Update MiniCPM_V_26 GPU example with save & load

python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-V-2_6/README.md

@@ -114,12 +114,21 @@ set SYCL_CACHE_PERSISTENT=1
   ```
   python ./chat.py --prompt 'What is in the image?' --stream
   ```
+- save model with low-bit optimization (if `LOWBIT_MODEL_PATH` does not exist)
+  ```
+  python ./chat.py --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --lowbit-path LOWBIT_MODEL_PATH --prompt 'What is in the image?'
+  ```
+- chat with saved model with low-bit optimization (if `LOWBIT_MODEL_PATH` exists):
+  ```
+  python ./chat.py --lowbit-path LOWBIT_MODEL_PATH --prompt 'What is in the image?'
+  ```
 
 > [!TIP]
 > For chatting in streaming mode, it is recommended to set the environment variable `PYTHONUNBUFFERED=1`.
 
 Arguments info:
 - `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the MiniCPM-V-2_6 (e.g. `openbmb/MiniCPM-V-2_6`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'openbmb/MiniCPM-V-2_6'`.
+- `--lowbit-path LOWBIT_MODEL_PATH`: argument defining the path to save/load the model with IPEX-LLM low-bit optimization. If it is an empty string, the original pretrained model specified by `REPO_ID_OR_MODEL_PATH` will be loaded. If it is an existing path, the saved model with low-bit optimization in `LOWBIT_MODEL_PATH` will be loaded. If it is a non-existing path, the original pretrained model specified by `REPO_ID_OR_MODEL_PATH` will be loaded, and the optimized low-bit model will be saved into `LOWBIT_MODEL_PATH`. It is default to be `''`, i.e. an empty string.
 - `--image-url-or-path IMAGE_URL_OR_PATH`: argument defining the image to be infered. It is default to be `'http://farm6.staticflickr.com/5268/5602445367_3504763978_z.jpg'`.
 - `--prompt PROMPT`: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be `'What is in the image?'`.
 - `--stream`: flag to chat in streaming mode

python/llm/example/GPU/HuggingFace/Multimodal/MiniCPM-V-2_6/chat.py

@@ -22,14 +22,21 @@
 import torch
 from PIL import Image
 from ipex_llm.transformers import AutoModel
-from transformers import AutoTokenizer
+from transformers import AutoTokenizer, AutoProcessor
 
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(description='Predict Tokens using `chat()` API for openbmb/MiniCPM-V-2_6 model')
     parser.add_argument('--repo-id-or-model-path', type=str, default="openbmb/MiniCPM-V-2_6",
                         help='The huggingface repo id for the openbmb/MiniCPM-V-2_6 model to be downloaded'
                              ', or the path to the huggingface checkpoint folder')
+    parser.add_argument("--lowbit-path", type=str,
+        default="",
+        help="The path to the saved model folder with IPEX-LLM low-bit optimization. "
+             "Leave it blank if you want to load from the original model. "
+             "If the path does not exist, model with low-bit optimization will be saved there."
+             "Otherwise, model with low-bit optimization will be loaded from the path.",
+    )
     parser.add_argument('--image-url-or-path', type=str,
                         default='http://farm6.staticflickr.com/5268/5602445367_3504763978_z.jpg',
                         help='The URL or path to the image to infer')
@@ -41,22 +48,43 @@
     args = parser.parse_args()
     model_path = args.repo_id_or_model_path
     image_path = args.image_url_or_path
+
+    lowbit_path = args.lowbit_path
+
+    if not lowbit_path or not os.path.exists(lowbit_path):
+        # Load model in 4 bit,
+        # which convert the relevant layers in the model into INT4 format
+        # When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the from_pretrained function.
+        # This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU.
+        model = AutoModel.from_pretrained(model_path, 
+                                        load_in_low_bit="sym_int4",
+                                        optimize_model=True,
+                                        trust_remote_code=True,
+                                        use_cache=True,
+                                        modules_to_not_convert=["vpm", "resampler"])
+
+        tokenizer = AutoTokenizer.from_pretrained(model_path,
+                                                  trust_remote_code=True)
+    else:
+        model = AutoModel.load_low_bit(lowbit_path, 
+                                       optimize_model=True,
+                                       trust_remote_code=True,
+                                       use_cache=True,
+                                       modules_to_not_convert=["vpm", "resampler"])
+        tokenizer = AutoTokenizer.from_pretrained(lowbit_path,
+                                                  trust_remote_code=True)
 
-    # Load model in 4 bit,
-    # which convert the relevant layers in the model into INT4 format
-    # When running LLMs on Intel iGPUs for Windows users, we recommend setting `cpu_embedding=True` in the from_pretrained function.
-    # This will allow the memory-intensive embedding layer to utilize the CPU instead of iGPU.
-    model = AutoModel.from_pretrained(model_path, 
-                                      load_in_low_bit="sym_int4",
-                                      optimize_model=True,
-                                      trust_remote_code=True,
-                                      use_cache=True,
-                                      modules_to_not_convert=["vpm", "resampler"])
-    model = model.half().to('xpu')
-    tokenizer = AutoTokenizer.from_pretrained(model_path,
-                                              trust_remote_code=True)
     model.eval()
 
+    if lowbit_path and not os.path.exists(lowbit_path):
+        processor = AutoProcessor.from_pretrained(model_path,
+                                                trust_remote_code=True)
+        model.save_low_bit(lowbit_path)
+        tokenizer.save_pretrained(lowbit_path)
+        processor.save_pretrained(lowbit_path)
+
+    model = model.half().to('xpu')
+
     query = args.prompt
     if os.path.exists(image_path):
        image = Image.open(image_path).convert('RGB')