Code

Fix torch compile, script, export#1031

Code

Merged

qubvel merged 58 commits intomainqubvel-org/segmentation_models.pytorch:mainfrom torch-compile-exportqubvel-org/segmentation_models.pytorch:torch-compile-exportCopy head branch name to clipboardJan 15, 2025

+2,418-2,289

Commits

Commits on Jan 12, 2025

• Move tests

  qubvel

  committed
  643c0b6

  Copy full SHA for 643c0b6

• Add compile test for encoders (to be optimized)

  qubvel

  committed
  7a937ab

  Copy full SHA for 7a937ab

• densnet

  qubvel

  committed
  9a7c768

  Copy full SHA for 9a7c768

• dpn

  qubvel

  committed
  34b8533

  Copy full SHA for 34b8533

• efficientnet

  qubvel

  committed
  a3618fa

  Copy full SHA for a3618fa

• inceptionresnetv2

  qubvel

  committed
  e3f6c70

  Copy full SHA for e3f6c70

• inceptionv4

  qubvel

  committed
  20b28be

  Copy full SHA for 20b28be

• mix-transformer

  qubvel

  committed
  d996165

  Copy full SHA for d996165

• mobilenet

  qubvel

  committed
  9e38154

  Copy full SHA for 9e38154

• mobileone

  qubvel

  committed
  c6e5d53

  Copy full SHA for c6e5d53

• resnet

  qubvel

  committed
  5a76722

  Copy full SHA for 5a76722

• senet

  qubvel

  committed
  36d056b

  Copy full SHA for 36d056b

• vgg

  qubvel

  committed
  aefcfd4

  Copy full SHA for aefcfd4

• xception

  qubvel

  committed
  e9628bf

  Copy full SHA for e9628bf

• Deprecate timm- encoders, remap to tu- most of them

  qubvel

  committed
  70262e5

  Copy full SHA for 70262e5

• Add tiny encoders and compile mark

  qubvel

  committed
  0b0b1c4

  Copy full SHA for 0b0b1c4

• Add conftest

  qubvel

  committed
  4c11682

  Copy full SHA for 4c11682

• Fix features

  qubvel

  committed
  70168b4

  Copy full SHA for 70168b4

Commits on Jan 13, 2025

• Merge branch 'main' into torch-compile-export

  qubvel

  committed
  8aed7ef

  Copy full SHA for 8aed7ef

• Add triggering compile tests on diff

  qubvel

  committed
  50c40d1

  Copy full SHA for 50c40d1

• Remove marks

  qubvel

  committed
  0764d5e

  Copy full SHA for 0764d5e

• Add test_compile stage to CI

  qubvel

  committed
  7cab4be

  Copy full SHA for 7cab4be

• Update requirements

  qubvel

  committed
  2622e0e

  Copy full SHA for 2622e0e

• Update makefile

  qubvel

  committed
  e12ee8d

  Copy full SHA for e12ee8d

• Update get_stages

  qubvel

  committed
  da0cd19

  Copy full SHA for da0cd19

• Fix weight loading for deprecate encoders

  qubvel

  committed
  7752969

  Copy full SHA for 7752969

• Fix weight loading for mobilenetv3

  qubvel

  committed
  409b820

  Copy full SHA for 409b820

• Format

  qubvel

  committed
  ae3cb8a

  Copy full SHA for ae3cb8a

• Add compile test for models

  qubvel

  committed
  ff278c9

  Copy full SHA for ff278c9

• Add torch.export test

  qubvel

  committed
  a806147

  Copy full SHA for a806147

• Disable export tests for dpn and inceptionv4

  qubvel

  committed
  aa5b088

  Copy full SHA for aa5b088

• Disable export for timm-eff-net

  qubvel

  committed
  df2f484

  Copy full SHA for df2f484

Commits on Jan 14, 2025

• Huge fix for torch scripting (except Unet++ and UperNet)

  qubvel

  committed
  7157501

  Copy full SHA for 7157501

• Fix scripting

  qubvel

  committed
  257da0b

  Copy full SHA for 257da0b

• Add test for torch script

  qubvel

  committed
  d4d4cf6

  Copy full SHA for d4d4cf6

• Add torch_script test to CI

  qubvel

  committed
  3cb8198

  Copy full SHA for 3cb8198

• Fix

  qubvel

  committed
  4f65d8f

  Copy full SHA for 4f65d8f

• Fix timm-effnet encoders

  qubvel

  committed
  70776ea

  Copy full SHA for 70776ea

• Make from_pretrained strict by default

  qubvel

  committed
  31bee79

  Copy full SHA for 31bee79

• Fix DeepLabV3 BC

  qubvel

  committed
  556b3aa

  Copy full SHA for 556b3aa

• Fix scripting for encoders

  qubvel

  committed
  f70d861

  Copy full SHA for f70d861

• Refactor test do not skip

  qubvel

  committed
  ead24b4

  Copy full SHA for ead24b4

• Fix encoders (mobilenet, inceptionv4)

  qubvel

  committed
  d44509a

  Copy full SHA for d44509a

• Update encoders table

  qubvel

  committed
  b2c13f1

  Copy full SHA for b2c13f1

• Fix export test

  qubvel

  committed
  73809e3

  Copy full SHA for 73809e3

• Fix docs

  qubvel

  committed
  bc1319e

  Copy full SHA for bc1319e

• Update warning

  qubvel

  committed
  d25dd47

  Copy full SHA for d25dd47

• Move pretrained settings

  qubvel

  committed
  4f3b37e

  Copy full SHA for 4f3b37e

• Add BC for timm- encoders

  qubvel

  committed
  06199b0

  Copy full SHA for 06199b0

• Fixing table

  qubvel

  committed
  51e0a67

  Copy full SHA for 51e0a67

• Update compile test

  qubvel

  committed
  524bcae

  Copy full SHA for 524bcae

Commits on Jan 15, 2025

• Change compile backend to eager

  qubvel

  committed
  a2b97d8

  Copy full SHA for a2b97d8

• Update docs

  qubvel

  committed
  17a4b70

  Copy full SHA for 17a4b70

• Fixup

  qubvel

  committed
  20564f2

  Copy full SHA for 20564f2

• Fix batchnorm typo

  qubvel

  committed
  5bbb1db

  Copy full SHA for 5bbb1db

• Add depth validation

  qubvel

  committed
  d121fec

  Copy full SHA for d121fec

• Update segmentation_models_pytorch/encoders/__init__.py

  Show description for 7bb9d37
  

  qubvel

  and

  adamjstewart

  authored
  7bb9d37

  Copy full SHA for 7bb9d37

• Style

  qubvel

  committed
  da24de9

  Copy full SHA for da24de9