Fix LLAVA example on CPU

[jenniew]

Fix LLAVA example issue on CPU.
Related issue: #11230

[JinBridger]

On linux it works fine. But on Windows, it throws following error:

platform: xiaoxin03
ipex-llm version: 2.1.0b20240613

(jinqiao-llm-0614) D:\jinqiao>python generate.py --repo-id-or-model-path D:/llm-models/llava-v1.5-7b --image-path-or-url https://llava-vl.github.io/static/images/monalisa.jpg
C:\Users\arda\miniforge3\envs\jinqiao-llm-0614\Lib\site-packages\bitsandbytes\cextension.py:34: UserWarning: The installed version of bitsandbytes was compiled without GPU support. 8-bit optimizers, 8-bit multiplication, and GPU quantization are unavailable.
  warn("The installed version of bitsandbytes was compiled without GPU support. "
'NoneType' object has no attribute 'cadam32bit_grad_fp32'
You are using a model of type llava to instantiate a model of type llava_llama. This is not supported for all configurations of models and can yield errors.
C:\Users\arda\miniforge3\envs\jinqiao-llm-0614\Lib\site-packages\huggingface_hub\file_download.py:1132: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use `force_download=True`.
  warnings.warn(
Traceback (most recent call last):
  File "D:\jinqiao\generate.py", line 292, in <module>
    tokenizer, model, image_processor, _ = load_pretrained_model(model_path=model_path,
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\jinqiao\generate.py", line 154, in load_pretrained_model
    model = LlavaLlamaForCausalLM.from_pretrained(
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\arda\miniforge3\envs\jinqiao-llm-0614\Lib\site-packages\transformers\modeling_utils.py", line 3706, in from_pretrained
    ) = cls._load_pretrained_model(
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\arda\miniforge3\envs\jinqiao-llm-0614\Lib\site-packages\transformers\modeling_utils.py", line 3835, in _load_pretrained_model
    raise ValueError(
ValueError: The current `device_map` had weights offloaded to the disk. Please provide an `offload_folder` for them. Alternatively, make sure you have `safetensors` installed if the model you are using offers the weights in this format.

By changing here to kwargs = {} it works well again on Windows.

ipex-llm/python/llm/example/CPU/PyTorch-Models/Model/llava/generate.py

Line 66 in f5ef940

kwargs = {"device_map": device_map}