github/DeepFaceLab
1
0
Fork 0
mirror of https://github.com/iperov/DeepFaceLab.git synced 2025-08-22 06:23:20 -07:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fixes

Browse source
This commit is contained in:
TalosOfCrete 2020-06-08 13:22:53 -05:00
commit 082394193a
1 changed files with 10 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

20
models/Model_SAEHD/Model.py
View file

@ -61,17 +61,17 @@ class SAEHDModel(ModelBase):
self.ask_batch_size(suggest_batch_size)
if self.is_first_run():
resolution = io.input_int("Resolution", default_resolution, add_info="64-512", help_message="More resolution requires more VRAM and time to train. Value will be adjusted to multiple of 16.")
resolution = io.input_int("Resolution", default_resolution, add_info="64-512", help_message="Higher resolution requires more VRAM and training time. Value will be adjusted to multiple of 16.")
resolution = np.clip ( (resolution // 16) * 16, 64, 512)
self.options['resolution'] = resolution
self.options['face_type'] = io.input_str ("Face type", default_face_type, ['h','mf','f','wf','head'], help_message="Half / mid face / full face / whole face / head. Half face has better resolution, but covers less area of cheeks. Mid face is 30% wider than half face. 'Whole face' covers full area of face including forehead. 'head' covers full head, but requires XSeg for src and dst faceset.").lower()
self.options['archi'] = io.input_str ("AE architecture", default_archi, ['df','liae','dfhd','liaehd','dfuhd','liaeuhd'], help_message="'df' keeps faces more natural.\n'liae' can fix overly different face shapes.\n'hd' are experimental versions.").lower()
self.options['separable'] = io.input_bool ("Use depthwise separable convolutions throughout model", default_separable, help_message="Use lighter and more effecient layers in the encoder, bottleneck, and decoder. This speeds up iterations (~15-200%) and reduces memory usage considerably in the encoder and decoder (~60%) - allowing for increased settings or faster training - but it may take more iterations for the model to become good and possibly hurt quality. Set this to false if you wish to only have these layers in certain parts of the model (e.g. only in the encoder) or if you do not want them at all.").lower()
self.options['separable'] = io.input_bool ("Use depthwise separable convolutions throughout model", default_separable, help_message="Use lighter and more effecient layers in the encoder, bottleneck, and decoder. This speeds up iterations (~15-200%) and reduces memory usage considerably in the encoder and decoder (~60%) - allowing for increased settings or faster training - but it may take more iterations for the model to become good and possibly hurt quality. Set this to false if you wish to only have these layers in certain parts of the model (e.g. only in the encoder) or if you do not want them at all.")
if not self.options['separable']:
self.options['separable_enc'] = io.input_bool ("Use depthwise separable convolutions in the encoder", default_separable_enc, help_message="This is the part of the model second most impacted by using these more efficient layers in terms of better iteration speed, memory savings, possible quality loss, and possible increase in required iterations as it may have anywhere from one to four downscaling operations for UHD and HD architectures respectively.").lower()
self.options['separable_inter'] = io.input_bool ("Use depthwise separable convolutions in the bottleneck", default_separable_inter, help_message="This is the part of the model least impacted by using these more efficient layers in terms of better iteration speed, memory savings, possible quality loss, and possible increase in required iterations as it only has a single upscaling operation regardless of architecture.").lower()
self.options['separable_dec'] = io.input_bool ("Use depthwise separable convolutions in the decoder", default_separable_dec, help_message="This is the part of the model most impacted by using these more efficient layers in terms of better iteration speed, memory savings, possible quality loss, and possible increase in required iterations as it has by far the most operations that these layers use out of all parts of the model.").lower()
self.options['separable_enc'] = io.input_bool ("Use depthwise separable convolutions in the encoder", default_separable_enc, help_message="This is the part of the model second most impacted by using these more efficient layers in terms of better iteration speed, memory savings, possible quality loss, and possible increase in required iterations as it may have anywhere from one to four downscaling operations for UHD and HD architectures respectively.")
self.options['separable_inter'] = io.input_bool ("Use depthwise separable convolutions in the bottleneck", default_separable_inter, help_message="This is the part of the model least impacted by using these more efficient layers in terms of better iteration speed, memory savings, possible quality loss, and possible increase in required iterations as it only has a single upscaling operation regardless of architecture.")
self.options['separable_dec'] = io.input_bool ("Use depthwise separable convolutions in the decoder", default_separable_dec, help_message="This is the part of the model most impacted by using these more efficient layers in terms of better iteration speed, memory savings, possible quality loss, and possible increase in required iterations as it has by far the most operations that these layers use out of all parts of the model.")
