#Devs using windows looking to team up to get it working?

103 messages · Page 1 of 1 (latest)

frank void
#

If there is anyone else on windows that contributes or would like to contribute to the project with me in a PR to see if we can get things working I'd be happy to team up.

I'm gonna start tinkering tonight and this weekend to see if I can get things working or at least rule out some potential causes for the issues on windows people are having based on what I have read.

If anyone here using windows does have the project working, I'd love to figure out how too. I think it could be helpful for anyone looking to try it out

frank void
#

Linting seems not to come up with anything that I can see so far.

Going to double check pathing to see if anything there might be causing a ruckus

silver adder
frank void
#

This is just ooba right? im unfamiliar with ooba

silver adder
#

Open-interpreter uses ooba to run text-generation-webui which is what runs the local models

strong beacon
#

I have it running with ooba, with issues. For every message, another ooba instance starts up.
With local API base or OpenAI, it runs fine.

silver adder
#

Working on it now ^^

silver adder
#

I have a fix...

frank void
#

Oh?

silver adder
#

@strong beacon @frank void

frank void
#

ill be right back to vc

silver adder
#

Ok👍

silver adder
#

@strong beacon Are there any other bugs you have noticed?

strong beacon
#

The other bugs for me were with the ooba dependencies, which you've largely fixed, sir ☺️

#

And json format issues but those are due to using certain models and I consider them out of scope for OI.

silver adder
#

There are 3 pull requests now, hopefully people can use open-interpreter on windows once merged

silver adder
#

We need better error handling... all the errors people report looks the same xD

strong beacon
#

I didn't create issues for that. Cause I know, say a Llama v2 based model needs a certain prompt template, or an API server like ooba might respond with an unexpected json format. That's not due to OI.

silver adder
#

I see

#

I separated the original pull request to 2 smaller pull request because I accidently deleted the original one while learning git.

frank void
#

having wild pathing issues because im a smooth brain thats all it seems to be catching me out on

silver adder
#

@junior bay Could we have a separate channel for development and bug squashing?

hybrid sluice
#

hey guys, I am interested in helping. I just installed the latest on top of the previous. it seems to have an issue with removing the backticks. It visually shows two at the beginning and 2 at the end and complains about the last line (where the backticks are, based on counting)

strong beacon
#

Hey Bart, we welcome any help 👍🏽

#

The backtick issue should be gone now since v0.1.8 if I'm not mistaken...

dreamy hound
#

hey @strong beacon can you please tell me which software we have to install before installing open-interpreter , because i am getting many errors

strong beacon
#

You need Python 3.10 or 3.11 (don't go with 3.12 yet)

#

and I would recommend you to use conda as well

#

Now, your question is a bit vague and I can't know what you're trying to setup...

#

For use of local models, the setup is complex. First off, do you use an Nvidia GPU?

#

If so, you'll need pytorch, but the OI setup should take care of that...

#

Let's continue there

dreamy hound
frank void
#

3.10+ is the way fellers

#

I think @silver adder fixed some of the backticks stuff and its in a PR to be merged.

Did you guys get it running on windows or are you currently trying to?

silver adder
frank void
#

weird. for me now its working fine

silver adder
#

If someone has some time to spare, please review this https://github.com/KillianLucas/ooba/pull/8. I just commented out the part that was restarting the llm server. I assume it is there to start the oobabooga if it crashes or dont start?

frank void
#

i'll send my system details and see what oculd the issue be. for me I had awful trouble

#

biggest takeaway for me was the env variable stuff but I wasnt thinking at all.

check "python --version" and your environment variables to make sure

silver adder
#

For me it is working fine now.

silver adder
frank void
#

its working great for me tbh

trail trench
#

Hey there, @silver adder! Just saw your ping about the ooba repo.

I’m only involved with the core open interpreter repo, but I’ll see if I can get someone to review and expedite these PRs.

silver adder
trail trench
#

Really appreciate the efforts to get this working well for Windows folks!

silver adder
strong beacon
#

This 👆

#

So selfless 🤣

silver adder
#

🤣

solemn hare
#

stuck on getting started fixed in any pr?

silver adder
solemn hare
#

Yes

silver adder
#

These two

#

To run interpreter in debug mode and local use interpreter -l -d It gives better error messages if you have any issues

solemn hare
#

Sure 👍

silver adder
#

Report back if it is working for you or you have an issue 🙂

frank void
#

Anton you're killing it

trail trench
#

I have it on good authority that these PRs should be looked at soon.

devout notch
devout notch
#

So i got open interpreter working on windows using LM studio and creating a server but now all my responses are returning the original prompt open interpreter uses like this: has anyone seen this?

Screenshot_2023-10-26_at_7.29.55_AM.png
frank void
#

what version of python arey you on now?

devout notch
#

3.10.8

plain silo
#

@devout notch was it as easy as changing the endpoint in the config? Been needing to try thisas well. just been having great luck with 3.5-16k and adding few shots. Thanks

devout notch
plain silo
#

Thank you Obi!

devout notch
#

Very chill experience getting it to work for my own endpoint via LM studio

#

But the responses are a bit wonky, need to tune the input to shape the object correctly and increase the input token limit

devout notch
devout notch
# plain silo Thank you Obi!

Of course, make sure to replace the url with your own, you can also replace the model by using - - model “Name of Model”. So far I have downloaded Some hugging face models and ran them through LM studio, that seems to be the easiest path forward to getting llama 2 running on open-interpreter locally

#

Using the - - models flag hasn’t been very fruitful for me

frank void
stiff cape
#

Thanks for taking the time to fix OI for windows, can't wait for the next update to OI @silver adder @frank void @strong beacon @devout notch

silver adder
#

^^

devout notch
#

are people having trouble getting mistral to work locally?

silver adder
#

yup

#

it's the PRs I made for Ooba, you can use it while waiting for them to be merged

devout notch
#

what problem did you solve? i have it running on windows 10 python 3.10.8 local mistral7b

#

with main branch

#

not trying to be snarky at all just trying to avoid rabbit holes myself

#

thanks for your work

silver adder
#

These

#

I just merged them into my fork so I could use interpreter on my machine easily

frank void
stiff cape
#

Killian said #general message

devout notch
#

I used OpenInterpreter with LLama 2 in 3 easy steps.

  1. After checking system requirements and loading my model into LM studio, get server running where you have a working endpoint. Something like: http://localhost:1234/v1/chat/completions
  2. run interpreter --api_base http://localhost:1234/v1/
  3. Test
mystic dagger
#

Hey there am new here tried install open-interpreter via the terminal after installation am getting this error of
"No module named 'model_map', tried installing model_map but unsuccessfully

silver adder
#

@mystic dagger Hey, what OS and python version do you have

frank void
torn moth
#

currently on windows, used to have this running but now that i did a fresh install , it wont run when i do "interpreter" , anyone having this issue?

silver adder