VTimeLLM/scripts/train/debug_deepspeed.py

#!/usr/bin/env python3
"""
Debug wrapper for DeepSpeed training
This script allows debugging the training process step by step
"""

import os
import sys
import subprocess
import argparse
from pathlib import Path

def setup_environment():
    """Setup environment variables for debugging"""
    env_vars = {
        'RANK': '1',
        'MASTER_PORT': '29571',
        'LOCAL_BATCH_SIZE': '2',
        'GRADIENT_ACCUMULATION_STEPS': '4',
        'TRANSFORMERS_OFFLINE': '1',
        'WANDB_PROJECT': 'vtimellm',
        'MODEL_VERSION': 'vicuna-v1-5-7b',
        'OUTPUT_DIR': './outputs/',
        'STAGE4': './outputs/vtimellm-vicuna-v1-5-7b-activitynet-stage4',
        'PYTHONPATH': f"{os.getcwd()}:{os.environ.get('PYTHONPATH', '')}",
        'CUDA_VISIBLE_DEVICES': '1',
        'TORCH_USE_CUDA_DSA': '1',
        'TRANSFORMERS_VERBOSITY': 'info',
        'TOKENIZERS_PARALLELISM': 'false'
    }
    
    for key, value in env_vars.items():
        os.environ[key] = value
    
    return env_vars

def check_required_files():
    """Check if all required files exist"""
    required_files = [
        "./checkpoints/vicuna-7b-v1.5",
        "./data/activitynet/mdpo-train.json",
        "./data/activitynet/videos/train",
        "./data/activitynet/clipvitl14-vtimellm.pth",
        "./checkpoints/vtimellm-vicuna-v1-5-7b-stage1/mm_projector.bin",
        "./checkpoints/vtimellm-vicuna-v1-5-7b-stage2",
        "./checkpoints/vtimellm-vicuna-v1-5-7b-stage3",
        "./checkpoints/vtimellm-vicuna-v1-5-7b-activitynet-stage4",
        "./scripts/zero2.json"
    ]
    
    missing_files = []
    for file_path in required_files:
        if not Path(file_path).exists():
            missing_files.append(file_path)
        else:
            print(f"✓ Found: {file_path}")
    
    if missing_files:
        print("✗ Missing files:")
        for file_path in missing_files:
            print(f"  {file_path}")
        return False
    
    return True

def check_gpu():
    """Check GPU availability"""
    try:
        result = subprocess.run(['nvidia-smi', '--query-gpu=name,memory.total,memory.free', '--format=csv,noheader,nounits'], 
                              capture_output=True, text=True)
        if result.returncode == 0:
            print("=== GPU Information ===")
            print(result.stdout)
            print("========================")
            return True
        else:
            print("Warning: nvidia-smi not available or no GPU found")
            return False
    except FileNotFoundError:
        print("Warning: nvidia-smi not found")
        return False

def create_output_dir():
    """Create output directory"""
    output_dir = "./outputs/vtimellm-vicuna-v1-5-7b-activitynet-stage5"
    Path(output_dir).mkdir(parents=True, exist_ok=True)
    print(f"Created output directory: {output_dir}")
    return output_dir

def run_training():
    """Run the training with DeepSpeed"""
    env_vars = setup_environment()
    
    print("=== Debug Environment Setup ===")
    for key, value in env_vars.items():
        print(f"{key}: {value}")
    print("================================")
    
    print("=== Checking Required Files ===")
    if not check_required_files():
        print("Error: Missing required files. Please check the paths.")
        return False
    
    print("=== Checking GPU ===")
    check_gpu()
    
    print("=== Creating Output Directory ===")
    create_output_dir()
    
    # DeepSpeed command
    cmd = [
        "deepspeed",
        "--include", f"localhost:{env_vars['RANK']}",
        "--master_port", env_vars['MASTER_PORT'],
        "vtimellm/train/train_dpo_mem.py",
        "--deepspeed", "./scripts/zero2.json",
        "--lora_enable", "True",
        "--lora_r", "8",
        "--lora_alpha", "128",
        "--training_stage", "3",
        "--finetuning", "True",
        "--model_name_or_path", "./checkpoints/vicuna-7b-v1.5",
        "--version", "v1",
        "--data_path", "./data/activitynet/mdpo-train.json",
        "--data_folder", "./data/activitynet/videos/train",
        "--feat_folder", "./data/activitynet/clipvitl14-vtimellm.pth",
        "--pretrain_mm_mlp_adapter", "./checkpoints/vtimellm-vicuna-v1-5-7b-stage1/mm_projector.bin",
        "--stage2_path", "./checkpoints/vtimellm-vicuna-v1-5-7b-stage2",
        "--stage3_path", "./checkpoints/vtimellm-vicuna-v1-5-7b-stage3",
        "--stage4_path", "checkpoints/vtimellm-vicuna-v1-5-7b-activitynet-stage4",
        "--output_dir", "./outputs/vtimellm-vicuna-v1-5-7b-activitynet-stage5",
        "--bf16", "True",
        "--max_steps", "100",
        "--per_device_train_batch_size", env_vars['LOCAL_BATCH_SIZE'],
        "--gradient_accumulation_steps", env_vars['GRADIENT_ACCUMULATION_STEPS'],
        "--evaluation_strategy", "no",
        "--save_strategy", "no",
        "--save_steps", "50000",
        "--save_total_limit", "10",
        "--learning_rate", "1e-6",
        "--freeze_mm_mlp_adapter", "True",
        "--weight_decay", "0.",
        "--warmup_ratio", "0.1",
        "--lr_scheduler_type", "cosine",
        "--logging_steps", "1",
        "--tf32", "True",
        "--model_max_length", "2048",
        "--gradient_checkpointing", "True",
        "--dataloader_num_workers", "4",
        "--lazy_preprocess", "True",
        "--report_to", "none",
        "--run_name", "vtimellm-vicuna-v1-5-7b-activitynet-stage5",
        "--gamma", "0.0",
        "--beta", "0.5",
        "--dpo_alpha", "1.0",
        "--train4dpo"
    ]
    
    print("=== Starting Debug Training ===")
    print(f"Command: {' '.join(cmd)}")
    print("================================")
    
    try:
        # Run the command
        result = subprocess.run(cmd, check=True)
        print("=== Training Completed Successfully ===")
        return True
    except subprocess.CalledProcessError as e:
        print(f"=== Training Failed with Error Code: {e.returncode} ===")
        return False
    except KeyboardInterrupt:
        print("=== Training Interrupted by User ===")
        return False

def main():
    parser = argparse.ArgumentParser(description="Debug wrapper for MDPO training")
    parser.add_argument("--check-only", action="store_true", help="Only check environment and files")
    parser.add_argument("--dry-run", action="store_true", help="Show command without executing")
    
    args = parser.parse_args()
    
    if args.check_only:
        setup_environment()
        check_required_files()
        check_gpu()
        create_output_dir()
        return
    
    if args.dry_run:
        env_vars = setup_environment()
        cmd = [
            "deepspeed",
            "--include", f"localhost:{env_vars['RANK']}",
            "--master_port", env_vars['MASTER_PORT'],
            "vtimellm/train/train_dpo_mem.py",
            # ... rest of arguments
        ]
        print("Command that would be executed:")
        print(" ".join(cmd))
        return
    
    success = run_training()
    sys.exit(0 if success else 1)

if __name__ == "__main__":
    main()