DS-LLM-TEMPLATE-FINETUNING/scripts/styling/inference.py

#!/usr/bin/env python3
"""
Styling Inference Script
Provides a command-line interface to run the styling inference pipeline
"""

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

def run_inference_with_config(config_path: str, instruction: str, input_text: str = "", max_tokens: int = 128, stream: bool = False):
    """Run inference using a YAML configuration file"""
    print(f"Running styling inference with config: {config_path}")
    print(f"Instruction: {instruction}")
    print(f"Input text: {input_text}")
    print(f"Max tokens: {max_tokens}")
    print(f"Streaming: {stream}")
    
    cmd = [
        "python", "pipelines/styling/inference.py",
        "--config", config_path,
        "--instruction", instruction,
        "--input-text", input_text,
        "--max-tokens", str(max_tokens)
    ]
    
    if stream:
        cmd.append("--stream")
    
    print(f"Running: {' '.join(cmd)}")
    
    try:
        result = subprocess.run(cmd, capture_output=True, text=True, check=True)
        print("✅ Inference completed successfully!")
        print("Output:")
        print(result.stdout)
        return result.stdout
    except subprocess.CalledProcessError as e:
        print(f"❌ Inference failed: {e}")
        print("Error output:")
        print(e.stderr)
        return None

def run_batch_inference_example(config_path: str, input_file: str, output_file: str, instruction: str, max_tokens: int = 128):
    """Run batch inference example"""
    print(f"=== Batch Inference Example ===")
    print(f"Config: {config_path}")
    print(f"Input file: {input_file}")
    print(f"Output file: {output_file}")
    print(f"Instruction: {instruction}")
    print(f"Max tokens: {max_tokens}")
    
    # Check if input file exists
    if not Path(input_file).exists():
        print(f"❌ Input file not found: {input_file}")
        return False
    
    # Read input texts
    with open(input_file, 'r', encoding='utf-8') as f:
        input_texts = [line.strip() for line in f if line.strip()]
    
    print(f"Found {len(input_texts)} texts to process")
    
    # Process each text
    results = []
    for i, text in enumerate(input_texts):
        print(f"\nProcessing text {i+1}/{len(input_texts)}: {text[:50]}...")
        result = run_inference_with_config(config_path, instruction, text, max_tokens)
        if result:
            results.append(result)
        else:
            print(f"❌ Failed to process text {i+1}")
    
    # Save results
    with open(output_file, 'w', encoding='utf-8') as f:
        for i, (input_text, result) in enumerate(zip(input_texts, results)):
            f.write(f"Input {i+1}: {input_text}\n")
            f.write(f"Output {i+1}: {result}\n")
            f.write("-" * 50 + "\n")
    
    print(f"✅ Batch inference completed! Results saved to: {output_file}")
    return True

def show_inference_features():
    """Show the features of the styling inference pipeline"""
    print("=== Styling Inference Pipeline Features ===")
    print()
    print("1. **Model Support**:")
    print("   - Trained LoRA models from training pipeline")
    print("   - Automatic model loading from config")
    print("   - Native Unsloth inference optimization")
    print()
    print("2. **Inference Modes**:")
    print("   - Single text inference with instruction")
    print("   - Batch file processing")
    print("   - Streaming generation")
    print()
    print("3. **Generation Control**:")
    print("   - Configurable max tokens")
    print("   - Same alpaca prompt format as training")
    print("   - Automatic response extraction")
    print()
    print("4. **Usage Examples**:")
    print("   - Single inference: --instruction 'style instruction' --input-text 'your text'")
    print("   - Streaming: add --stream flag")
    print("   - Batch: use batch subcommand with input/output files")

def create_inference_example():
    """Create an inference example using the formal style configuration"""
    print("=== Inference Example: Formal Style Transfer ===")
    print()
    
    # Check if we have the required files
    config_path = "configs/styling/formal.yaml"
    
    if not Path(config_path).exists():
        print(f"❌ Configuration file not found: {config_path}")
        print("   Please run the data processor first to create the configuration")
        return False
    
    print("✅ Found configuration file!")
    print(f"   Config: {config_path}")
    print()
    
    # Example text
    example_text = "Hey, what's up? I'm gonna go grab some food later."
    
    print(f"Example text: {example_text}")
    print()
    
    # Run inference
    success = run_inference_with_config(
        config_path=config_path,
        instruction="Rewrite the following text in a formal style",
        input_text=example_text
    )
    
    if success:
        print("✅ Example inference completed successfully!")
        return True
    else:
        print("❌ Example inference failed!")
        return False

def main():
    """Main inference function"""
    parser = argparse.ArgumentParser(description="Styling Inference Pipeline")
    subparsers = parser.add_subparsers(dest="command", help="Available commands")
    
    # Inference command
    infer_parser = subparsers.add_parser("infer", help="Run single inference")
    infer_parser.add_argument("--config", type=str, required=True, help="Path to YAML configuration file")
    infer_parser.add_argument("--instruction", type=str, required=True, help="Style instruction")
    infer_parser.add_argument("--input-text", type=str, default="", help="Input text to style")
    infer_parser.add_argument("--max-tokens", type=int, default=128, help="Maximum new tokens to generate")
    infer_parser.add_argument("--stream", action="store_true", help="Enable streaming generation")
    
    # Batch inference command
    batch_parser = subparsers.add_parser("batch", help="Run batch inference")
    batch_parser.add_argument("--config", type=str, required=True, help="Path to YAML configuration file")
    batch_parser.add_argument("--input-file", type=str, required=True, help="Input file with texts to style")
    batch_parser.add_argument("--output-file", type=str, required=True, help="Output file for styled texts")
    batch_parser.add_argument("--instruction", type=str, required=True, help="Style instruction")
    batch_parser.add_argument("--max-tokens", type=int, default=128, help="Maximum new tokens to generate")
    
    # Features command
    subparsers.add_parser("features", help="Show available features")
    
    # Example command
    subparsers.add_parser("example", help="Run example inference")
    
    args = parser.parse_args()
    
    if args.command == "infer":
        run_inference_with_config(
            args.config, 
            args.instruction, 
            args.input_text, 
            args.max_tokens, 
            args.stream
        )
    elif args.command == "batch":
        run_batch_inference_example(args.config, args.input_file, args.output_file, args.instruction, args.max_tokens)
    elif args.command == "features":
        show_inference_features()
    elif args.command == "example":
        create_inference_example()
    else:
        parser.print_help()

if __name__ == "__main__":
    main()