mirror of
https://github.com/deepfakes/faceswap.git
synced 2026-10-22 01:15:59 +08:00
torzdf
837bc2d51d
* FaceSwap 3 (#1515) * Update extract pipeline * Update requirements + setup for nvidia * Remove allow-growth option * tf.keras to keras updates * lib.model.losses - Port + fix all loss functions for Keras3 * lib.model - port initializers, layers. normalization to Keras3 * lib.model.autoclip to Keras 3 * Update mixed precision layer storage * model file to .keras format * Restructure nn_blocks to initialize layers in __init__ * Tensorboard - Trainer: Add Torch compatible Tensorboard callbacks - GUI event reader remove TF dependency * Loss logging - Flush TB logs on save - Replace TB live iterator for GUI * Backup models on total loss drop rather than per side * Update all models to Keras3 Compat * Remove lib.model.session * Update clip ViT to Keras 3 * plugins.extract.mask.unet-dfl - Fix for Keras3/Torch backend * Port AdaBelief to Keras 3 * setup.py: - Add --dev flag for dev tool install * Fix Keras 3 syntax * Fix LR Finder for Keras 3 * Fix mixed precision switching for Keras 3 * Add more optimizers + open up config setting * train: Remove updating FS1 weights to FS2 models * Alignments: Remove support for legacy .json files * tools.model: - Remove TF Saved Format saving - Fix Backup/Restore + Nan-Scan * Fix inference model creation for Keras 3 * Preview tool: Fix for Keras3 * setup.py: Configure keras backend * train: Migration of FS2 models to FS3 * Training: Default coverage to 100% * Remove DirectML backend * Update setup for MacOS * GUI: Force line reading to UTF-8 * Remove redundant Tensorflow references * Remove redundant code * Legacy model loading: Fix TFLamdaOp scalar ops and DepthwiseConv2D * Add vertical offset option for training * Github actions: Add more python versions * Add python version to workflow names * Github workflow: Exclude Python 3.12 for macOS * Implement custom training loop * Fs3 - Add RTX5xxx and ROCm 6.1-6.4 support (#1511) * setup.py: Add Cuda/ROCm version select options * bump minimum python version to 3.11 * Switch from setup.cgf to pyproject.toml * Documentation: Update all docs to use automodapi * Allow sysinfo to run with missing packages + correctly install tk under Linux * Bugfix: dot naming convention in clip models * lib.config: Centralise globally rather than passing as object - Add torch DataParallel for multi-gpu training - GUI: Group switches together when generating cli args - CLI: Remove deprecated multi-character argparse args - Refactor: - Centralise tensorboard reading/writing + unit tests - Create trainer plugin interfaces + add original + distributed * Update installers
94 lines
3.0 KiB
Python
94 lines
3.0 KiB
Python
#!/usr/bin/env python3
|
|
""" The default options for the faceswap Dfl_SAE Model plugin.
|
|
|
|
Defaults files should be named `<plugin_name>_defaults.py`
|
|
|
|
Any qualifying items placed into this file will automatically get added to the relevant config
|
|
.ini files within the faceswap/config folder and added to the relevant GUI settings page.
|
|
|
|
The following variable should be defined:
|
|
|
|
Parameters
|
|
----------
|
|
HELPTEXT: str
|
|
A string describing what this plugin does
|
|
|
|
Further plugin configuration options are assigned using:
|
|
>>> <config_item> = ConfigItem(...)
|
|
|
|
where <config_item> is the name of the configuration option to be added (lower-case, alpha-numeric
|
|
+ underscore only) and ConfigItem(...) is the [`~lib.config.objects.ConfigItem`] data for the
|
|
option.
|
|
|
|
See the docstring/ReadtheDocs documentation required parameters for the ConfigItem object.
|
|
Items will be grouped together as per their `group` parameter, but otherwise will be processed in
|
|
the order that they are added to this module.
|
|
from lib.config import ConfigItem
|
|
"""
|
|
from lib.config import ConfigItem
|
|
|
|
|
|
HELPTEXT = "DFL SAE Model (Adapted from https://github.com/iperov/DeepFaceLab)"
|
|
|
|
|
|
input_size = ConfigItem(
|
|
datatype=int,
|
|
default=128,
|
|
group="size",
|
|
info="Resolution (in pixels) of the input image to train on.\n"
|
|
"BE AWARE Larger resolution will dramatically increase VRAM requirements.\n"
|
|
"\nMust be divisible by 16.",
|
|
rounding=16,
|
|
min_max=(64, 256),
|
|
fixed=True)
|
|
|
|
architecture = ConfigItem(
|
|
datatype=str,
|
|
default="df",
|
|
group="network",
|
|
info="Model architecture:"
|
|
"\n\t'df': Keeps the faces more natural."
|
|
"\n\t'liae': Can help fix overly different face shapes.",
|
|
choices=["df", "liae"],
|
|
gui_radio=True,
|
|
fixed=True)
|
|
|
|
autoencoder_dims = ConfigItem(
|
|
datatype=int,
|
|
default=0,
|
|
group="network",
|
|
info="Face information is stored in AutoEncoder dimensions. If there are not enough "
|
|
"dimensions then certain facial features may not be recognized."
|
|
"\nHigher number of dimensions are better, but require more VRAM."
|
|
"\nSet to 0 to use the architecture defaults (256 for liae, 512 for df).",
|
|
rounding=32,
|
|
min_max=(0, 1024),
|
|
fixed=True)
|
|
|
|
encoder_dims = ConfigItem(
|
|
datatype=int,
|
|
default=42,
|
|
group="network",
|
|
info="Encoder dimensions per channel. Higher number of encoder dimensions will help "
|
|
"the model to recognize more facial features, but will require more VRAM.",
|
|
rounding=1,
|
|
min_max=(21, 85),
|
|
fixed=True)
|
|
|
|
decoder_dims = ConfigItem(
|
|
datatype=int,
|
|
default=21,
|
|
group="network",
|
|
info="Decoder dimensions per channel. Higher number of decoder dimensions will help "
|
|
"the model to improve details, but will require more VRAM.",
|
|
rounding=1,
|
|
min_max=(10, 85),
|
|
fixed=True)
|
|
|
|
multiscale_decoder = ConfigItem(
|
|
datatype=bool,
|
|
default=False,
|
|
group="network",
|
|
info="Multiscale decoder can help to obtain better details.",
|
|
fixed=True)
|