else:
self.options['separable_enc'] = True
self.options['separable_inter'] = True
@ -99,15 +99,15 @@ class SAEHDModel(ModelBase):
if self.is_first_run() or ask_override:
if self.options['face_type'] == 'wf' or self.options['face_type'] == 'head':
self.options['masked_training'] = io.input_bool ("Masked training", default_masked_training, help_message="This option is available only for 'whole_face' type. Masked training clips training area to full_face mask, thus network will train the faces properly. When the face is trained enough, disable this option to train all area of the frame. Merge with 'raw-rgb' mode, then use Adobe After Effects to manually mask and compose whole face include forehead.")
self.options['masked_training'] = io.input_bool ("Masked training", default_masked_training, help_message="This option is available only for 'whole_face' type. Masked training clips training area to full_face mask, thus the network will train the faces properly. When the face is trained enough, disable this option to train the entire frame. Merge with 'raw-rgb' mode, then use Adobe After Effects to manually mask and compose whole face including forehead.")
self.options['eyes_prio'] = io.input_bool ("Eyes priority", default_eyes_prio, help_message='Helps to fix eye problems during training like "alien eyes" and wrong eyes direction ( especially on HD architectures ) by forcing the neural network to train eyes with higher priority. before/after https://i.imgur.com/YQHOuSR.jpg ')
self.options['eyes_prio'] = io.input_bool ("Eyes priority", default_eyes_prio, help_message='Helps to fix eye problems during training, like "alien eyes" and incorrect eye directions (especially on HD architectures) by forcing the neural network to train the eyes with higher priority. before/after https://i.imgur.com/YQHOuSR.jpg ')
if self.is_first_run() or ask_override:
self.options['models_opt_on_gpu'] = io.input_bool ("Place models and optimizer on GPU", default_models_opt_on_gpu, help_message="When you train on one GPU, by default model and optimizer weights are placed on GPU to accelerate the process. You can place they on CPU to free up extra VRAM, thus set bigger dimensions.")
self.options['lr_dropout'] = io.input_bool ("Use learning rate dropout", default_lr_dropout, help_message="When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness and reduce subpixel shake for less amount of iterations.")
self.options['lr_dropout'] = io.input_bool ("Use learning rate dropout", default_lr_dropout, help_message="When the face is trained enough, you can enable this option to get extra sharpness and reduce subpixel shake for fewer iterations.")
self.options['random_warp'] = io.input_bool ("Enable random warp of samples", default_random_warp, help_message="Random warp is required to generalize facial expressions of both faces. When the face is trained enough, you can disable it to get extra sharpness and reduce subpixel shake for fewer iterations.")
self.options['gan_power'] = np.clip ( io.input_number ("GAN power", default_gan_power, add_info="0.0 .. 10.0", help_message="Train the network in Generative Adversarial manner. Accelerates the speed of training. Forces the neural network to learn small details of the face. You can enable/disable this option at any time. Typical value is 1.0"), 0.0, 10.0 )
@ -117,7 +117,7 @@ class SAEHDModel(ModelBase):
self.options['true_face_power'] = 0.0
self.options['face_style_power'] = np.clip ( io.input_number("Face style power", default_face_style_power, add_info="0.0..100.0", help_message="Learn the color of the predicted face to be the same as dst inside mask. If you want to use this option with 'whole_face' you have to use XSeg trained mask. Warning: Enable it only after 10k iters, when predicted face is clear enough to start learn style. Start from 0.001 value and check history changes. Enabling this option increases the chance of model collapse."), 0.0, 100.0 )
self.options['bg_style_power'] = np.clip ( io.input_number("Background style power", default_bg_style_power, add_info="0.0..100.0", help_message="Learn the area outside mask of the predicted face to be the same as dst. If you want to use this option with 'whole_face' you have to use XSeg trained mask. For whole_face you have to use XSeg trained mask. This can make face more like dst. Enabling this option increases the chance of model collapse. Typical value is 2.0"), 0.0, 100.0 )
self.options['bg_style_power'] = np.clip ( io.input_number("Background style power", default_bg_style_power, add_info="0.0..100.0", help_message="Learn the area outside of the mask of the predicted face to be the same as dst. If you want to use this option with 'whole_face' you have to use XSeg trained mask. For whole_face you have to use XSeg trained mask. This can make face more like dst. Enabling this option increases the chance of model collapse. Typical value is 2.0"), 0.0, 100.0 )
self.options['ct_mode'] = io.input_str (f"Color transfer for src faceset", default_ct_mode, ['none','rct','lct','mkl','idt','sot'], help_message="Change color distribution of src samples close to dst samples. Try all modes to find the best.")
self.options['clipgrad'] = io.input_bool ("Enable gradient clipping", default_clipgrad, help_message="Gradient clipping reduces chance of model collapse, sacrificing speed of training.")