Finished training, but it hasn't #357

New Issue

MrMustachio43 · 2023-08-29T17:28:56Z

MrMustachio43 commented

2023-08-29 17:28:56 +00:00

The first image is what was shown after it had "finished training". nothing further happened.
The second image shows my config, as you'll see it should have done an extra 400 epochs.

The third + fourth image shows that there are 500 audio files but only 150 has been transcribed and processed. This is because I added 150 files followed by an extra 200, in the preparation stage I did processes and transcribe them and they were added to the whisper file, but not train.txt file.

this is probably a rookie mistake, but I don't know what I've done wrong

The first image is what was shown after it had "finished training". nothing further happened. The second image shows my config, as you'll see it should have done an extra 400 epochs. The third + fourth image shows that there are 500 audio files but only 150 has been transcribed and processed. This is because I added 150 files followed by an extra 200, in the preparation stage I did processes and transcribe them and they were added to the whisper file, but not train.txt file. this is probably a rookie mistake, but I don't know what I've done wrong

Screenshot 2023-08-29 181001.png

103 KiB

Screenshot 2023-08-29 181018.png

107 KiB

Screenshot 2023-08-29 181106.png

53 KiB

Screenshot 2023-08-29 181210.png

70 KiB

mrq commented

2023-08-29 20:22:53 +00:00

In the Training > Prepare Dataset tab, you can click "(Re)Create Datasets" for it force regenerate the train.txt file again, although it's odd that it seemed to have done the prior steps without doing that. I would also, to be safe, rename the train.txt file to something like train.txt.bkp just in case it just doesn't want to overwrite that file.

I'll have to take a look at the spaghetti code again and see if there's an oversight somewhere with it not regenerating the file.

In the `Training > Prepare Dataset` tab, you can click "(Re)Create Datasets" for it force regenerate the `train.txt` file again, although it's odd that it seemed to have done the prior steps without doing that. I would also, to be safe, rename the `train.txt` file to something like `train.txt.bkp` just in case it just doesn't want to overwrite that file. I'll have to take a look at the spaghetti code again and see if there's an oversight somewhere with it not regenerating the file.

Sign in to join this conversation.