ValueError: `params` cannot be accessed from model when the model is created with `_do_init=False`.

[jchauhan]

Describe the bug
Getting the following error while running llama model after training using EasyDel and converting to Hugging face.

python serve_llama_tpu_easydel.py 
Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████████████| 3/3 [00:03<00:00,  1.14s/it]
Traceback (most recent call last):
  File "/home/XXX/research/EasyDeL/serve_llama_tpu_easydel.py", line 26, in <module>
    params=model.params,
  File "/home/XXX/research/EasyDeL/.venv/lib/python3.10/site-packages/transformers/modeling_flax_utils.py", line 271, in params
    raise ValueError(
ValueError: `params` cannot be accessed from model when the model is created with `_do_init=False`. You must call `init_weights` manually and store the params outside of the model and pass it explicitly where needed.

To Reproduce

• Using Kaggle to traiin llama-2-7b-chat-hf model on certain data
• convert and publish to Hugging face
• load the model using EasyDel example

[jchauhan]

Possible Solution,

Take _do_init as parameter. currently it is hardcoded

lib/python/EasyDel/modules/auto_easydel_model.py

        ed_model = module(
            config=cfg,
            _do_init=False,
            dtype=dtype,
            param_dtype=param_dtype,
            precision=precision,
            input_shape=input_shape
        )

Is there any other better solution while publishing model to avoid the error?

[erfanzar]

you don't need to re-init parameters in case of converting or serving or even training this is happing cause you not using library correctly can I have access to your code