff
Merge remote-tracking branch 'upstream/master' into eca
5 years ago
904c618040
Update EcaModule.py
...
Make pylint happy
(commas, unused imports, missed imports)
5 years ago
db91ba053b
EcaModule(CamelCase)
...
CamelCased EcaModule.
Renamed all instances of ecalayer to EcaModule.
eca_module.py->EcaModule.py
5 years ago
5c4991a088
Add PyTorch trained EfficientNet-ES weights from Andrew Lavin
5 years ago
d04ff95eda
Merge branch 'master' into eca
5 years ago
d63ae121d5
Clean up eca_module code
...
functionally similar
adjusted rwightman's version of reshaping and viewing.
Use F.pad for circular eca version for cleaner code
5 years ago
d66819d1f3
Indentation mistake. Fixes #81
5 years ago
f87fcd7e88
Implement Eca modules
...
implement ECA module by
1. adopting original eca_module.py into models folder
2. adding use_eca layer besides every instance of SE layer
5 years ago
4808b3c32f
Bump version for PyPi update, fix few out of date README items/mistakes, add README updates for TF EfficientNet-B8 (RandAugment)
5 years ago
7d07ebb660
Adding some configs to sknet, incl ResNet50 variants from 'Compounding ... Assembled Techniques' paper and original SKNet50
5 years ago
a9d2424fd1
Add separate zero_init_last_bn function to support more block variety without a mess
5 years ago
355aa152d5
Just leave it float for now, will look at fp16 later. Remove unused reference code.
5 years ago
ef457555d3
BlockDrop working on GPU
5 years ago
3ff19079f9
Missed nn_ops.py from last commit
5 years ago
9f11b4e8a2
Add ConvBnAct layer to parallel integrated SelectKernelConv, add support for DropPath and DropBlock to ResNet base and SK blocks
5 years ago
cefc9b7761
Move SelectKernelConv to conv2d_layers and more
...
* always apply attention in SelectKernelConv, leave MixedConv for no attention alternative
* make MixedConv torchscript compatible
* refactor first/previous dilation name to make more sense in ResNet* networks
5 years ago
9abe610931
Used wrong channel var for split
5 years ago
58e28dc7e7
Move Selective Kernel blocks/convs to their own sknet.py file
5 years ago
a93bae6dc5
A SelectiveKernelBasicBlock for more experiments
5 years ago
ad087b4b17
Missed bias=False in selection conv
5 years ago
c8b3d6b81a
Initial impl of Selective Kernel Networks. Very much a WIP.
5 years ago
1daa303744
Add support to Dataset for class id mapping file, clean up a bit of old logic. Add results file arg for validation and update script.
5 years ago
91534522f9
Add newly added TF ported EfficientNet-B8 weights (RandAugment)
5 years ago
12dbc74742
New ResNet50 JSD + RandAugment weights
5 years ago
2f41905ba5
Update ResNet50 weights to AuxMix trained 78.994 top-1. A few commentes re 'tiered_narrow' tn variant.
5 years ago
d9a6a9d0af
Merge pull request #74 from rwightman/augmix-jsd
...
AugMix, JSD loss, SplitBatchNorm (Auxiliary BN), and more
5 years ago
3eb4a96eda
Update AugMix, JSD, etc comments and references
5 years ago
a28117ea46
Add tiered narrow ResNet (tn) and weights for seresnext26tn_32x4d
5 years ago
833066b540
A few minor things in SplitBN
5 years ago
7547119891
Add SplitBatchNorm. AugMix, Rand/AutoAugment, Split (Aux) BatchNorm, Jensen-Shannon Divergence, RandomErasing all working together
5 years ago
2e955cfd0c
Update RandomErasing with some improved arg names, tweak to aspect range
5 years ago
3cc0f91e23
Fix augmix variable name scope overlap, default non-blended mode
5 years ago
ec0dd4053a
Add updated RandAugment trained EfficientNet-B0 trained weights from @michaelklachko
5 years ago
40fea63ebe
Add checkpoint averaging script. Add headers, shebangs, exec perms to all scripts
5 years ago
4666cc9aed
Add --pin-mem arg to enable dataloader pin_memory (showing more benefit in some scenarios now), also add --torchscript arg to validate.py for testing models with jit.script
5 years ago
53001dd292
ResNet / Res2Net additions:
...
* ResNet torchscript compat
* output_stride arg supported to limit network stride via dilations (support for dilation added to Res2Net)
* allow activation layer to be changed via act_layer arg
5 years ago
f96b3e5e92
InceptionResNetV2 torchscript compatible
5 years ago
19d93fe454
Add selecsls60 weights
5 years ago
0062c15fb0
Update checkpoint url with modelzoo compatible ones.
5 years ago
b5315e66b5
Streamline SelecSLS model without breaking checkpoint compat. Move cfg handling out of model class. Update feature/pooling behaviour to match current.
5 years ago
d59a756c16
Run PyCharm autoformat on selecsls and change mix cap variables and model names to all lower
5 years ago
fb3a0f4bb8
Merge pull request #65 from mehtadushy/selecsls
...
Incorporate SelecSLS Models
5 years ago
19fc205a4d
Update comments on the new SE-ResNeXt26 models
5 years ago
acc3ed2b8c
Add EfficientNet-B3 weights, trained from scratch with RA.
