Code

Major update #25

Code

Open

franckma31 wants to merge 98 commits intomasterdeel-ai/deel-torchlip:masterfrom developdeel-ai/deel-torchlip:developCopy head branch name to clipboard

+10,648-2,223

Commits

Commits on Jun 14, 2022

• chore: update versions of pre-commit hooks

  Show description for ca39a7f
  cofri

  committed
  ca39a7f

  Copy full SHA for ca39a7f

• fix: align PyTorch version in tox with requirements.txt

  cofri

  committed
  74dcef2

  Copy full SHA for 74dcef2

• chore: remove NumPy fixed version in requirements

  Show description for 9bd9956
  cofri

  committed
  9bd9956

  Copy full SHA for 9bd9956

Commits on Jun 15, 2022

• Merge pull request #13 from deel-ai/chore/remove_numpy_version

  Show description for e8c11cc
  cofri

  authored
  e8c11cc

  Copy full SHA for e8c11cc

Commits on Jun 17, 2022

• feat: new "disjoint_neurons" parameter in FrobeniusLinear

  Show description for 7b519b9
  cofri

  committed
  7b519b9

  Copy full SHA for 7b519b9

Commits on Jun 20, 2022

• Merge pull request #14 from deel-ai/feat/frobenius_disjoint_neurons

  Show description for 29de308
  cofri

  authored
  29de308

  Copy full SHA for 29de308

Commits on Jun 21, 2022

• keep the layer names while exporting

  

  Florian

  committed
  0de161e

  Copy full SHA for 0de161e

Commits on Jun 23, 2022

• Merge pull request #16 from HeinrichAD/feature/keep-layer-names-flag

  Show description for 3886890
  cofri

  authored
  3886890

  Copy full SHA for 3886890

Commits on Nov 29, 2023

• feat: add parametrization

  y-prudent

  committed
  55fd48c

  Copy full SHA for 55fd48c

• chore: update py version in workflows (3.6 deprec)

  y-prudent

  committed
  80a75a0

  Copy full SHA for 80a75a0

Commits on Nov 30, 2023

• feat: add in-place vanilla model conversion

  y-prudent

  committed
  5ab989b

  Copy full SHA for 5ab989b

Commits on Jan 15, 2024

• fix: asymptotic coef value for lconv

  Show description for effd7e6
  y-prudent

  committed
  effd7e6

  Copy full SHA for effd7e6

Commits on Feb 5, 2024

• chore: vanilla_model added in deel/torchlip/__init__.py

  y-prudent

  committed
  87f78e9

  Copy full SHA for 87f78e9

Commits on Apr 15, 2024

• remove parametrization for lipschitz param when coef_lip==1.0

  

  Franck Mamalet

  committed
  b8a0b3d

  Copy full SHA for b8a0b3d

• add support for softHKR loss: SoftHKRMulticlassLoss warning alpha\in[0.1]

  

  Franck Mamalet

  committed
  c75092b

  Copy full SHA for c75092b

• add reduce mean for softHKR loss

  

  Franck Mamalet

  committed
  432f706

  Copy full SHA for 432f706

Commits on Jun 5, 2024

• detach to avoid graph expansion

  

  Franck Mamalet

  committed
  4583add

  Copy full SHA for 4583add

Commits on Oct 8, 2024

• test uniformization using pytest

  

  Franck Mamalet

  committed
  f96fd0c

  Copy full SHA for f96fd0c

• update using uft

  

  Franck Mamalet

  committed
  ae90761

  Copy full SHA for ae90761

• update using uft

  

  Franck Mamalet

  committed
  4325e43

  Copy full SHA for 4325e43

• linter

  

  Franck Mamalet

  committed
  554e499

  Copy full SHA for 554e499

• flake8 errors

  

  Franck Mamalet

  committed
  37a54f5

  Copy full SHA for 37a54f5

• flake8 errors test_metrics

  

  Franck Mamalet

  committed
  8a05fc7

  Copy full SHA for 8a05fc7

Commits on Oct 9, 2024

• seed on test normalizers

  

  Franck Mamalet

  committed
  016b9e1

  Copy full SHA for 016b9e1

• linter on test_layers

  

  Franck Mamalet

  committed
  64aa108

  Copy full SHA for 64aa108

• linter on test_layers

  

  Franck Mamalet

  committed
  7918a2b

  Copy full SHA for 7918a2b

• avoid F401 linter error

  

  Franck Mamalet

  committed
  f1688dc

  Copy full SHA for f1688dc

• test normalizer all close value

  

  Franck Mamalet

  committed
  576de52

  Copy full SHA for 576de52

• F401 linter

  

  Franck Mamalet

  committed
  496fe04

  Copy full SHA for 496fe04

• update kr_loss for nan support, and modify hkr losses to use 0<= alpha <= 1.0

  

  Franck Mamalet

  committed
  929bed4

  Copy full SHA for 929bed4

• add warning when alpha > 1. in SoftHKRMulticlassLoss

  

