The pgml.transform function will automatically download a custom model from HuggingFace. Here is an example:

SELECT pgml.transform(
  task   => '{
    "task": "text-generation",
    "model": "meta-llama/Meta-Llama-3.1-8B-Instruct",
    "token": "hf_123"
  }'::JSONB,
  inputs => ARRAY['AI is going to'],
  args   => '{
    "max_new_tokens": 100,
    "trust_remote_code": true
  }'::JSONB
);

Replace model with your model, and if the repo is gated, make sure and set a token. Change the task to the relevant task.

The "pgml meta data" tables are for models trained in the database. I don't think you should have to worry about those.

@SilasMarvin , I understand it will work for hugging face, I am looking for inhouse developed customized model to load into pgml. Downloading from hugging face not allowed due to restrictions. I have pytorch .pt file . How i can import into pgml.
Its downloadable model..

@SilasMarvin - Do you have update on this ?

@SilasMarvin @jonatas @kianmeng @solidsnack - - Do you have update on this ?

I haven't worked on this project @viswa-optimusprime. Not sure where did you got my name from 🤔