Commit Graph

752 Commits (e0c4eec4b66dc14ae96097c7b4a7ef2af45ba309)

Author SHA1 Message Date
Alexander Soare b25ff96768 wip - pre-rebase
3 years ago
Alexander Soare e051dce354 Make all models FX traceable
3 years ago
Alexander Soare cf4561ca72 Add FX based FeatureGraphNet capability
3 years ago
Alexander Soare 0149ec30d7 wip - attempting to rebase
3 years ago
Alexander Soare 02c3a75a45 wip - make it possible to use fx graph in train and eval mode
3 years ago
Alexander Soare bc3d4eb403 wip -rebase
3 years ago
Alexander Soare ab3ac3f25b Add FX based FeatureGraphNet capability
3 years ago
Ross Wightman 9ec3210c2d More TFDS parser cleanup, support improved TFDS even_split impl (on tfds-nightly only currently).
3 years ago
Ross Wightman ba65dfe2c6 Dataset work
3 years ago
Ross Wightman ddc29da974 Add ResNet101 and ResNet152 weights from higher aug RSB recipes. 81.93 and 82.82 top-1 at 224x224.
3 years ago
Ross Wightman b328e56f49 Update eca_halonext26ts weights to a better set
3 years ago
Ross Wightman 2ddef942b9 Better fix for #954 that doesn't break torchscript, pull torch._assert into timm namespace when it exists
3 years ago
Ross Wightman 4f0f9cb348 Fix #954 by bringing traceable _assert into timm to allow compat w/ PyTorch < 1.8
3 years ago
Ross Wightman a41de1f666 Add interpolation mode handling to transforms. Removes InterpolationMode warning. Works for torchvision versions w/ and w/o InterpolationMode enum. Fix #738.
3 years ago
Ross Wightman ed41d32637 Add repr to auto_augment and random_erasing impl
3 years ago
Ross Wightman ae72d009fa Add weights for lambda_resnet50ts, halo2botnet50ts, lamhalobotnet50ts, updated halonet50ts
3 years ago
Ross Wightman b745d30a3e Fix formatting of last commit
3 years ago
Ross Wightman 3478f1d7f1 Traceability fix for vit models for some experiments
3 years ago
Ross Wightman f658a72e72 Cleanup re-use of Dropout modules in Mlp modules after some twitter feedback :p
3 years ago
Thomas Viehmann f805ba86d9 use .unbind instead of explicitly listing the indices
3 years ago
Ross Wightman 57992509f9 Fix some formatting in utils/model.py
3 years ago
Ross Wightman 0fe4fd3f1f add d8 and e8 regnetz models with group size 8
3 years ago
Ross Wightman 25e7c8c5e5 Update broken resnetv2_50 weight url, add resnetv1_101 a1h recipe weights for 224x224 train
3 years ago
Ross Wightman b6caa356d2 Fixed eca_botnext26ts_256 weights added, 79.27
3 years ago
Ross Wightman c02334d9fa Add weights for regnetz_d and haloregnetz_c, update regnetz_c weights. Add commented PyTorch XLA code for halo attention
3 years ago
Ross Wightman 02daf2ab94 Add option to include relative pos embedding in the attention scaling as per references. See discussion #912
3 years ago
masafumi 047a5ec05f Fix bugs that Mixup does not work device=cpu
3 years ago
Ross Wightman cd34913278 Remove some outdated comments, botnet networks working great now.
3 years ago
Ross Wightman 6ed4cdccca Update lambda_resnet26t weights with better set
3 years ago
ICLR Author 44d6d51668 Add ConvMixer
3 years ago
Ross Wightman a85df34993 Update lambda_resnet26rpt weights to 78.9, add better halonet26t weights at 79.1 with tweak to attention dim
3 years ago
Ross Wightman b544ad4d3f regnetz model default cfg tweaks
3 years ago
Ross Wightman e5da481073 Small post-merge tweak for freeze/unfreeze, add to __init__ for utils
3 years ago
Ross Wightman 5ca72dcc75 Merge branch 'freeze-functionality' of https://github.com/alexander-soare/pytorch-image-models into alexander-soare-freeze-functionality
3 years ago
Ross Wightman e2b8d44ff0 Halo, bottleneck attn, lambda layer additions and cleanup along w/ experimental model defs
3 years ago
Alexander Soare 431e60c83f Add acknowledgements for freeze_batch_norm inspiration
3 years ago
Ross Wightman fbf59c04ee Change crop ratio on correct resnet50 variant.
