Update some files

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import torch
 import math
+import os
 from transformers import AutoTokenizer, AutoModel, AutoProcessor
 from decord import VideoReader, cpu
 from PIL import Image
@@ -17,10 +18,18 @@ transform = Compose([
 ])
 
 # === 模型加载 ===
+PERSISTENT_DIR = "/data/internvl3_model"  # 持久路径
 MODEL_NAME = "OpenGVLab/InternVL3-14B"
 
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True)
-processor = AutoProcessor.from_pretrained(MODEL_NAME, trust_remote_code=True)
+# 如果第一次运行：下载模型并缓存到 /data
+if not os.path.exists(PERSISTENT_DIR):
+    print("Downloading model to persistent storage...")
+    from transformers import snapshot_download
+    snapshot_download(repo_id=MODEL_NAME, local_dir=PERSISTENT_DIR, trust_remote_code=True)
+
+# 模型加载（从本地）
+tokenizer = AutoTokenizer.from_pretrained(PERSISTENT_DIR, trust_remote_code=True)
+processor = AutoProcessor.from_pretrained(PERSISTENT_DIR, trust_remote_code=True)
 
 def split_model(model_path):
     from transformers import AutoConfig
@@ -47,10 +56,10 @@ def split_model(model_path):
     device_map[f'language_model.model.layers.{num_layers - 1}'] = 0
     return device_map
 
-device_map = split_model(MODEL_NAME)
+device_map = split_model(PERSISTENT_DIR)
 
 model = AutoModel.from_pretrained(
-    MODEL_NAME,
+    PERSISTENT_DIR,
     torch_dtype=torch.bfloat16,
     low_cpu_mem_usage=True,
     use_flash_attn=True,