Code

Faceswap 3 beta#1392

Code

Merged

torzdf merged 136 commits intodeepfakes:fs3deepfakes/faceswap:fs3from torzdf:fs3torzdf/faceswap:fs3Copy head branch name to clipboardJun 7, 2024

+7,081-6,286

Commits

Commits on Mar 3, 2024

• Replace TF entry point code

  torzdf

  committed
  f8ca0a2

  Copy full SHA for f8ca0a2

• Update extract pipeline

  torzdf

  committed
  e3b1060

  Copy full SHA for e3b1060

• Update requirements + setup for nvidia

  torzdf

  committed
  2eebe9a

  Copy full SHA for 2eebe9a

• Remove allow-growth option

  torzdf

  committed
  e86e0ca

  Copy full SHA for e86e0ca

• tf.keras to keras updates

  torzdf

  committed
  b5808fe

  Copy full SHA for b5808fe

Commits on Mar 4, 2024

• Basic training functionality (original)

  torzdf

  committed
  6d49eaa

  Copy full SHA for 6d49eaa

Commits on Mar 14, 2024

• lib.model.losses - Port + fix all loss functions for Keras3

  torzdf

  committed
  bd1a137

  Copy full SHA for bd1a137

• lib.model - port initializers, layers. normalization to Keras3

  torzdf

  committed
  17234b9

  Copy full SHA for 17234b9

• training - Fix previews and LR_Finder

  torzdf

  committed
  288a011

  Copy full SHA for 288a011

• lib.logger - Prevent crash on custom loglevels when launching scripts external to FS

