ModuleNotFoundError: No module named 'LIHQ.QVI.demo'

Question

ModuleNotFoundError: No module named 'LIHQ.QVI.demo'

nexus-collab opened this issue 2 years ago · comments

nexus-collab commented 2 years ago

I get this error every time I try to run the collab

John Gettings · Answer 1 · Mon Nov 28 2022 23:50:42 GMT+0800 (China Standard Time)

On which cell do you get this error?

John Gettings · Answer 2 · Tue Nov 29 2022 00:38:22 GMT+0800 (China Standard Time)

Okay I found it.
It looks like gdown has a bug in it that wasn't allowing me to download demo.py. I upgraded to the latest version in requirements.txt and it seems to be working for me now.

nexus-collab · Answer 3 · Tue Nov 29 2022 01:33:55 GMT+0800 (China Standard Time)

It is running now but still get an error every time (attached screenshot), it says video complete but when I go to that folder it is just some example footage not my footage [image: image.png]

…

_____________________ Sincerely, Corey Wade 323.739.7664

On Mon, Nov 28, 2022 at 11:50 AM John Gettings ***@***.***> wrote: Closed #2 <#2> as completed. — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3QLBILM7NDGI4I45IDXTG3WKTPF7ANCNFSM6AAAAAASM6GO2I> . You are receiving this because you authored the thread.Message ID: ***@***.***>

John Gettings · Answer 4 · Tue Nov 29 2022 02:53:47 GMT+0800 (China Standard Time)

I'm not seeing the image

John Gettings · Answer 5 · Tue Nov 29 2022 09:20:12 GMT+0800 (China Standard Time)

There are a lot of options within the primary LIHQ colab but if you look at the 'Examples' colab you'll find it can be pretty simple as well.
All you need is audio file(s) in the audio folder, a cropped image of a face (specify location with the 'face' argument in run.py) , click run and find the video output in: LIHQ ->output -> finalVidsOut

If you complete those three steps it should work, as long as the audio is clean and the face is within specifications mentioned in the colab. Everything else is optional.

nexus-collab · Answer 6 · Tue Nov 29 2022 09:29:27 GMT+0800 (China Standard Time)

The issue is it completes all frames but when it tries to generate a video it crashes and disconnects, deleting everything I have uploaded as well as any outputs, happens every time. I brought image size down to 512x512, still doesn't work

…

_____________________ Sincerely, Corey Wade 323.739.7664

On Mon, Nov 28, 2022 at 8:20 PM John Gettings ***@***.***> wrote: There are a lot of options within the primary LIHQ colab but if you look at the 'Examples' colab you'll find it can be pretty simple as well. All you need is audio file(s) in the audio folder, a cropped image of a face (specify location with the 'face' argument in run.py) , click run and find the video output in: LIHQ ->output -> finalVidsOut If you complete those three steps it should work, as long as the audio is clean and the face is within specifications mentioned in the colab. Everything else is optional. — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3QLBIMRZKRROAN2IBEEYNLWKVK5PANCNFSM6AAAAAASM6GO2I> . You are receiving this because you authored the thread.Message ID: ***@***.***>

John Gettings · Answer 7 · Tue Nov 29 2022 10:56:58 GMT+0800 (China Standard Time)

Face image size shouldn't matter too much.

What is the length of your audio? As it states in the name, the tool has a LONG inference time and uses a couple models that requires high RAM use. This tool is not really meant to create a 15 minute video. I would recommend keeping audio clips to under 30 seconds and splicing together finished video clips as necessary. Be sure to look at 'ref_vid_offset' argument if you want the spliced clips to look seamless.

If you are still running out of memory go into colab settings: Runtime -> Change Runtime Type -> select high-RAM option.
I have tested on a T4 and have never received OOM, so check your GPU and see if it is anything less powerful than that. It may be a problem. You can subscribe to google colab+ and get more RAM, better GPUs, more frequently.