5 years ago
2404361f62
correct asset paths
5 years ago
31939311f6
Added SelecSLS models
5 years ago
1f4498f217
Add ResNet deep tiered stem and model weights for seresnext26t_32x4d and seresnext26d_32x4d
5 years ago
32012a44fd
Added SelecSLS model
5 years ago
73b78459dc
Add update RandAugment MixNet-XL weights
5 years ago
3afc2a4dc0
Some cleanup/improvements to AugMix impl:
...
* make 'increasing' levels for Contrast, Color, Brightness, Saturation ops
* remove recursion from faster blending mix
* add config striing parsing for AugMix
5 years ago
232ab7fb12
Working on an implementation of AugMix with JensenShannonDivergence loss that's compatible with my AutoAugment and RandAugment impl
5 years ago
a435ea1327
Change reduce_bn to distribute_bn, add ability to choose between broadcast and reduce (mean). Add crop_pct arg to allow selecting validation crop while training.
5 years ago
3bff2b21dc
Add support for keeping running bn stats the same across distributed training nodes before eval/save
5 years ago
0161de0127
Switch RandoErasing back to on GPU normal sampling
5 years ago
ff421e5e09
New PyTorch trained EfficientNet-B2 weights with my RandAugment impl
5 years ago
3bef524f9c
Finish with HRNet, weights and models updated. Improve consistency in model classifier/global pool treatment.
5 years ago
6ca0828166
Update EfficientNet comments, MobileNetV3 non-TF create fns, fix factory arg checks, bump PyTorch version req to 1.2
5 years ago
eccbadca74
Update EfficientNet comments
5 years ago
902d32fb16
Renamed gen_efficientnet.py -> efficientnet.py
5 years ago
5a0a8de7e3
ResNet updates:
...
* remove redundant GluonResNet model/blocks and use the code in ResNet for Gluon weights
* change SEModules back to using AdaptiveAvgPool instead of mean, PyTorch issue long fixed
5 years ago
a39cc43374
Bring EfficientNet and MobileNetV3 up to date with my gen-efficientnet repo
...
* Split MobileNetV3 and EfficientNet model files and put builder and blocks in own files (getting too large)
* Finalize CondConv EfficientNet variant
* Add the AdvProp weights files and B8 EfficientNet model
* Refine the feature extraction module for EfficientNet and MobileNetV3
5 years ago
ad93347548
Initial HRNet classification model commit
5 years ago
2393708650
Missed stashing of out_indices in model
5 years ago
35e8f0c5e7
Fixup a few comments, add PyTorch version aware Flatten and finish as_sequential for GenEfficientNet
5 years ago
7ac6db4543
Missed activations.py
5 years ago
506df0e3d0
Add CondConv support for EfficientNet into WIP for GenEfficientNet Feature extraction setup
5 years ago
576d360f20
Bring in JIT version of optimized swish activation from gen_efficientnet as default (while working on feature extraction functionality here).
5 years ago
7b83e67f77
Pass drop connect arg through to EfficientNet models
5 years ago
31453b039e
Update Auto/RandAugment comments, README, more.
...
* Add a weighted choice option for RandAugment
* Adjust magnitude noise/std naming, config
5 years ago
4243f076f1
Adding RandAugment to AutoAugment impl, some tweaks to AA included
5 years ago
0d58c50fb1
Add TF RandAug weights for B5/B7 EfficientNet models.
5 years ago
c099374771
Map pretrained checkpoint to cpu to avoid issue with some pretrained checkpoints still having CUDA tensors. Fixes #42
5 years ago
b93fcf0708
Add Facebook Research Semi-Supervised and Semi-Weakly Supervised ResNet model weights.
5 years ago
a9eb484835
Add memory efficient Swish impl
5 years ago
d3ba34ee7e
Fix Mobilenet V3 model name for sotabench. Minor res2net cleanup.
5 years ago
2680ad14bb
Add Res2Net and DLA to README
5 years ago
adbf770f16
Add Res2Net and DLA models w/ pretrained weights. Update sotabench.
5 years ago
4002c0d4ce
Fix AutoAugment abs translate calc
5 years ago
c06274e5a2
Add note on random selection of magnitude value
5 years ago
b750b76f67
More AutoAugment work. Ready to roll...
5 years ago
25d2088d9e
Working on auto-augment
5 years ago
aff194f42c
Merge pull request #32 from rwightman/opt
...
More optimizer work
5 years ago
64966f61f7
Add Nvidia's NovogGrad impl from Jasper (cleaner/faster than current) and Apex Fused optimizers
5 years ago
3d9c8a6489
Add support for new AMP checkpointing support w/ amp.state_dict
5 years ago
ba3c97c3ad
Some Lookahead cleanup and fixes
5 years ago
e9d2ec4d8e
Merge pull request #31 from rwightman/opt
...
Optimizers and more
5 years ago
fac58f609a
Add RAdam, NovoGrad, Lookahead, and AdamW optimizers, a few ResNet tweaks and scheduler factory tweak.
...
* Add some of the trendy new optimizers. Decent results but not clearly better than the standards.
* Can create a None scheduler for constant LR
* ResNet defaults to zero_init of last BN in residual
* add resnet50d config
5 years ago
81875d52a6
Update sotabench model list, add Mean-Max pooling DPN variants, disable download progress
5 years ago
f37e633e9b
Merge remote-tracking branch 'origin/re-exp' into opt
5 years ago
Ross Wightman