training doesn't proceeded.

[androboy]

After run the training,
Below logs printed and progress bar moved to left and right,
But any action was not comming.
No logs update, No new files at output directories.

Code: Select all

Loading...
Setting Faceswap backend to AMD
02/03/2022 15:31:29 INFO Log level set to: INFO
02/03/2022 15:31:29 INFO Setting up for PlaidML
02/03/2022 15:31:29 INFO Using GPU(s): ['Advanced Micro Devices, Inc. - gfx902 (experimental)']
02/03/2022 15:31:29 INFO Successfully set up for PlaidML
02/03/2022 15:31:34 INFO Model A Directory: 'D:\Entertain\Game\NewWorld\sam1\sa1_fa' (520 images)
02/03/2022 15:31:34 INFO Model B Directory: 'D:\Entertain\Game\NewWorld\sam1\na1_fa' (481 images)
02/03/2022 15:31:34 INFO Training data directory: D:\Entertain\Game\NewWorld\sam1\model_1
02/03/2022 15:31:34 INFO ===================================================
02/03/2022 15:31:34 INFO Starting
02/03/2022 15:31:34 INFO Press 'Stop' to save and quit
02/03/2022 15:31:34 INFO ===================================================
02/03/2022 15:31:35 INFO Loading data, this may take a while...
02/03/2022 15:31:35 INFO Loading Model from Original plugin...
Using plaidml.keras.backend backend.
02/03/2022 15:31:36 INFO No existing state file found. Generating.
02/03/2022 15:31:36 INFO Opening device "opencl_amd_gfx902.0"
02/03/2022 15:31:38 INFO Loading Trainer from Original plugin...
02/03/2022 15:31:43 INFO Analyzing Ops: 445 of 1447 operations complete
02/03/2022 15:31:45 INFO Analyzing Ops: 643 of 1447 operations complete
02/03/2022 15:31:47 INFO Analyzing Ops: 877 of 1447 operations complete

And after click the stop button,
Below logs following.

Code: Select all

Sending Exit Signal
02/03/2022 15:45:26 INFO Exit requested! The trainer will complete its current cycle, save the models and quit (This can take a couple of minutes depending on your training speed).
02/03/2022 15:47:25 ERROR Timeout reached sending Exit Signal
Terminating Process...
Terminated
Process exited

What I have to do?

[torzdf]

It looks like you are using an integrated GPU. Most likely you will never get anywhere with it, but you can try using the Lightweight model and setting a very low batchsize and going from there.