killianlucas / ooba Goto Github PK

Hi, Running on Colab with T4 GPU and Gitpod ends with the following message:

Getting started...
Server took too long to start

---------------------------------------------------------------------------

Exception                                 Traceback (most recent call last)

[/usr/local/lib/python3.10/dist-packages/ooba/llm.py](https://localhost:8080/#) in __init__(self, path, cpu, verbose, first_time)
    104             else:
--> 105                 raise Exception("Server took too long to start")
    106 

Exception: Server took too long to start


During handling of the above exception, another exception occurred:

Exception                                 Traceback (most recent call last)

2 frames

[/usr/local/lib/python3.10/dist-packages/ooba/llm.py](https://localhost:8080/#) in __init__(self, path, cpu, verbose, first_time)
    103                     time.sleep(1.5)
    104             else:
--> 105                 raise Exception("Server took too long to start")
    106 
    107 

Exception: Server took too long to start

The model dolphin-2.6-mixtral-8x7b-GGUFis really big, but it worked in webui, this issue is also known at open-interpreter, is there any timeout param to increase?
Another issue GPU is not detected, it falls back to CPU on Colab.

Looks like this issue is better here, as its the ooba check that is failing:

OpenInterpreter/open-interpreter#713

It goes to a 100% and beyond. 1980 or so..

I know it's early days, but consider this some early adopter's feedback ;)

Running llm = ooba.llm(path) on Windows is really s.l.o.w. This is after the git clone and switch branch. i can see it creating the conda env and everything, so I appreciate it needs to download stuff. Which, in my case, is pretty fast with a fiber connection.

I've tried interrupting the process and running things manually, but I couldn't figure out every single step. Looking at some of the scripts, this is all pretty smart! 🤓
No worries, I'll just let it sit there and come back later, but I wanted to let you know. Will try and reproduce on WSL later.

So since #1 is just sitting there, I'm giving WSL2 on Win11 a go.

It falls over and I think I see why.

if system == "Linux":
        if "microsoft" in platform.uname().release.lower():
            self.start_script = os.path.join(REPO_DIR, "start_wsl.bat")

We can't run a .bat script from WSL 😊
Since the script basically calls wsl and runs wsl.sh, I chose to run that manually and all ran smoothly!