backup

app.py

@@ -2,7 +2,7 @@ import os
 import gradio as gr
 import torch
 from PIL import Image
-from transformers import AutoConfig, AutoModelForCausalLM, AutoProcessor
+from transformers import MllamaForConditionalGeneration, AutoProcessor
 from peft import PeftModel
 from huggingface_hub import login
 import spaces
@@ -31,18 +31,30 @@ login(token=os.environ["HF_TOKEN"], add_to_git_credential=True)
 base_model_path = "taesiri/BugsBunny-LLama-3.2-11B-Vision-Instruct-Medium-FullModel"
 # lora_weights_path = "taesiri/BugsBunny-LLama-3.2-11B-Vision-Base-Medium-LoRA"
 
-try:
-    processor = AutoProcessor.from_pretrained(base_model_path, trust_remote_code=True)
-    model = AutoModelForCausalLM.from_pretrained(
-        base_model_path,
-        torch_dtype=torch.bfloat16,
-        device_map="cuda",
-        trust_remote_code=True,
-    )
-    model.tie_weights()
-except Exception as e:
-    print(f"Error loading model: {e}")
-    raise
+# processor = AutoProcessor.from_pretrained(base_model_path)
+# model = MllamaForConditionalGeneration.from_pretrained(
+#     base_model_path,
+#     torch_dtype=torch.bfloat16,
+#     device_map="cuda",
+# )
+
+from transformers import AutoModelForCausalLM, AutoProcessor, LlamaTokenizer
+import torch
+
+model_path = "taesiri/BugsBunny-LLama-3.2-11B-Vision-Instruct-Medium-FullModel"
+
+# Load the processor (handles both text and vision inputs)
+processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True)
+
+# Load the model
+model = AutoModelForCausalLM.from_pretrained(
+    model_path, torch_dtype=torch.bfloat16, device_map="cuda", trust_remote_code=True
+)
+
+# If you specifically need the tokenizer
+tokenizer = LlamaTokenizer.from_pretrained(model_path)
+
+model.tie_weights()
 
 # model = PeftModel.from_pretrained(model, lora_weights_path)