  torzdf

  committed
  5d39aa1

  Copy full SHA for 5d39aa1

Commits on Mar 17, 2024

• lib.model.autoclip to Keras 3

  torzdf

  committed
  7a939fd

  Copy full SHA for 7a939fd

• Update mixed precision layer storage

  torzdf

  committed
  d6aae0b

  Copy full SHA for d6aae0b

• model file to .keras format

  torzdf

  committed
  43d4a7b

  Copy full SHA for 43d4a7b

• typo fix

  torzdf

  committed
  233bbaa

  Copy full SHA for 233bbaa

• Set loss debug logging to trace

  torzdf

  committed
  f3c9a40

  Copy full SHA for f3c9a40

• Restructure nn_blocks to initialize layers in __init__

  torzdf

  committed
  a6ba24b

  Copy full SHA for a6ba24b

• lib.model - standardise class init logging

  torzdf

  committed
  b991ba2

  Copy full SHA for b991ba2

Commits on Mar 19, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  8f516e2

  Copy full SHA for 8f516e2

Commits on Mar 20, 2024

• Tensorboard

  Show description for 3c74a2f
  torzdf

  committed
  3c74a2f

  Copy full SHA for 3c74a2f

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  4eb4aa5

  Copy full SHA for 4eb4aa5

Commits on Mar 21, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  a037fd8

  Copy full SHA for a037fd8

Commits on Mar 22, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  b60c681

  Copy full SHA for b60c681

• Loss logging

  Show description for 8edcf3f
  torzdf

  committed
  8edcf3f

  Copy full SHA for 8edcf3f

• Backup models on total loss drop rather than per side

  torzdf

  committed
  dc99b5c

  Copy full SHA for dc99b5c

Commits on Mar 23, 2024

• Update all models to Keras3 Compat

  torzdf

  committed
  106cacf

  Copy full SHA for 106cacf

• Clean up console output

  torzdf

  committed
  02d551c

  Copy full SHA for 02d551c

• Fix learn mask and multiscale preview outputs

  torzdf

  committed
  dcbe353

  Copy full SHA for dcbe353

Commits on Mar 24, 2024

• linting

  torzdf

  committed
  2f5615b

  Copy full SHA for 2f5615b

• pylint: reduce min-public-methods to 1 + linting

  torzdf

  committed
  568462e

  Copy full SHA for 568462e

• lib.logger/utils - remove TF specific log handling + linting

  torzdf

  committed
  e4638f8

  Copy full SHA for e4638f8

Commits on Mar 25, 2024

• Remove lib.model.session

  torzdf

  committed
  d29a7df

  Copy full SHA for d29a7df

• Extract - explicitly pass batch_size to predict

  torzdf

  committed
  d58d00e

  Copy full SHA for d58d00e

• Exclude GPU for CUDA backends

  torzdf

  committed
  fb4ed01

  Copy full SHA for fb4ed01

• plugins.model - remove exclude_gpu

  torzdf

  committed
  be2e8c5

  Copy full SHA for be2e8c5

Commits on Mar 26, 2024

• Add exclude_gpus properties to ROCm and DirectML

  torzdf

  committed
  8c79afd

  Copy full SHA for 8c79afd

• Enable CPU mode for compatible extractors

  torzdf

  committed
  ed0fe85

  Copy full SHA for ed0fe85

• Fix optimizer saving

  torzdf

  committed
  3fbfef2

  Copy full SHA for 3fbfef2

Commits on Mar 27, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  41fbd36

  Copy full SHA for 41fbd36

• bug/typo fixes

  torzdf

  committed
  1c0e39c

  Copy full SHA for 1c0e39c

Commits on Mar 28, 2024

• Update clip ViT to Keras 3

  torzdf

  committed
  3724c54

  Copy full SHA for 3724c54

• lib.cli - Deprecate multi-character switches

  torzdf

  committed
  a6aeeff

  Copy full SHA for a6aeeff

Commits on Mar 29, 2024

• tools cli - deprecate multi character switches

  torzdf

  committed
  eb54229

  Copy full SHA for eb54229

• convert - remove trainer cli-arg

  torzdf

  committed
  24a5bb6

  Copy full SHA for 24a5bb6

• Remove colab references from project

  torzdf

  committed
  e00cf63

  Copy full SHA for e00cf63

• Extract: Supress Keras compile warnings for FAN + Unet

  torzdf

  committed
  6b90a1a

  Copy full SHA for 6b90a1a

Commits on Mar 30, 2024

• plugins.extract.mask.unet-dfl - Fix for Keras3/Torch backend

  torzdf

  committed
  245f707

  Copy full SHA for 245f707

Commits on Mar 31, 2024

• plugins.extract - linting + typing

  torzdf

  committed
  6f1bd1d

  Copy full SHA for 6f1bd1d

Commits on Apr 2, 2024

• Port AdaBelief to Keras 3

  torzdf

  committed
  232c21d

  Copy full SHA for 232c21d

Commits on Apr 3, 2024

• linting: Remove pylint hint from logger

  torzdf

  committed
  38a5c02

  Copy full SHA for 38a5c02

• Merge remote-tracking branch 'upstream/staging' into fs3

  torzdf

  committed
  05e8819

  Copy full SHA for 05e8819

• typofix

  torzdf

  committed
  7e73d85

  Copy full SHA for 7e73d85

• Merge remote-tracking branch 'upstream/staging' into fs3

  torzdf

  committed
  9016f7a

  Copy full SHA for 9016f7a

• Merge remote-tracking branch 'upstream/staging' into fs3

  torzdf

  committed
  dbeec7a

  Copy full SHA for dbeec7a

• Merge remote-tracking branch 'upstream/staging' into fs3

  torzdf

  committed
  8ebcbb0

  Copy full SHA for 8ebcbb0

• Merge remote-tracking branch 'upstream/staging' into fs3

  torzdf

  committed
  817ec0c

  Copy full SHA for 817ec0c

Commits on Apr 5, 2024

• Fix GUI for refactored lib.cli.args

  torzdf

  committed
  849f3e0

  Copy full SHA for 849f3e0

• Merge remote-tracking branch 'upstream/staging' into fs3

  torzdf

  committed
  50e8de3

  Copy full SHA for 50e8de3

Commits on Apr 16, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  be9817b

  Copy full SHA for be9817b

• Use saving API for load_model

  torzdf

  committed
  de09ec9

  Copy full SHA for de09ec9

• setup.py:

  Show description for 61bd910
  torzdf

  committed
  61bd910

  Copy full SHA for 61bd910

• extract: Fix keras.device to call keras.backend.device_scope

  torzdf

  committed
  2328c5e

  Copy full SHA for 2328c5e

Commits on Apr 17, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  5554a25

  Copy full SHA for 5554a25

• setup.py add more dev requirements

  torzdf

  committed
  cc5a320

  Copy full SHA for cc5a320

• training: Linting, fixes and docstrings

  Show description for 5bcd748
  torzdf

  committed
  5bcd748

  Copy full SHA for 5bcd748

Commits on Apr 18, 2024

• Bugfixes for video file alignments storage:

