Upload root_scripts/test_single_q.py with huggingface_hub

root_scripts/test_single_q.py

@@ -0,0 +1,67 @@
+
+import torch
+from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
+from PIL import Image
+import os, re
+
+MODEL_PATH = "/workspace/rl4phyx/models/Qwen2.5-VL-3B-Instruct"
+IMAGE_PATH = "/workspace/rl4phyx/RL4Phyx/ZeroSearch/One-Shot-RLVR/data/train/physics_vlm/mechanics/100.png"
+
+PROMPT = """Look at the image and answer the physics question.
+
+A patient with a dislocated shoulder is put into a traction apparatus as shown in figure. The pulls A and B have equal magnitudes and must combine to produce an outward traction force of 12.8 N on the patient's arm.
+
+Question: How large should these pulls be?
+
+Options:
+A:"7.55N", B:"5.55N", C:"7.65N", D:"6.65N"
+
+Please analyze the problem step by step and provide your final answer within \\boxed{}."""
+
+print("Loading model...", flush=True)
+model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
+    MODEL_PATH, torch_dtype=torch.bfloat16, device_map="auto",
+    attn_implementation="sdpa"
+)
+processor = AutoProcessor.from_pretrained(MODEL_PATH)
+print("Model loaded!", flush=True)
+
+image = Image.open(IMAGE_PATH).convert("RGB")
+print(f"Image loaded: {image.size}", flush=True)
+
+results = []
+for trial in range(10):
+    messages = [{"role": "user", "content": [
+        {"type": "image", "image": image},
+        {"type": "text", "text": PROMPT}
+    ]}]
+    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+    inputs = processor(text=[text], images=[image], return_tensors="pt", padding=True).to(model.device)
+    
+    with torch.no_grad():
+        out = model.generate(**inputs, max_new_tokens=1024, temperature=0.7, do_sample=True, top_p=0.9)
+    
+    response = processor.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
+    
+    boxed = re.search(r'\\boxed\{(.+?)\}', response)
+    answer = boxed.group(1) if boxed else "N/A"
+    
+    # Check if correct (GT is A / 7.55N)
+    correct = "A" in answer or "7.55" in answer
+    results.append({"trial": trial+1, "answer": answer, "correct": correct})
+    
+    print(f"\n=== Trial {trial+1}/10 ===")
+    print(f"Extracted answer: {answer}")
+    print(f"Correct: {correct}")
+    print(f"Full response:\n{response}")
+    print(f"{'='*50}", flush=True)
+
+print(f"\n\n{'='*60}")
+print(f"SUMMARY: GT = A (7.55N)")
+print(f"{'='*60}")
+correct_count = sum(1 for r in results if r["correct"])
+for r in results:
+    mark = "OK" if r["correct"] else "WRONG"
+    print(f"  Trial {r['trial']:2d}: {r['answer']:20s} [{mark}]")
+print(f"\nAccuracy: {correct_count}/10 ({correct_count*10}%)")
+print(f"{'='*60}", flush=True)