  Franck Mamalet

  committed
  a361056

  Copy full SHA for a361056

• update losses to support any target (target>0 for true value), alpha in [0,1], and min_margin/2.0

  

  Franck Mamalet

  committed
  4007e77

  Copy full SHA for 4007e77

• add vanilla export to InvertibleUpsampling class

  

  Franck Mamalet

  committed
  fbedaac

  Copy full SHA for fbedaac

• linter loss.py

  

  Franck Mamalet

  committed
  f4c38c2

  Copy full SHA for f4c38c2

• add MultiMarginLoss test based on pytorch implementation (warning no hot encoding)

  

  Franck Mamalet

  committed
  20272e2

  Copy full SHA for 20272e2

• add support for reduction in binary losses

  

  Franck Mamalet

  committed
  15f8ffe

  Copy full SHA for 15f8ffe

• add support for reduction in multiclass losses

  

  Franck Mamalet

  committed
  2702d20

  Copy full SHA for 2702d20

• add support for reduction in softHKR loss

  

  Franck Mamalet

  committed
  8920cb2

  Copy full SHA for 8920cb2

Commits on Oct 11, 2024

• add supported lipschitz layer test in Sequential

  

  Franck Mamalet

  committed
  d68aee1

  Copy full SHA for d68aee1

• add support for Tau Cross ENtropy loasses and tests

  

  Franck Mamalet

  committed
  c24ccf8

  Copy full SHA for c24ccf8

• add support for molti gpu in binary losses

  

  Franck Mamalet

  committed
  115bd95

  Copy full SHA for 115bd95

• linters

  

  Franck Mamalet

  committed
  efc8ca9

  Copy full SHA for efc8ca9

• add support multi_gpu for all KR losses

  

  Franck Mamalet

  committed
  7c5cfc4

  Copy full SHA for 7c5cfc4

• linters

  

  Franck Mamalet

  committed
  0a9b9a4

  Copy full SHA for 0a9b9a4

Commits on Oct 13, 2024

• paranthesis tricks bjork

  

  Franck Mamalet

  committed
  3c6bece

  Copy full SHA for 3c6bece

Commits on Oct 14, 2024

• switch from n_iter to eps stopping criteria for Bjorck and spectral norm

  

  Franck Mamalet

  committed
  dbc5c1a

  Copy full SHA for dbc5c1a

• switch from n_iter to eps stopping criteria + add spectral normaization in bjorck_ initializer

  

  Franck Mamalet

  committed
  a83eb67

  Copy full SHA for a83eb67

• linters

  

  Franck Mamalet

  committed
  130e40a

  Copy full SHA for 130e40a

• update setup.cfg to move to python 39,310,311 and pt{1.10.2,1.13.1,2.1.2,2.3.1,2.4.1}

  

  Franck Mamalet

  committed
  96333bd

  Copy full SHA for 96333bd

• linters corrections

  

  Franck Mamalet

  committed
  4c8d32b

  Copy full SHA for 4c8d32b

• linters corrections

  

  Franck Mamalet

  committed
  9a21dcc

  Copy full SHA for 9a21dcc

• Update python-lints.yml

  franckma31

  authored
  13cdb98

  Copy full SHA for 13cdb98

• Update python-tests.yml

  franckma31

  authored
  5e0d3ba

  Copy full SHA for 5e0d3ba

• update workflow github

  

  Franck Mamalet

  committed
  a2729a5

  Copy full SHA for a2729a5

• update workflow github

  

  Franck Mamalet

  committed
  9696ab4

  Copy full SHA for 9696ab4

• update requirements and pytorch version

  

  Franck Mamalet

  committed
  2cfa164

  Copy full SHA for 2cfa164

• update version with file + cleaning

  

  Franck Mamalet

  committed
  d7481a2

  Copy full SHA for d7481a2

• Merge pull request #20 from deel-ai/feat-parametrization-for-normalization

  Show description for 49b1a50
  franckma31

  authored
  49b1a50

  Copy full SHA for 49b1a50

Commits on Nov 7, 2024

• update docstring in losses

  franckma31

  committed
  9658d27

  Copy full SHA for 9658d27

• clean test_condense

  franckma31

  committed
  26fd0e3

  Copy full SHA for 26fd0e3

• remove warning on eps_spectral

  franckma31

  committed
  e83741e

  Copy full SHA for e83741e

• remove reference to Keras in test files

  franckma31

  committed
  f5acef3

  Copy full SHA for f5acef3

Commits on Nov 13, 2024

• replace Reshape layer by torch.nn.UnShuffle for tests

  franckma31

  committed
  9c2d4a1

  Copy full SHA for 9c2d4a1

• updated atol testing value due to numerical imprecision on random initialization

  franckma31

  committed
  014764d

  Copy full SHA for 014764d

Commits on Nov 27, 2024

• update docs notebooks with new losses

  franckma31

  committed
  6d7d3a1

  Copy full SHA for 6d7d3a1

• Merge pull request #22 from deel-ai/feat/update_normalizers

  Show description for 4360edf
  cofri

  authored
  4360edf

  Copy full SHA for 4360edf

• update groupsort and groupsort2 activ to be consistent with deellip (there were a bug when processing images)

