Python keeps crashing #24

Closed
opened 2023-02-12 03:49:26 +00:00 by rowsinator · 6 comments

Not sure what the issue is, but python keeps crashing right when it gets to the part where it starts "Generating autoregressive samples". I'm not sure what to do here as it doesn't seem to give an error message, just hard crashes.

I have attached a picture of what the output looks like in the terminal

Computer specs:
OS: Windows 10 21H1
Python version: Python 3.10.8
CPU: AMD 5900x
GPU: AMD 6900XT
RAM: 32 gigs

image

Text output from terminal:

PS G:\Ai Stuff\TorToiSe\tortoise-tts> .\start.bat

Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Initializating voice-fixer
initialized voice-fixer
Initializating TorToiSe...
Hardware acceleration found: dml
TorToiSe initialized, ready for generation.
Size of best fit: 11168279
Chunk size exceeded, clamping to: 1000000
[1/1] Generating line: As I walk through the valley, in the shadow of death, I take a look at my life, and realize there is nothing left.
Generating autoregressive samples
PS G:\Ai Stuff\TorToiSe\tortoise-tts> .\start.bat

Running on local URL: http://127.0.0.1:7860

Initializating voice-fixer
initialized voice-fixer
Hardware acceleration found: dml
TorToiSe initialized, ready for generation.
Reading from latent: ./voices\suntsulady\cond_latents.pth
[1/1] Generating line: As I walk through the valley, in the shadow of death, I take a look at my life, and realize there is nothing left.
Generating autoregressive samples

Not sure what the issue is, but python keeps crashing right when it gets to the part where it starts "Generating autoregressive samples". I'm not sure what to do here as it doesn't seem to give an error message, just hard crashes. I have attached a picture of what the output looks like in the terminal Computer specs: OS: Windows 10 21H1 Python version: Python 3.10.8 CPU: AMD 5900x GPU: AMD 6900XT RAM: 32 gigs ![image](/attachments/711e2da5-0ca7-40e7-86d2-c243d0a385ba) Text output from terminal: PS G:\Ai Stuff\TorToiSe\tortoise-tts> .\start.bat Running on local URL: http://127.0.0.1:7860 To create a public link, set `share=True` in `launch()`. Initializating voice-fixer initialized voice-fixer Initializating TorToiSe... Hardware acceleration found: dml TorToiSe initialized, ready for generation. Size of best fit: 11168279 Chunk size exceeded, clamping to: 1000000 [1/1] Generating line: As I walk through the valley, in the shadow of death, I take a look at my life, and realize there is nothing left. Generating autoregressive samples PS G:\Ai Stuff\TorToiSe\tortoise-tts> .\start.bat Running on local URL: http://127.0.0.1:7860 Initializating voice-fixer initialized voice-fixer Hardware acceleration found: dml TorToiSe initialized, ready for generation. Reading from latent: ./voices\suntsulady\cond_latents.pth [1/1] Generating line: As I walk through the valley, in the shadow of death, I take a look at my life, and realize there is nothing left. Generating autoregressive samples
Owner

How strange.

The only culprit I can think of is some funny duddy business from using python3.10, but it lets you run it fine up to it so it shouldn't be that.

  • It can't be from OOMing, your 6900XT has the same VRAM as my 6800XT
    • I doubt you upped the sample batch size, as that doesn't auto-calculate yet, and would throw a real OOM anyways
    • The other vague OOM error of The GPU device instance has been suspended. would be thrown as well (since that happens if I don't use Stable Diffusion on DirectML with --medvram).
  • Shouldn't be driver related, as I haven't updated mine in a few months I believe.
  • It also can't be from too long of a prompt, since I've used something a hair longer and I had no issues.

I suppose I'll throw up a python3.10 venv and test myself.

