Spaces:

Kangarroar

/

ApplioRVC-Inference

like 59

No application file

I cannot merge it but i modified copying the line 93, also, who are you? ^^

First of all, I am using a translator because I am not good at English. I hope the meaning is conveyed well.
I found this space in the hugging face when I was looking for a way to model it with diff-svc, but an error occurred and I duplicated this space with my personal space and made it run on cpu, but it's too slow.

I don't know if it's possible, but if possible, I'd like to fix the error and use t4 gpu.

We have run into the same issues then? Dockerfile, ports, error 403 for file uploader?

yes, the same error occurred for me, and another error is occurring now.

For sure! Im really grateful that you showed up and helped me with the ports error! My saving method was working fine but now it's weird, locally it works, so im lost there, ill add you

You were added with write powers, tell me if it works for you to modify files or if you can

Thank you! To be honest, there have been many attempts to correct the error after duplicating it, so I'm not sure if I can fix it again this time, but I'll try it.

Your audio input works and it is:

audio = st.file_uploader("Choose your audio", type=["wav"])

if audio:

    # Get the bytes data of the file

    bytes_data = audio.getvalue()

    # Get the file extension

    file_extension = os.path.splitext(audio.name)[1]

    # Create a temporary file to store the audio file

    with tempfile.NamedTemporaryFile(mode="wb", suffix=file_extension, delete=False, dir=temp_dir) as temp:
        # Write the bytes to the temporary file

        temp.write(bytes_data)

        audio_temp_file = temp.name```

I did a reboot now because for some reason it loaded everything but wont load the streamlit page

I was surprised if it was because I fixed the code because 410 error occurred. I'm glad it's not because of me.

Dont worry, this may be because I try this code locally and i run windows, so my path temp is different

Error 410 seems weird but its random or due to huggingface, because when i run this on my pc there is no error 410

Im gonna factory reboot, because its a bug or so it seems

Do you have Telegram or Discord or something so we can talk better via messages? I really wanna make this work and make rendering accesible and user friendly

Oh, I see. Just in case, it's night here, and I have something else to do a little later, so I can sleep and wake up and modify the code tomorrow.

Well, actually, I'm not very knowledgeable this way either. I don't know much about it, but I tried a lot of things on the Internet and solved errors, so it probably won't help much if we discuss it.

Dont worry! This really made my day, a pull from a project i thought no one would care, and you fixed the uploading issues for port errors! Thanks!!
Dont worry, thanks

Hmm unfortunately, 410 errors seem to continue to occur. In the case of my space, it works well for some reason, but I think that might be the problem if the error occurs after modifying the code of the docker. I hope it's a hugging face problem. If the same error continues to occur after some time, it would be better to revert the corrected code. I have no choice but to look for another way.

Maybe something is crashing but since this doesnt show errors on console via streamlit
I updated something to see how it handles

It's not a docker error, i have a clone copy of app.py in another space and it works fine but the config.yaml doesnt, this seems to be huggingface problem

Oh, that's a relief then! But it's time to go, so I think I'll have to go do something.
I'm sorry, but I'll come back tomorrow.

Dont worry, we will have to wait until huggingface fixes our allocation

Hello. Can you check the space I duplicated? I tried to fix it in my own way, but I got an error that the nvidia graphics card was not assigned.

I am not good at coding (?) or programming (?) like you, so I think it would be better to modify it in a duplicated space than to modify it directly. Now I have a hard time downloading pyworld, but I'm lucky enough to solve it somehow.

Hmm.. I want to check if there is an error, so I will save the existing requirements.txt, docker.txt, and app.py separately and apply the code I modified.

it is possible to upload wav files, yaml files, and ckpt files, and it seems that it runs rendering, but it does not seem to be able to print them out. Maybe. I'm not sure how to modify it.

Im checking, because I had a lot of errors here like you, with pyworld and stuff, now the page is fixed and gets the files, but it doesnt want to show the rendered file, this is weird? Im checking your space

I modified your space and tested it. Maybe you can check the diff-svc space!

Yes,, That's right. I put in the Ckpt file, yaml file, and music file, pressed the render button, and checked the log, and the rendering work seemed to be in progress, but it doesn't show the result.

It's weird that is not showing the result... Im looking for a fix

But its rendering! Its a big progress and this may be almost done!

Im still trying to find where this came from, because before it didnt got fails for temp or rendering...

I think it will be solved if we fix one or two more places.
Uploading ckpt, yaml, wav/mp3 works well, but it doesn't show the rendering results, so it seems like a path problem with the rendered wav file. However, in the duplicated space, this error has never occurred, so it is difficult to solve.

The duplicated space shows the rendered?

hmm, In the past, the duplicated space showed the result of rendering, but the duplicated space was not assigned a graphics card, so I couldn't test it.
This error has never occurred in yls diff or latte diff.

I was able to fix again the file loader and rendering, but we are running again into not showing the file

What is the duplicated space that shows the rendered file?

The code I've modified so far has almost been able to be modified by searching the Internet, but in this case, there is not much information even if I search the Internet. I'm not sure if I can be of any help, but I'll keep thinking about it.

Right now I'm happy that the log console shows the rendering again, we just need to find why it doesn't show the wav

Can you link me the space that shows the rendered?

this - https://huggingface.co/spaces/Mangla/Inference
Unlike the space being modified now, it loads the ckpt file and yaml from the repository. The sound file is uploaded by the user, runs rendering, and lets you hear the results.

But I didn't fix this in the part that shows the results, and it showed me the rendered wav file right away.

Oh, that space runs on another version of Linux, we are using right now Ubuntu 20.04

Maybe that's why? I'm gonna test this

I'd be glad if it worked.

IT WORKS! IT WORKS! TRY TESTING IT!!!

Holy I'm gonna test it right now!

haha, I want to test it right now, but I get 410 errors. Sadly, I'll test it again after some time.

For some weird reason the browser uses a cached version when we open the page,,,, huggingface bug

This is definitely a hug face bug. It is confirmed that 410 error occurs on smartphone, but it works properly on PC. I checked that 410 error appears or does not appear depending on the browser on the smartphone. Anyway, you did this! I'm so happy to use this space.

I mean I was able to upload everything and it showed the result, either way im using incognito mode because my pc browser just loads an old version??? man i love huggingface,,,,,,,,

Hmm.. Maybe my translator is weird in interpretation. We can share the joy of working properly, right?

When I access it from Google Chrome on my smartphone, 410 error occurs, but when I access it from Chrome's Secret tab, it works properly again. a strange bug

I've just tested the operation, and it works! Thank you again! We are pleased to be able to use diff-svc at a fast speed.

You did great too! Without your first pull i would have fallen into not keeping up with fixing this!

and you managed to fix a lot of stuff

I'm so glad it helped!I hope a lot of people can use it.

If it's not too much trouble, why don't you make the values of pndm_speedup and add_noise_step=500 adjustable? like key value.

It's an easy work, and it'll be convenient for people to use.

I plan on doing that, but first i need to measure how heavy the rendering is by itself on the T4 by multiple people, if it goes well, i can do modifiable speedup

Certainly. It could be much for many people to use!