  Show description for e024e42
  torzdf

  committed
  e024e42

  Copy full SHA for e024e42

• Fix Keras 3 syntax

  torzdf

  committed
  3b5d3b1

  Copy full SHA for 3b5d3b1

• Fix LR Finder for Keras 3

  torzdf

  committed
  4c2494d

  Copy full SHA for 4c2494d

Commits on Apr 19, 2024

• lib: Linting, typing and minor clean up/fixes

  torzdf

  committed
  4886750

  Copy full SHA for 4886750

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  20266c9

  Copy full SHA for 20266c9

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  39aae11

  Copy full SHA for 39aae11

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  d5c0dd8

  Copy full SHA for d5c0dd8

• extract: Fix device selection for plugins

  torzdf

  committed
  ac3cda3

  Copy full SHA for ac3cda3

Commits on Apr 20, 2024

• Fix mixed precision switching for Keras 3

  torzdf

  committed
  acf6b24

  Copy full SHA for acf6b24

Commits on Apr 21, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  98fb03a

  Copy full SHA for 98fb03a

Commits on Apr 23, 2024

• Add more optimizers + open up config setting

  torzdf

  committed
  d99fc8c

  Copy full SHA for d99fc8c

Commits on Apr 25, 2024

• train: Remove updating FS1 weights to FS2 models

  torzdf

  committed
  0c68712

  Copy full SHA for 0c68712

• Alignments: Remove support for legacy .json files

  torzdf

  committed
  f266121

  Copy full SHA for f266121

• train: Fix Keras imports in model plugins

  torzdf

  committed
  c7e864a

  Copy full SHA for c7e864a

• Interim fix: Disable loading of legacy models

  torzdf

  committed
  8c09bb1

  Copy full SHA for 8c09bb1

• tools.model:

  Show description for 4c4b510
  torzdf

  committed
  4c4b510

  Copy full SHA for 4c4b510

Commits on Apr 29, 2024

• Update deprecated LeakyReLu arg

  torzdf

  committed
  6dcf34c

  Copy full SHA for 6dcf34c

Commits on May 2, 2024

• Fix inference model creation for Keras 3

  torzdf

  committed
  4632b9d

  Copy full SHA for 4632b9d

• Preview tool: Fix for Keras3

  torzdf

  committed
  a493bfe

  Copy full SHA for a493bfe

Commits on May 3, 2024

• setup.py: Configure keras backend

  Show description for 29fcf11
  torzdf

  committed
  29fcf11

  Copy full SHA for 29fcf11

• - Bump pytorch max version

  Show description for d81d206
  torzdf

  committed
  d81d206

  Copy full SHA for d81d206

• bugfix: Lion optimizer

  torzdf

  committed
  e7aec0d

  Copy full SHA for e7aec0d

Commits on May 4, 2024

• Bugfix: Select .keras file for models in GUI

  torzdf

  committed
  5cb050c

  Copy full SHA for 5cb050c

• bugfix: Model - prevent ConvAware from re-calculating on model load

  torzdf

  committed
  6fa27cb

  Copy full SHA for 6fa27cb

Commits on May 5, 2024

• train: Migration of FS2 models to FS3

  torzdf

  committed
  48e17a2

  Copy full SHA for 48e17a2

Commits on May 7, 2024

• Training: Default coverage to 100%

  torzdf

  committed
  6dbb01d

  Copy full SHA for 6dbb01d

• Bugfix: GUI Stats:

