Ross Wightman
827a3d6010
Add current checkpoints output back to CheckpointSaver (via logger)
6 years ago
Ross Wightman
63961b36a2
Missed pnasnet entrypoint
6 years ago
Ross Wightman
6cc214bd7a
Consistency in model entrypoints
...
* move pretrained entrypoint arg to first pos to be closer to torchvision/hub
* change DPN weight URLS to my github location
6 years ago
Ross Wightman
b20bb58284
Distributed tweaks
...
* Support PyTorch native DDP as fallback if APEX not present
* Support SyncBN for both APEX and Torch native (if torch >= 1.1)
* EMA model does not appear to need DDP wrapper, no gradients, updated from wrapped model
6 years ago
Ross Wightman
6fc886acaf
Remove all prints, change most to logging calls, tweak alignment of batch logs, improve setup.py
6 years ago
Ross Wightman
1d7f2d93a6
Python >= 3.6 requirement
6 years ago
Ross Wightman
aa4354f466
Big re-org, working towards making pip/module as 'timm'
6 years ago
Ross Wightman
871f4c1b0c
Change some absolute imports to relative
6 years ago
Ross Wightman
b9f8d40b10
Fix pretrained override logic for validate, checkpoint always trump pretrained flag during model create
6 years ago
Ross Wightman
0e1fd11ad8
Merge pull request #12 from rwightman/ema-cleanup
...
Model weights Exponential Moving Average
6 years ago
Ross Wightman
1019414fd2
Add ported Tensorflow EfficientNet B4/B5 weights
6 years ago
Ross Wightman
c9a61b7d98
Add gitignore
6 years ago
Ross Wightman
c6b8259ad6
Update README.md
...
Add blurb about python environment
6 years ago
Ross Wightman
f3134973b5
Add the MnasNet-B1 variant weights, add A1/B1 model names as in the stand-alone repo, remove a bit of unused code
6 years ago
Ross Wightman
c1a84ecb22
dataset not passed through PrefetchLoader for inference script. Fix #10
...
* also, make top5 configurable for lower class count cases
6 years ago
Ross Wightman
2060e433c0
Add native PyTorch weights for SE-MnasNet aka MnasNet-A1
6 years ago
Ross Wightman
7dab6d1ec7
Default to img_size in model default_cfg, defer output folder creation until later in the init sequence
6 years ago
Ross Wightman
7d17394bdc
Add native PyTorch weights for EfficientNet-B0 w/ top-1 > 76.9
...
* also add pooling details to default cfg for efficiennets so testtimepool wrapper works
6 years ago
Ross Wightman
9bcd65181b
Add exponential moving average for model weights + few other additions and cleanup
...
* ModelEma class added to track an EMA set of weights for the model being trained
* EMA handling added to train, validation and clean_checkpoint scripts
* Add multi checkpoint or multi-model validation support to validate.py
* Add syncbn option (APEX) to train script for experimentation
* Cleanup interface of CheckpointSaver while adding ema functionality
6 years ago
Ross Wightman
105d5702d7
Update README.md
6 years ago
Ross Wightman
6688d0669f
PyTorch trained MobileNetV3 weights that match/best paper. Update fbnetc weights to improved copy.
6 years ago
Ross Wightman
89e77ec049
Merge pull request #7 from zhunzhong07/patch-2
...
Correct name
6 years ago
Zhun Zhong
b32361c9f8
Correct name
6 years ago
Ross Wightman
ff99625603
Missed a few models in the model/_all_ list for senet
6 years ago
Ross Wightman
99122aac1c
Replace ResNet-34 default weights with a great result from my experiments.
6 years ago
Ross Wightman
48ab3cf070
Add Apache LICENSE file
6 years ago
Ross Wightman
4bb5e9b224
Ported Tensorflow pretrained EfficientNet weights and some model cleanup
...
* B0-B3 weights ported from TF with close to paper accuracy
* Renamed gen_mobilenet to gen_efficientnet since scaling params go well beyond 'mobile' specific
* Add Tensorflow preprocessing option for closer images to source repo
6 years ago
Ross Wightman
4efecfdc47
Add drop_connect impl to try during training, fix a few comments
6 years ago
Ross Wightman
0fc4cca2ff
Cleanup unused member with old name
6 years ago
Ross Wightman
7a6d61566e
Add EfficientNet impl, change existing depth_multipler -> channel_multiplier as definitions have been confused
6 years ago
Ross Wightman
6bff9c75dc
Cleanup model_factory imports, consistent __all__ for models, fixed inception_v4 weight url
6 years ago
Ross Wightman
e6c14427c0
More appropriate/correct loss name
6 years ago
Ross Wightman
99ab1b1276
Update README.md
6 years ago
Ross Wightman
c2de3a922b
Update README.md
6 years ago
Ross Wightman
337702d51b
Merge pull request #2 from zhunzhong07/patch-1
...
Fix bug for prefetcher
6 years ago
Zhun Zhong
127487369f
Fix bug for prefetcher
...
Set input and target to Cuda when without using prefetcher.
6 years ago
Ross Wightman
e8cf619005
Update gluon_resnet.py
...
Update header comment
6 years ago
Ross Wightman
7419e9835f
Add MxNet Gluon ResNet variants w/ converted pretrained weights. Very well trained set of models.
6 years ago
Ross Wightman
2da0b4dbc1
Add inception_v3 models via torchvision, 4 different pretrained weight choices
6 years ago
Ross Wightman
af1a68d2e1
Update README.md
6 years ago
Ross Wightman
8ceceef889
Densenet should default to bicubic interpolation, update model links to 'cpu fix' from long ago
6 years ago
Ross Wightman
a4516fe0fb
Add pretrained weights for FBNet-C
6 years ago
Ross Wightman
a72e9b67ee
Update README.md
6 years ago
Ross Wightman
4d2056722a
Mixup and prefetcher improvements
...
* Do mixup in custom collate fn if prefetcher enabled, reduces performance impact
* Move mixup code to own file
* Add arg to disable prefetcher
* Fix no cuda transfer when prefetcher off
* Random erasing when prefetcher off wasn't changed to match new args, fixed
* Default random erasing to off (prob = 0.) for train
6 years ago
Ross Wightman
780c0a96a4
Change args for RandomErasing so only one required for pixel/color mode
6 years ago
Ross Wightman
76539d905e
Some transform/data/loader refactoring, hopefully didn't break things
...
* factor out data related constants to own file
* move data related config helpers to own file
* add a variant of RandomResizeCrop that randomizes interpolation method
* remove old Numpy version of RandomErasing
* cleanup torch version of RandomErasing and use it in either GPU loader batch mode or single image cpu Transform
6 years ago
Ross Wightman
e3377b0409
Add some info to README
6 years ago
Ross Wightman
db8ad25a23
MobileNetV3 appears correct based on paper update, cleaned up comments and compacted last block def
6 years ago
Ross Wightman
20d66beead
Move RMSpropTF another step closer to Tensorflow impl
...
* init square_avg with one instead of zero as per TF
* match TF order of ops for square_avg accumulation
* move LR scaling to momentum buffer accumulator as per TF
* add decoupled weight decay flag (not in TF)
6 years ago
Ross Wightman
89147a91e6
Remove adabound optimizer, never got it working well on larger datasets
6 years ago