Update app.py

app.py

@@ -6,8 +6,11 @@ import torch
 import spaces                # <-- NEW: import spaces for ZeroGPU
 import gradio as gr
 from PIL import Image
-from transformers import AutoProcessor, AutoModelForVision2Seq
+from transformers import AutoProcessor, AutoModelForVision2Seq, AutoConfig
 from peft import PeftModel
+import transformers, accelerate, torch, numpy as np
+
+print("VERSIONS:", transformers.__version__, accelerate.__version__, torch.__version__, np.__version__)
 
 # 0. Environment tweaks for Accelerate (unchanged)
 os.environ["ACCELERATE_USE_SLOW_RETRIEVAL"] = "true"
@@ -22,9 +25,11 @@ dtype = torch.float16        # <-- use float16 for GPU
 # 2. Load processor
 processor = AutoProcessor.from_pretrained(BASE_MODEL, trust_remote_code=True)
 
+cfg = AutoConfig.from_pretrained(BASE_MODEL, trust_remote_code=True)
 # 3. Load base model on CPU; stream shards to save RAM
 base = AutoModelForVision2Seq.from_pretrained(
     BASE_MODEL,
+    config=cfg,
     torch_dtype=dtype,
     low_cpu_mem_usage=True,
     device_map={"": "cpu"},