  Show description for 35188a9
  torzdf

  committed
  35188a9

  Copy full SHA for 35188a9

• bugfix: Gui stats. Reset session if log files have been replaced

  torzdf

  committed
  c2a2157

  Copy full SHA for c2a2157

• numexpr: Default to use 2/3rds of available cores

  torzdf

  committed
  cc655ae

  Copy full SHA for cc655ae

Commits on May 8, 2024

• Prevent model summaries from becoming truncated

  torzdf

  committed
  e558cd6

  Copy full SHA for e558cd6

• fix mixed precision layer storage to be compatible with ConvNext

  torzdf

  committed
  7505073

  Copy full SHA for 7505073

• Bugfix: ClipV unnest model outputs

  torzdf

  committed
  98c3e26

  Copy full SHA for 98c3e26

• Bugfix: Create keras config folder if it does not exist

  torzdf

  committed
  3607156

  Copy full SHA for 3607156

Commits on May 9, 2024

• ClipV: Don't load weights by name to supress Keras warning message

  torzdf

  committed
  00af7d5

  Copy full SHA for 00af7d5

Commits on May 10, 2024

• clip: Supress weights mismatch warning

  torzdf

  committed
  5065d2d

  Copy full SHA for 5065d2d

Commits on May 11, 2024

• Bugfix: Mean output on FFL to prevent NaNs

  torzdf

  committed
  2b3aabe

  Copy full SHA for 2b3aabe

Commits on May 14, 2024

• Remove DirectML backend

  torzdf

  committed
  fa7f979

  Copy full SHA for fa7f979

Commits on May 15, 2024

• setup: update for Linux

  torzdf

  committed
  12a485a

  Copy full SHA for 12a485a

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  349ee75

  Copy full SHA for 349ee75

Commits on May 16, 2024

• bugfix: Legacy model loading with unnested inbound nodes

  torzdf

  committed
  9bf5f03

  Copy full SHA for 9bf5f03

• Support Python up to 3.12

  torzdf

  committed
  98f5d86

  Copy full SHA for 98f5d86

• Lower max python version to 3.11

  torzdf

  committed
  0deebed

  Copy full SHA for 0deebed

Commits on May 17, 2024

• Bugfix: Porting Keras 2 models that use ClipV to Keras 3

  torzdf

  committed
  4875caf

  Copy full SHA for 4875caf

• Update setup for MacOS

  torzdf

  committed
  de5a834

  Copy full SHA for de5a834

• Allow python up to 3.12 + Fix locale deprecation

  torzdf

  committed
  d552e75

  Copy full SHA for d552e75

Commits on May 18, 2024

• GUI: Force line reading to UTF-8

  torzdf

  committed
  243b383

  Copy full SHA for 243b383

• Remove redundant Tensorflow references

  torzdf

  committed
  067862b

  Copy full SHA for 067862b

Commits on May 20, 2024

• Disable distribution + remove redumdant tf references

  torzdf

  committed
  527778b

  Copy full SHA for 527778b

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  c58a24e

  Copy full SHA for c58a24e

• pytest fixes

  torzdf

  committed
  fbc6d04

  Copy full SHA for fbc6d04

• Remove redundant code

  torzdf

  committed
  e01da8d

  Copy full SHA for e01da8d

• pytest: Set KERAS_BACKEND EnvVar

  torzdf

  committed
  16da164

  Copy full SHA for 16da164

Commits on May 21, 2024

• lib.model.normalization: bugfixes

  torzdf

  committed
  dd122e5

  Copy full SHA for dd122e5

• lib.model.loss.LaplacianPyramidLoss: Shape dtype bugfix

  torzdf

  committed
  6ee13c4

  Copy full SHA for 6ee13c4

• lib.model.layers: Minor bugfixes

  torzdf

  committed
  60eda73

  Copy full SHA for 60eda73

• scripts.conver.ConvertItem: default item bugfix

  torzdf

  committed
  3b81da6

  Copy full SHA for 3b81da6

• PyTest fixes

  torzdf

  committed
  fd6582f

  Copy full SHA for fd6582f

• bugfix: Don't error on TestTensorBoardLogs teardown

  torzdf

  committed
  830e16e

  Copy full SHA for 830e16e

• bugfix: Mutable types in DataClass

  torzdf

  committed
  dc23009

  Copy full SHA for dc23009

• github workflow: Fix macOS

  torzdf

  committed
  8d8388f

  Copy full SHA for 8d8388f

• setup.py bugfix: Don't error on cuda key lookup

  torzdf

  committed
  381ca7d

  Copy full SHA for 381ca7d

Commits on May 22, 2024

• Github workflow: All loss calculations on CPU

  torzdf

  committed
  c7f494c

  Copy full SHA for c7f494c

• tests: Place ML ops explicitly on CPU

  torzdf

  committed
  b8cecd5

  Copy full SHA for b8cecd5

• Legacy model loading: Fix TFLamdaOp scalar ops and DepthwiseConv2D

  torzdf

  committed
  7e2f861

  Copy full SHA for 7e2f861

Commits on May 26, 2024

• Add vertical offset option for training

  torzdf

  committed
  760bcaa

  Copy full SHA for 760bcaa

Commits on Jun 4, 2024

• Merge remote-tracking branch 'upstream/master' into fs3

  torzdf

  committed
  7943f24

  Copy full SHA for 7943f24

• Typing and linting

  torzdf

  committed
  7bce2fb

  Copy full SHA for 7bce2fb

• Github actions: Add more python versions

  torzdf

  committed
  a4d0003

  Copy full SHA for a4d0003

• Add python version to workflow names

  torzdf

  committed
  0ef3f62

  Copy full SHA for 0ef3f62

• Github workflow: Exclude Python 3.12 for macOS

  torzdf

  committed
  1d3e8b2

  Copy full SHA for 1d3e8b2

Commits on Jun 7, 2024

• Implement custom training loop

  torzdf

  committed
  3e5966d

  Copy full SHA for 3e5966d

• Docs update and pin keras

  torzdf

  committed
  4e353df

  Copy full SHA for 4e353df