H2O LLM Studio using Google Colab?

[Anox135]

Hey,
I want to run the h20-llmstudio using google colab (not using CLI but the no-code interface). Is there any way I can do this? I have tried different ways, but it all fails when I want to connect to localhost:10101. Do you have any Idea how I can connect to localhost when using colab? Or any other tips? Thanks in advance

What I have tried:
Using ngrok
Using pinggy to create a public URL for localhost

[pascal-pfeiffer]

Ngrok is one way that certainly works. How did you try it?

[Anox135]

You can see everything that I have tried and the errors here:
https://colab.research.google.com/drive/1SBdJ_PoSjTBSvDCRqrHzrgoMe8aLDP-w?usp=sharing

Thanks in advance for any tips.

[pascal-pfeiffer]

I guess your wave server was never running, could you try this instead:

# !pip install -r requirements.txt
!make setup

# !app.run_server(mode='external')
from subprocess import Popen, run
try:
    app_process.kill()
except:
    pass
app_process = Popen(["make", "wave"])

!ngrok config add-authtoken XXXX

from pyngrok import ngrok
url = ngrok.connect(addr=10101)
url = url.public_url
print(f"Successfully established ngrok tunnel, visit the app at {url}.")

[Anox135]

Okay, thanks. It kinda works now. I can start the llm studio but when I start it, it cannot train any model.

I am not sure but I think it has to do with this issue: When running:

!app.run_server(mode = 'external') from subprocess import Popen, run try: app_process.kill() except: pass app_process = Popen(["make", "wave"])

This error comes up:
/bin/bash: -c: line 1: syntax error near unexpected token 'mode' /bin/bash: -c: line 1: 'app.run_server(mode = 'external')'

Thank you very much for your help, I am kinda stuck at this problem rn.

[pascal-pfeiffer]

# !app.run_server(mode='external')

this should be commented out.

[pascal-pfeiffer]

closing due to inactivity
Assuming that it is solved.