Delete files batch_inference.py with huggingface_hub

batch_inference.py

@@ -1,147 +0,0 @@
-"""CAD-Coder Batch Inference Script using vLLM"""
-import argparse
-import json
-import os
-import re
-import csv
-from tqdm import tqdm
-from vllm import LLM, SamplingParams
-
-
-def parse_args():
-    parser = argparse.ArgumentParser(description='CAD-Coder Batch Inference')
-    parser.add_argument('--model_path', type=str, default="gudo7208/CAD-Coder",
-                        help='Model path or HuggingFace Hub name')
-    parser.add_argument('--data_path', type=str, required=True,
-                        help='Test data path (JSON or JSONL format)')
-    parser.add_argument('--output_dir', type=str, default="./output",
-                        help='Output directory')
-    parser.add_argument('--tensor_parallel_size', type=int, default=1,
-                        help='Number of GPUs for tensor parallelism')
-    parser.add_argument('--temperature', type=float, default=0.7,
-                        help='Sampling temperature')
-    parser.add_argument('--top_p', type=float, default=0.7,
-                        help='Top-p sampling')
-    parser.add_argument('--max_tokens', type=int, default=4096,
-                        help='Maximum tokens to generate')
-    return parser.parse_args()
-
-
-def load_data(data_path):
-    """Load test data from JSON or JSONL format"""
-    if data_path.endswith('.json'):
-        with open(data_path, 'r', encoding='utf-8') as f:
-            return json.load(f)
-    elif data_path.endswith('.jsonl'):
-        data = []
-        with open(data_path, 'r', encoding='utf-8') as f:
-            for line in f:
-                data.append(json.loads(line))
-        return data
-    else:
-        raise ValueError(f"Unsupported file format: {data_path}")
-
-
-def build_prompt(user_content):
-    """Build chat format prompt for Qwen model"""
-    return (f"<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n"
-            f"<|im_start|>user\n{user_content}<|im_end|>\n"
-            f"<|im_start|>assistant\n")
-
-
-def extract_code(response):
-    """Extract Python code from response, supporting multiple formats"""
-    # Try to extract from \boxed{```python ... ```}
-    boxed_match = re.search(r'\\boxed\{```python\n(.*?)```\}', response, re.DOTALL)
-    if boxed_match:
-        return boxed_match.group(1).strip()
-
-    # Try to extract from ```python ... ```
-    code_match = re.search(r'```python\n(.*?)```', response, re.DOTALL)
-    if code_match:
-        return code_match.group(1).strip()
-
-    # Return raw response if no code block found
-    return response
-
-
-def main():
-    args = parse_args()
-
-    # Initialize vLLM model
-    print(f"Loading model from {args.model_path}...")
-    llm = LLM(
-        model=args.model_path,
-        tensor_parallel_size=args.tensor_parallel_size,
-        trust_remote_code=True,
-        dtype="float16",
-        gpu_memory_utilization=0.9
-    )
-
-    # Set sampling parameters
-    sampling_params = SamplingParams(
-        temperature=args.temperature,
-        top_p=args.top_p,
-        max_tokens=args.max_tokens
-    )
-
-    # Load test data
-    print(f"Loading data from {args.data_path}...")
-    test_data = load_data(args.data_path)
-    # test_data = test_data[:50]  # 测试时只取50条，正式运行时注释此行
-    print(f"Loaded {len(test_data)} samples")
-
-    # Prepare prompts and file names
-    prompts = []
-    file_names = []
-
-    for item in test_data:
-        # Extract file name from model_path
-        model_path = item.get('model_path', f"sample_{len(file_names)}")
-        file_name = os.path.basename(model_path).replace('.pth', '')
-        file_names.append(file_name)
-
-        # Extract user content and build prompt
-        user_content = item['messages'][0]['content']
-        prompts.append(build_prompt(user_content))
-
-    # Generate with vLLM
-    print("Generating...")
-    outputs = llm.generate(prompts, sampling_params)
-
-    # Create output directory
-    os.makedirs(args.output_dir, exist_ok=True)
-
-    # Prepare CSV data
-    csv_data = []
-
-    # Save results
-    for file_name, output in tqdm(zip(file_names, outputs), total=len(outputs), desc="Saving"):
-        response = output.outputs[0].text
-        python_code = extract_code(response)
-
-        # Save Python code
-        code_path = os.path.join(args.output_dir, f"{file_name}.py")
-        with open(code_path, "w", encoding="utf-8") as f:
-            f.write(python_code)
-
-        # Save full response
-        txt_path = os.path.join(args.output_dir, f"{file_name}.txt")
-        with open(txt_path, "w", encoding="utf-8") as f:
-            f.write(response)
-
-        # Add to CSV data
-        csv_data.append([file_name, response, python_code])
-
-    # Save CSV summary
-    csv_path = os.path.join(args.output_dir, "results.csv")
-    with open(csv_path, "w", encoding="utf-8", newline="") as f:
-        writer = csv.writer(f)
-        writer.writerow(["id", "responses", "code"])
-        writer.writerows(csv_data)
-
-    print(f"Results saved to {args.output_dir}")
-
-
-if __name__ == "__main__":
-    main()