Error during "Step 2 - run the training"

[oMauOficial]

Traceback (most recent call last): File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\queueing.py", line 459, in call_prediction output = await route_utils.call_process_api( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\route_utils.py", line 232, in call_process_api output = await app.get_blocks().process_api( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\blocks.py", line 1533, in process_api result = await self.call_function( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\blocks.py", line 1151, in call_function prediction = await anyio.to_thread.run_sync( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\anyio\to_thread.py", line 56, in run_sync return await get_async_backend().run_sync_in_worker_thread( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 2144, in run_sync_in_worker_thread return await future File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 851, in run result = context.run(func, args) File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\utils.py", line 678, in wrapper response = f(args, kwargs) File "C:\MR\Finetune\xtts-finetune-webui\xtts_demo.py", line 339, in train_model os.remove(run_dir) PermissionError: [WinError 5] Acesso negado: 'C:\MR\Finetune\xtts-finetune-webui\finetune_models\run' Traceback (most recent call last): File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\queueing.py", line 459, in call_prediction output = await route_utils.call_process_api( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\route_utils.py", line 232, in call_process_api output = await app.get_blocks().process_api( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\blocks.py", line 1533, in process_api result = await self.call_function( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\blocks.py", line 1151, in call_function prediction = await anyio.to_thread.run_sync( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\anyio\to_thread.py", line 56, in run_sync return await get_async_backend().run_sync_in_worker_thread( File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 2144, in run_sync_in_worker_thread return await future File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 851, in run result = context.run(func, args) File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\utils.py", line 678, in wrapper response = f(args, kwargs) File "C:\MR\Finetune\xtts-finetune-webui\xtts_demo.py", line 339, in train_model os.remove(run_dir) PermissionError: [WinError 5] Acesso negado: 'C:\MR\Finetune\xtts-finetune-webui\finetune_models\run'

The above exception was the direct cause of the following exception:

Traceback (most recent call last): File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\queueing.py", line 497, in process_events response = await self.call_prediction(awake_events, batch) File "C:\MR\Finetune\xtts-finetune-webui\venv\lib\site-packages\gradio\queueing.py", line 468, in call_prediction raise Exception(str(error) if show_error else None) from error Exception: None

[oMauOficial]

I managed to solve it simply by deleting the run folder (C:\xtts-finetune-webui\finetune_models\run) '-'

[kohai203]

I managed to solve it simply by deleting the run folder (C:\xtts-finetune-webui\finetune_models\run) '-'

Thanks ! It helped me but right after this i'm getting :

" PermissionError: [WinError 32] couldn't access... /xtts-finetune-webui/finetune_models/run/training/GPT_XTTS_FT-May-19-2024_10+30AM-6cb8066\trainer_0_log.txt' "

Can someone help ?

[oMauOficial]

@kohai203 If you are trying to train a second voice for example, try deleting the previous voice files generated in the "wavs" folder (C:\Finetune\xtts-finetune-webui\finetune_models\dataset\wavs"). Leave only the files that you are going to train at the moment, delete the "run" folder completely and try again, it worked for me =)

[PrometheusDante]

I'm getting this error as well:

PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'D:/VER_13/_AI/xtts-finetune-webui/finetune_models/run/training/GPT_XTTS_FT-May-22-2024_02+25AM-56af2c8\trainer_0_log.txt'

And I've tried deleting the run folder as well as deleting everything inside finetune_models to restart the process, but to no avail.

[anastalaz]

I have the same problem. Anybody found a solution?

[Tenidus]

I ran into this exact issue so I restested using a previously trained model using it's reference audio and it worked which made me realize my new reference audio for the new voice must be bad. I recreated the new reference files and tried learning the new model and it worked. One thing I know I did with the new voices original reference was in Audacity I saved it as 4800mhz vs 4400mhz which could or could not be the culprit but that was the only variation I did. So double check your reference audio being used for training is good.

[PrometheusDante]

Alright, I've tested some more settings and approaches and I'm pretty sure this error is simply the result of the evaluation of the input audio amounting to 0 useable samples. I don't know what the criteria are, but the evaluation is very strictly sorting out your inputs. Would be nice to get feedback on why a sample didn't qualify to edit them better in preprocessing. For now, I've managed to train models by following the advice and shared observations in this post: https://www.reddit.com/r/Oobabooga/comments/1c09ank/so_you_want_to_finetune_an_xtts_model_let_me_help/ Quintessentially summarized as: Using multiple audio inputs with 44.1khz, 16bit, mono, single sentences that do not exceed the maximum input audio size (default 11s) and using about 50-500 such files depending on the complexity of your target voice, wheres specific accents and dialects benefit from a larger dataset and more epochs.

I'm currently running a training where I input very clear and noiseless audio (video game voice acting). Exported in 78 small files, with only 22 making it through the eval step. So keep in mind that you might make a few hundred samples to have the majority of them revoked from the get-go for reasons undisclosed.

[biollas]

Alright, I've tested some more settings and approaches and I'm pretty sure this error is simply the result of the evaluation of the input audio amounting to 0 useable samples. I don't know what the criteria are, but the evaluation is very strictly sorting out your inputs. Would be nice to get feedback on why a sample didn't qualify to edit them better in preprocessing. For now, I've managed to train models by following the advice and shared observations in this post: https://www.reddit.com/r/Oobabooga/comments/1c09ank/so_you_want_to_finetune_an_xtts_model_let_me_help/ Quintessentially summarized as: Using multiple audio inputs with 44.1khz, 16bit, mono, single sentences that do not exceed the maximum input audio size (default 11s) and using about 50-500 such files depending on the complexity of your target voice, wheres specific accents and dialects benefit from a larger dataset and more epochs.

I'm currently running a training where I input very clear and noiseless audio (video game voice acting). Exported in 78 small files, with only 22 making it through the eval step. So keep in mind that you might make a few hundred samples to have the majority of them revoked from the get-go for reasons undisclosed.

Yeah this same problem seems to be destroying everything I try to do.