Open computersrmyfriends opened 3 months ago
Got Started with this:
diffusers==0.27.2 numpy==1.26.4 scipy==1.13.0 torch==2.2.0 transformers==4.39.3
Hope it completes the process.
Nice efforts. 🙏👍
Yeah, things are pretty barebones right now. I'll put a requirements doc on my TODO. Thanks for putting in the effort to get it running. Hope it works for you!
It stopped after a while for me on a A4000 with 16G VRAM gpu
Finally, it went through and got stuck here:
sing the model-agnostic default `max_length` (=1500) to control the generation length. We recommend setting `max_new_tokens` to control the maximum length of the generation.
Traceback (most recent call last):
File "/root/vworkspace/Robot-Podcast-Generator/main.py", line 148, in <module>
generate_theme_song()
File "/root/vworkspace/Robot-Podcast-Generator/gen_music.py", line 21, in generate_theme_song
scipy.io.wavfile.write(
File "/opt/conda/lib/python3.10/site-packages/scipy/io/wavfile.py", line 766, in write
fid = open(filename, 'wb')
FileNotFoundError: [Errno 2] No such file or directory: 'Podcast Generator/tunes/themesong.wav'
After another try, it's something else. Is something missing? Maybe it didn't generate the themesong.wav?
Looks like it was a directory naming issue. generate_theme_song was trying to write to "Podcast Generator" which is an old name for this repo. I just pushed a fix. Thanks for finding that.
I'm surprised that an A4000 struggled with the image generator, although that is admittedly a hungry boy. If you have similar struggles in the future, try editing line 24 of gen_img where the logo is scaled up to 1400x1400 (a recommended size for podcast banners). You can comment that out or just scale the image up to a smaller size like 700x700.
There's no requirements.txt.
How to setup and get it running?