  

  Franck Mamalet

  authored and

  franckma31

  committed
  73467a2

  Copy full SHA for 73467a2

• change to_NCHW to process numpy arrays

  

  Franck Mamalet

  authored and

  franckma31

  committed
  c8bdf64

  Copy full SHA for c8bdf64

• Add HouseHolder activation and adapt pytest

  

  Franck Mamalet

  authored and

  franckma31

  committed
  c38ebb6

  Copy full SHA for c38ebb6

• update shape information in test_activation

  

  Franck Mamalet

  authored and

  franckma31

  committed
  fc0a3ba

  Copy full SHA for fc0a3ba

• add ScaledGlobalL2NormPool2d class

  

  Franck Mamalet

  authored and

  franckma31

  committed
  d0820eb

  Copy full SHA for d0820eb

• add SpectralConvTranspose2d layer and update test for tf compatibility

  Show description for 4eba8f2
  

  Franck Mamalet

  authored and

  franckma31

  committed
  4eba8f2

  Copy full SHA for 4eba8f2

• cleaning

  

  Franck Mamalet

  authored and

  franckma31

  committed
  30922f9

  Copy full SHA for 30922f9

• layer and batch centering

  

  Mathieu SERRURIER

  authored and

  franckma31

  committed
  01d8b4a

  Copy full SHA for 01d8b4a

• update layer and batch centering + pytests

  franckma31

  committed
  defbcf6

  Copy full SHA for defbcf6

• simplify householder reshape computation + correct test

  franckma31

  committed
  ad4fc38

  Copy full SHA for ad4fc38

• clean normalization

  franckma31

  committed
  7305a91

  Copy full SHA for 7305a91

• add description comments on added layers

  franckma31

  committed
  493156a

  Copy full SHA for 493156a

• spectral conv 1d

  

  Mathieu SERRURIER

  authored and

  franckma31

  committed
  184bdd4

  Copy full SHA for 184bdd4

• add support for spectralConv1d

  franckma31

  committed
  60961af

  Copy full SHA for 60961af

• add padconv support + tests

  franckma31

  committed
  4ac8ff6

  Copy full SHA for 4ac8ff6

• modify InvertibleUpDownSampling t o use torch PixelShuffle and Unshuffle; modify pytest (warning only support 2D inputs, and single value kernel size)

  franckma31

  committed
  c8ca131

  Copy full SHA for c8ca131

• add a LipResidual layer and tests

  franckma31

  committed
  7227db4

  Copy full SHA for 7227db4

• add support for replicate padding + trick to change symmetric+1 padding in replicate

  franckma31

  committed
  d19ad7d

  Copy full SHA for d19ad7d

• clean and linter

  franckma31

  committed
  ccdc261

  Copy full SHA for ccdc261

Commits on Dec 2, 2024

• update for compatibility with tensorflow tests

  franckma31

  committed
  1fd81f4

  Copy full SHA for 1fd81f4

• updates on docs et notebook + resolve issue #17

  franckma31

  committed
  58d8fba

  Copy full SHA for 58d8fba

Commits on Dec 9, 2024

• update vanilla export with PixelUnshuffle variable in InvertibleUp(Down)Sampling

  franckma31

  committed
  c6bf80f

  Copy full SHA for c6bf80f

• modify vanilla_model functio to support parametrization

  franckma31

  committed
  16b0059

  Copy full SHA for 16b0059

• update to solve comment #23 (comment)

  franckma31

  committed
  34105b9

  Copy full SHA for 34105b9

• use LPPool2d in ScaledL2NormPool2d + pooling tests

  franckma31

  committed
  f5e28ce

  Copy full SHA for f5e28ce

• use lp_lppool2d in ScaledAdaptativeL2NormPool2d ( #23 (comment)) + name change Global to Adaptative

  franckma31

  committed
  a8721bb

  Copy full SHA for a8721bb

Commits on Dec 10, 2024

• remove same case for padding_mode to solve #23 (comment) and #23 (comment)

  franckma31

  committed
  76a6e69

  Copy full SHA for 76a6e69

Commits on Jan 14, 2025

• Merge pull request #23 from deel-ai/feat/new_layers

  Show description for b23e090
  thib-s

  authored
  b23e090

  Copy full SHA for b23e090

Commits on Jan 24, 2025

• change version to 0.1.1->1.0

  franckma31

  committed
  baa4ee2

  Copy full SHA for baa4ee2

• Merge pull request #24 from deel-ai/feat/new_layers

  Show description for 4975e50
  cofri

  authored
  4975e50

  Copy full SHA for 4975e50

Commits on Feb 3, 2025

• deprecated spectral_ and bjorck_ initialization

  franckma31

  committed
  892a8fe

  Copy full SHA for 892a8fe

Commits on Feb 10, 2025

• Merge pull request #26 from deel-ai/feat/new_layers

  Show description for 3308dd7
  cofri

  authored
  3308dd7

  Copy full SHA for 3308dd7