Commits

Branch selector

 1331c14

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 23, 2022

• Add train benchmark results, adjust name scheme for inference and train benchmark files.

  rwightman

  committed
  1331c14

  Copy full SHA for 1331c14

Commits on Jan 21, 2022

• Merge pull request huggingface#1105 from kozistr/refactor/remove-condition

  Show description for a517bf6
  rwightman

  authored
  a517bf6

  Copy full SHA for a517bf6

• refactor: remove if-condition

  kozistr

  committed
  56a6b38

  Copy full SHA for 56a6b38

• version 0.5.5

  rwightman

  committed
  4476776

  Copy full SHA for 4476776

Commits on Jan 20, 2022

• Add update NCHW and NHWC inference benchmark numbers for current models. Flip name of 'sam' vit models in results files

  rwightman

  committed
  499c474

  Copy full SHA for 499c474

Commits on Jan 19, 2022

• Last batch of small model weights (for now). mobilenetv3_small 050/075/100 and updated mnasnet_small with lambc/lamb optimizer.

  rwightman

  committed
  83b40c5

  Copy full SHA for 83b40c5

• Merge pull request huggingface#1094 from Mi-Peng/lars

  Show description for 7f73252
  rwightman

  authored
  7f73252

  Copy full SHA for 7f73252

• fix lars

  

  Mi-Peng

  committed
  cdcd0a9

  Copy full SHA for cdcd0a9

• Update results csvs for latest release

  rwightman

  committed
  2d4b7e7

  Copy full SHA for 2d4b7e7

• Add AvgPool2d anti-aliasing support to ResNet arch (as per OpenAI CLIP models), add a few blur aa models as well

  rwightman

  committed
  1aa617c

  Copy full SHA for 1aa617c

Commits on Jan 17, 2022

• Add --fuser arg to train/validate/benchmark scripts to select jit fuser type

  rwightman

  committed
  f0f9ecc

  Copy full SHA for f0f9ecc

• Add Dino pretrained weights (no head) for vit models. Add support to tests and helpers for models w/ no classifier (num_classes=0 in pretrained cfg)

  rwightman

  committed
  010b486

  Copy full SHA for 010b486

• unbiased=False for torch.var_mean path of ConvNeXt LN. Fix huggingface#1090

  rwightman

  committed
  738a9cd

  Copy full SHA for 738a9cd

Commits on Jan 16, 2022

• Default conv_mlp to False across the board for ConvNeXt, causing issues on more setups than it's improving right now...

  rwightman

  committed
  e0c4eec

  Copy full SHA for e0c4eec

Commits on Jan 15, 2022

• Add ConvNeXt 22k->1k fine-tuned and 384 22k-1k fine-tuned weights after testing

  rwightman

  committed
  b669f4a

  Copy full SHA for b669f4a

• Update README.md

  rwightman

  authored
  6dcbaf2

  Copy full SHA for 6dcbaf2

• Giant/gigantic vits snuck through in a test a broke GitHub test runner, add filter

  rwightman

  committed
  a8d103e

  Copy full SHA for a8d103e

• Extra vit_huge model likely to cause test issue (non in21k variant), adding to filters

  rwightman

  committed
  ef72ad4

  Copy full SHA for ef72ad4

• Update REAMDE.md. Sneak in g/G (giant / gigantic?) ViT defs from scaling paper

  rwightman

  committed
  e967c72

  Copy full SHA for e967c72

• Add some more small model weights lcnet, mnas, mnv2

  rwightman

  committed
  9ca3437

  Copy full SHA for 9ca3437

• Version 0.5.4

  rwightman

  committed
  fa6463c

  Copy full SHA for fa6463c

• Fix stem width for really small mobilenetv3 arch defs

  rwightman

  committed
  fa81164

  Copy full SHA for fa81164

• Add missing dropout for head reset in ConvNeXt default head

  rwightman

  committed
  edd3d73

  Copy full SHA for edd3d73

• Some convnext cleanup, remove in place mul_ for gamma, breaking symbolic trace, cleanup head a bit...

  rwightman

  committed
  b093dcb

  Copy full SHA for b093dcb

• Add initial ConvNeXt impl (mods of official code)

  rwightman

  committed
  18934de

  Copy full SHA for 18934de

• Fix MobileNetV2 head conv size for multiplier < 1.0. Add some missing modification copyrights, fix starting date of some old ones.

  rwightman

  committed
  656757d

  Copy full SHA for 656757d

Commits on Jan 8, 2022

• Fix out_indices handling breakage, should have left as per vgg approach.

  rwightman

  committed
  ccfeb06

  Copy full SHA for ccfeb06

Commits on Jan 7, 2022

• Fix huggingface#1078, DarkNet has 6 feature maps. Make vgg and darknet out_indices handling/comments equivalent

  rwightman

  committed
  a9f9148

  Copy full SHA for a9f9148

• visformer supports spatial feat map, update pool_size in pretrained cfg to match

  rwightman

  committed
  c21b216

  Copy full SHA for c21b216

• Fix fbnetv3 pretrained cfg changes

  rwightman

  committed
  9c11dfd

  Copy full SHA for 9c11dfd

• FBNetV3 timm trained weights added for b/d/g variants. Update version to 0.5.2 for pypi release.

  rwightman

  committed
  1406cdd

  Copy full SHA for 1406cdd

• Leaving repeat aug sampler indices as tensor thrashes worker shared process memory

  rwightman

  committed
  02ae11e

  Copy full SHA for 02ae11e

• Add lcnet_100 and mnasnet_small weights

  rwightman

  committed
  4df51f3

  Copy full SHA for 4df51f3

• Remove deprecated bn-tf train arg and create_model handler. Add evos/evob models back into fx test filter until norm_norm_norm branch merged.

  rwightman

  committed
  5ccf682

  Copy full SHA for 5ccf682

• Add more small model defs for MobileNetV3/V2/LCNet

  rwightman

  committed
  b9a715c

  Copy full SHA for b9a715c

Pagination

Previous

Next