3 years ago
Ross Wightman ae1ff5792f Clean a1/a2/3 rsb _0 checkpoints properly, fix v2 loading.
3 years ago
Ross Wightman 93901e992f Version bump to 0.5.0 for pending release post RSB and ATTN updates
3 years ago
Ross Wightman da0d39bedd Update default crop_pct for byoanet
3 years ago
Ross Wightman cc9bedf373 Add initial ResNet Strikes Back weights for ResNet50 and ResNetV2-50 models
3 years ago
Ross Wightman 64495505b7 Add updated lambda resnet26 and botnet26 checkpoints with fixes applied
3 years ago
Ross Wightman b2094f4ee8 support bits checkpoints in avg/load
3 years ago
Ross Wightman 007bc39323 Some halo and bottleneck attn code cleanup, add halonet50ts weights, use optimal crop ratios
3 years ago
Alexander Soare 65c3d78b96 Freeze unfreeze functionality finalized. Tests added
3 years ago
Alexander Soare 0cb8ea432c wip
3 years ago
Ross Wightman b1c2e3eb92 Match rel_pos_indices attr rename in conv branch
3 years ago
Ross Wightman b49630a138 Add relative pos embed option to LambdaLayer, fix last transpose/reshape.
3 years ago
Ross Wightman d657e2cc0b Remove dead code line from efficientnet
3 years ago
Ross Wightman 0ca687f224 Make 'regnetz' model experiments closer to actual RegNetZ, bottleneck expansion, expand from in_chs, no shortcut on stride 2, tweak model sizes
3 years ago
leondgarse 51eaf9360d
Remove a duplicate layer creation in byobnet.py
3 years ago
Ross Wightman b81e79aae9 Fix bottleneck attn transpose typo, hopefully these train better now..
3 years ago
Ross Wightman 80075b0b8a Add worker_seeding arg to allow selecting old vs updated data loader worker seed for (old) experiment repeatability
3 years ago
Ross Wightman 6478bcd02c Fix regnetz_d conv layer name, use inception mean/std
3 years ago
Ross Wightman 0387e6057e Update binary cross ent impl to use thresholding as an option (convert soft targets from mixup/cutmix to 0, 1)
3 years ago
Ross Wightman f8a63a3b71 Add worker_init_fn to loader for numpy seed per worker
3 years ago
Ross Wightman 515121cca1 Use reshape instead of view in std_conv, causing issues in recent PyTorch in channels_last
3 years ago
Ross Wightman da06cc61d4 ResNetV2 seems to work best without zero_init residual
3 years ago
Ross Wightman 8e11da0ce3 Add experimental RegNetZ(ish) models for training / perf trials.
3 years ago
Alexander Soare 6bbc50beb4 make it possible to provide norm_layer via create_model
3 years ago
nateraw adcb74f87f 🎨 Import load_state_dict_from_url directly
3 years ago
nateraw e65a2cba3d 🎨 cleanup and add a couple comments
3 years ago
nateraw 2b6ade24b3 🎨 write model card to enable inference
3 years ago
Ross Wightman cf5ac2800c BotNet models were still off, remove weights for bad configs. Add good SE-HaloNet33-TS weights.