I can't think of anything to suggest you try to help diagnose it outside of:

  • try a smaller prompt
  • set your sample batch size under Settings to 1 (if you did set it to anything else)
  • humor yourself by ticking Low VRAM and reloading the TTS / restarting the web UI and trying there
  • keep an eye on your GPU under Task Manager > Performance > GPU # (it just works)
  • I might push out a way to override falling back to CPU mode with DirectML,
How strange. The only culprit I can think of is some funny duddy business from using python3.10, but it lets you run it fine up to it so it shouldn't be that. * It can't be from OOMing, your 6900XT has the same VRAM as my 6800XT - I *doubt* you upped the sample batch size, as that doesn't auto-calculate yet, and would throw a real OOM anyways - The other vague OOM error of `The GPU device instance has been suspended.` would be thrown as well (since that happens if I don't use Stable Diffusion on DirectML with `--medvram`). * Shouldn't be driver related, as I haven't updated mine in a few months I believe. * It also can't be from too long of a prompt, since I've used something a hair longer and I had no issues. I suppose I'll throw up a python3.10 venv and test myself. I can't think of anything to suggest you try to help diagnose it outside of: * try a smaller prompt * set your sample batch size under Settings to 1 (if you did set it to anything else) * humor yourself by ticking Low VRAM and reloading the TTS / restarting the web UI and trying there * keep an eye on your GPU under Task Manager > Performance > GPU # (it just works) * I might push out a way to override falling back to CPU mode with DirectML,
Owner

Ah, exact same silent crash on python3.10, but without the popup saying Python crashed. image

Install python3.9. Strange though, it didn't complain at all during setup.

You shouldn't need to uninstall 3.10, as the most recent python install will be used when it adds itself to the PATH env var (as that's what I exactly did a week ago).

Ah, exact same silent crash on python3.10, but without the popup saying Python crashed. ![image](/attachments/8d25ca63-d72b-4448-9483-d97cfe8eb677) Install python3.9. Strange though, it didn't complain at all during setup. You *shouldn't* need to uninstall 3.10, as the most recent python install will be used when it adds itself to the PATH env var (as that's what I exactly did a week ago).
Author

So, I know you said that I didn't need to uninstall python 3.10, but I did anyway, as I don't think I have a dependency on it. I am now using version 3.9.13 and it still crashes as soon as it gets to the generating stage, I've tried this with a few different sound samples, and can provide the one I was using if you like.

the output is more or less the same but here is a pic

image

So, I know you said that I didn't need to uninstall python 3.10, but I did anyway, as I don't think I have a dependency on it. I am now using version 3.9.13 and it still crashes as soon as it gets to the generating stage, I've tried this with a few different sound samples, and can provide the one I was using if you like. the output is more or less the same but here is a pic ![image](/attachments/494eca34-061c-47da-b557-91c9103a2f96)
Author

I have now gone through all other suggestions:

  • try a smaller prompt - Did not work
  • set your sample batch size under Settings to 1 (if you did set it to anything else) - Did not work
  • humor yourself by ticking Low VRAM and reloading the TTS / restarting the web UI and trying there - Did not work
  • keep an eye on your GPU under Task Manager > Performance > GPU # (it just works) - GPU utilization did not go above 15% as far as I can tell, I used the task manager, but can use other tools if you deem it necessary. CPU usages was relatively high on all cores, but not maxing out or anything.
I have now gone through all other suggestions: * **try a smaller prompt** - Did not work * **set your sample batch size under Settings to 1 (if you did set it to anything else)** - Did not work * **humor yourself by ticking Low VRAM and reloading the TTS / restarting the web UI and trying there** - Did not work * **keep an eye on your GPU under Task Manager > Performance > GPU # (it just works)** - GPU utilization did not go above 15% as far as I can tell, I used the task manager, but can use other tools if you deem it necessary. CPU usages was relatively high on all cores, but not maxing out or anything.
Owner

Did you delete your tortoise-venv folder and re-run setup after installing python3.9?

Did you delete your `tortoise-venv` folder and re-run setup after installing python3.9?
Author

That worked. In the rentry and the readme, think it might be worth mentioning that deleting the tortoise-venv folder after trying to fix an install might be necessary.

Might also be worth mentioning in the rentry and the readme that only 3.9 can be used more prominently, but me having a later version installed is my (and chocolaty's) fault.

Thank you so much for your help.

That worked. In the rentry and the readme, think it might be worth mentioning that deleting the tortoise-venv folder after trying to fix an install might be necessary. Might also be worth mentioning in the rentry and the readme that only 3.9 can be used more prominently, but me having a later version installed is my (and chocolaty's) fault. Thank you so much for your help.
Sign in to join this conversation.
No Label
No Milestone
No project
No Assignees
2 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: mrq/tortoise-tts#24
No description provided.