continuedev / continue

⏩ Continue is the leading open-source AI code assistant. You can connect any models and any context to build custom autocomplete and chat experiences inside VS Code and JetBrains
https://docs.continue.dev/
Apache License 2.0
18.7k stars 1.58k forks source link

Also add NVIDIA as a provider #2517

Open lakshits11 opened 2 weeks ago

lakshits11 commented 2 weeks ago

Validations

Problem

Please add NVIDIA as a provider

Solution

No response

Patrick-Erichsen commented 2 weeks ago

Hi @lakshits11 , we had some docs for this but it looks like they accidentally got removed in a recent refactor.

We do support NVIDIA as an OpenAI compatible provider: https://docs.continue.dev/customize/model-providers/openai#openai-compatible-servers--apis

And for embeddings: https://github.com/continuedev/continue/pull/2233/files#diff-f5ee2eb28e79019bdf138776df40073fefe4c5c12c03877cb1befdda47ef6418R71-R85

lakshits11 commented 2 weeks ago

For some reason I am not able to get nvidia to work. I m using below configuration

{
      "title": "mistral",
      "provider": "nvidia",
      "model": "mistralai/mistral-large-2-instruct,
      "apiKey": "[API_KEY]",
      "apiBase": "https://integrate.api.nvidia.com/v1",
}

Am i missing something?

lakshits11 commented 2 weeks ago

Also, the docs have a reference for kindo https://docs.continue.dev/customize/model-providers/more/kindo

But I dont find their name in provider autocomplete suggestions

robertpiosik commented 2 weeks ago

@lakshits11 change provider to openai, leave apiBase

{
      "apiKey": "",
      "title": "Llama-3.1-Nemotron-70B-Instruct",
      "model": "nvidia/llama-3.1-nemotron-70b-instruct",
      "provider": "openai",
      "apiBase": "https://integrate.api.nvidia.com/v1"
    }
lakshits11 commented 2 weeks ago

Thank you this was exactly the model I was gonna try

lakshits11 commented 1 week ago

hy it still doesnt work :( i got my api key but still it does not print any output but the credit points are getting deducted

EDIT : adding this got it to work "useLegacyCompletionsEndpoint": true,

Also a humble reuest, I think your docs and extension have a huge mismatch. docs dont have nvidia, but the config.json autocomplete shows nvidia as provider and on hovering it, it shows description about vLLM

{161C8E56-E918-49C7-93DD-C220B928DDD2}

pls update the docs. Also I will be happy to do so. Thanks, please consider my request

robertpiosik commented 1 week ago

Yeah, the latest update broke it. Thanks for your post, I had no idea why it suddenly stopped working...