3 years ago
Ross Wightman 24720abe3b Merge branch 'master' into attn_update
3 years ago
Ross Wightman 1c9284c640 Add BeiT 'finetuned' 1k weights and pretrained 22k weights, pretraining specific (masked) model excluded for now
3 years ago
Ross Wightman f8a215cfe6 A few more crossvit tweaks, fix training w/ no_weight_decay names, add crop option for scaling, adjust default crop_pct for large img size to 1.0 for better results
3 years ago
Ross Wightman 7ab2491ab7 Better handling of crossvit for tests / forward_features, fix torchscript regression in my changes
3 years ago
Ross Wightman f1808e0970 Post crossvit merge cleanup, change model names to reflect input size, cleanup img size vs scale handling, fix tests
3 years ago
Ross Wightman 4027412757 Add resnet33ts weights, update resnext26ts baseline weights
3 years ago
Richard Chen 9fe5798bee fix bug for reset classifier and fix for validating the dimension
3 years ago
Richard Chen 3718c5a5bd fix loading pretrained model
3 years ago
Richard Chen bb50b69a57 fix for torch script
3 years ago
nateraw abf9d51bc3 🚧 wip
3 years ago
Ross Wightman 5bd04714e4 Cleanup weight init for byob/byoanet and related
3 years ago
Ross Wightman 8642401e88 Swap botnet 26/50 weights/models after realizing a mistake in arch def, now figuring out why they were so low...
3 years ago
Ross Wightman 5f12de4875 Add initial AttentionPool2d that's being trialed. Fix comment and still trying to improve reliability of sgd test.
3 years ago
Ross Wightman 76881d207b Add baseline resnet26t @ 256x256 weights. Add 33ts variant of halonet with at least one halo in stage 2,3,4
3 years ago
Ross Wightman 484e61648d Adding the attn series weights, tweaking model names, comments...
3 years ago
Ross Wightman fb94350896 Update training script and loader factory to allow use of scheduler updates, repeat augment, and bce loss
3 years ago
Ross Wightman f262137ff2 Add RepeatAugSampler as per DeiT RASampler impl, showing promise for current (distributed) training experiments.
3 years ago
Ross Wightman ba9c1108a1 Add a BCE loss impl that converts dense targets to sparse /w smoothing as an alternate to CE w/ smoothing. For training experiments.
3 years ago
Ross Wightman 29a37e23ee LR scheduler update:
3 years ago
nateraw 28d2841acf 💄 apply isort
3 years ago
Ross Wightman 492c0a4e20 Update HaloAttn comment
3 years ago
nateraw e72c989973 add ability to push to hf hub
3 years ago
Richard Chen 7ab9d4555c add crossvit
3 years ago
Ross Wightman 3b9032ea48 Use Tensor.unfold().unfold() for HaloAttn, fast like as_strided but more clarity
3 years ago
Ross Wightman 78933122c9 Fix silly typo
3 years ago
Ross Wightman 2568ffc5ef Merge branch 'master' into attn_update
3 years ago
Ross Wightman 708d87a813 Fix ViT SAM weight compat as weights at URL changed to not use repr layer. Fix #825. Tweak optim test.
3 years ago
Ross Wightman 8449ba210c Improve performance of HaloAttn, change default dim calc. Some cleanup / fixes for byoanet. Rename resnet26ts to tfs to distinguish (extra fc).
3 years ago
Ross Wightman a8b65695f1 Add resnet26ts and resnext26ts models for non-attn baselines
3 years ago
Ross Wightman a5a542f17d Fix typo
3 years ago
Ross Wightman 925e102982 Update attention / self-attn based models from a series of experiments:
3 years ago
Ross Wightman d667351eac Tweak accuracy topk safety. Fix #807
3 years ago
Yohann Lereclus 35c9740826 Fix accuracy when topk > num_classes
3 years ago
Ross Wightman a16a753852 Add lamb/lars to optim init imports, remove stray comment
3 years ago
Ross Wightman c207e02782 MOAR optimizer changes. Woo!
3 years ago
Ross Wightman a426511c95 More optimizer cleanup. Change all to no longer use .data. Improve (b)float16 use with adabelief. Add XLA compatible Lars.
3 years ago