Running example, hangs and then is killed

[smithlt]

It hangs for a long time. Only output is:

$ python3.8 demo.py Killed

Any ideas? It's a Google Cloud instance.

$ uname -r 4.19.0-23-cloud-amd64

[gkucsko]

which demo file is this? does it come from a fork or what does it contain?

[smithlt]

It's the first demo from the README

Edit: I also tried removing workbook audio playback (the last line) to see if that was the problem, no dice.

[smithlt]

Here's the complete file

$ cat demo.py
from bark import SAMPLE_RATE, generate_audio, preload_models
from IPython.display import Audio
from scipy.io.wavfile import write as write_wav

# download and load all models
preload_models()

# generate audio from text
text_prompt = """
     Hello, my name is Suno. And, uh — and I like pizza. [laughs] 
     But I also have other interests such as playing tic tac toe.
"""
audio_array = generate_audio(text_prompt)

[mcamac]

which Google Cloud instance type is this? CPU or GPU? If you run preload_models() by itself does it still crash?

[smithlt]

Machine configuration

Machine type n1-standard-1 CPU platform Intel Skylake Architecture — vCPUs to core ratio

—

Custom visible cores — Display device Disabled Enable to use screen capturing and recording tools GPUs 1 x NVIDIA T4

[smithlt]

Seems to hang in the same way when only preload_models() is included.

A bad dependency? Is there a docker image out there for running bark?

[gkucsko]

don't you need to turn this into a proper script to run it? have you tried running it in a notebook or ipython?

[smithlt]

what does proper script mean

[gkucsko]

hm, actually can you just try running it in ipython or a notebook and see what happens? wondering if you are for some reason just not seeing the logging messages that download the models etc.

[smithlt]

---instance-1:/opt/repos/bark$ ipython
Python 3.9.2 (default, Feb 28 2021, 17:03:44) 
Type 'copyright', 'credits' or 'license' for more information
IPython 8.13.1 -- An enhanced Interactive Python. Type '?' for help.

In [1]: 

In [1]: from bark import SAMPLE_RATE, generate_audio, preload_models
   ...: from IPython.display import Audio
   ...: from scipy.io.wavfile import write as write_wav
   ...: 
   ...: # download and load all models
   ...: preload_models()
   ...: 
   ...: # generate audio from text
   ...: text_prompt = """
   ...:      Hello, my name is Suno. And, uh — and I like pizza. [laughs] 
   ...:      But I also have other interests such as playing tic tac toe.
   ...: """
   ...: audio_array = generate_audio(text_prompt)
No GPU being used. Careful, inference might be very slow!
Killed

Maybe getting it to use a GPU (different instance) might help?

[gkucsko]

would definitely make it faster. on cpu it can take minutes or longer per sample

[gkucsko]

gonna close for inactivity. feel free to reopen if not fixed

[Alnounou-UHCL]

I am getting the same issue. I do not know why it is killed.

I successfully ran it 2 times, but starting from the 3rd attempt it hangs and is killed. I am also running on CPU without a discreet GPU on this laptop.

[mpastewski]

I experienced the same issue when running Bark on GCP, and I resolved it by increasing the amount of RAM to 16 GB.

Also, when running Bark, you'll quickly run out of storage, so also keep an eye on that.

[Alnounou-UHCL]

About that, do you know of a way to clear the allocated storage? When using a GPU I could clear the allocated memory as well as reduce the model sizes and that resolved my issues. I do not know how to do that for CPU runs, I can reduce the model sizes, but clearing them after I am done generating an audio file is where I'm stuck at.