I cannot load "gemma_2b_en" model from my own notebook environment

[davidshen84]

Hi,

When I try keras_nlp.models.GemmaCausalLM.from_preset("gemma_2b_en"), I got 403 error.

Downloading from https://www.kaggle.com/api/v1/models/keras/gemma/keras/gemma_2b_en/2/download/config.json...
---------------------------------------------------------------------------
HTTPError                                 Traceback (most recent call last)
File /usr/local/lib/python3.11/dist-packages/kagglehub/exceptions.py:49, in kaggle_api_raise_for_status(response, resource_handle)
     48 try:
---> 49     response.raise_for_status()
     50 except requests.HTTPError as e:

File /usr/local/lib/python3.11/dist-packages/requests/models.py:1021, in Response.raise_for_status(self)
   1020 if http_error_msg:
-> 1021     raise HTTPError(http_error_msg, response=self)

HTTPError: 403 Client Error: Forbidden for url: https://www.kaggle.com/api/v1/models/keras/gemma/keras/gemma_2b_en/2/download/config.json

The above exception was the direct cause of the following exception:
...

Looks like I need to login to Kaggle first. But I am running the code from my own jupyter notebook environment, and there's no instruction about how to do the login.

I understand Kaggle wants users to login to access their resources. But for keras_nlp, should it by default use a public resource?

[davidshen84]

OK, after digging the Internet for a while, I found the solution.

1. Create a kaggle account
2. Go to https://www.kaggle.com/models/google/gemma, then request access
3. Go to https://www.kaggle.com/settings, and "Create New Token".
   This will download a json file to your system. It contains your kaggle account credentials
4. Put the kaggle.json file to ~/.kaggle/kaggle.json (or %userprofile%\.kaggle\kaggle.json on Windows)

Then you should be able to download the model.