refactor: restructure project - scripts at root level

inference.py

@@ -0,0 +1,103 @@
+#!/usr/bin/env python3
+"""
+Inference script for trained LoRA adapter.
+
+Loads the trained model and generates Cyron summaries.
+"""
+
+import argparse
+from pathlib import Path
+
+
+def load_model(model_path):
+    """Load trained LoRA model."""
+    from peft import PeftModel
+    from transformers import AutoModelForCausalLM, AutoTokenizer
+    
+    print(f"Loading base model from {model_path}...")
+    base_model = AutoModelForCausalLM.from_pretrained(
+        model_path,
+        device_map="auto",
+        torch_dtype="auto",
+    )
+    
+    print(f"Loading LoRA weights from {model_path}...")
+    model = PeftModel.from_pretrained(base_model, model_path)
+    tokenizer = AutoTokenizer.from_pretrained(model_path)
+    
+    return model, tokenizer
+
+
+def generate_summary(model, tokenizer, task, files_changed=None, tests_run=False, test_count=0, commit=False, push=False, errors_seen=False):
+    """Generate a Cyron summary for a task."""
+    
+    # Build prompt
+    prompt = f"Task: {task}"
+    if files_changed:
+        prompt += f"\nFiles: {', '.join(files_changed[:3])}"
+    
+    if tests_run:
+        prompt += f"\nTests run: {test_count}"
+    if commit:
+        prompt += "\nCommit: yes"
+    if push:
+        prompt += "\nPush: yes"
+    if errors_seen:
+        prompt += "\nErrors seen: yes"
+    
+    prompt += "\n\nGenerate a Cyron summary:"
+    
+    # Tokenize
+    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
+    
+    # Generate
+    outputs = model.generate(
+        **inputs,
+        max_new_tokens=500,
+        do_sample=True,
+        temperature=0.7,
+        top_p=0.9,
+        repetition_penalty=1.1,
+    )
+    
+    # Decode
+    generated = tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
+    
+    return generated
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Generate Cyron summaries with LoRA")
+    parser.add_argument("--model", type=str, default="output/ornith-35b-lora-cyron", 
+                       help="Path to trained LoRA model")
+    parser.add_argument("--task", type=str, required=True, help="Task description")
+    parser.add_argument("--files", type=str, nargs="*", default=[], help="Files changed")
+    parser.add_argument("--tests-run", action="store_true", help="Tests were run")
+    parser.add_argument("--test-count", type=int, default=0, help="Number of tests")
+    parser.add_argument("--commit", action="store_true", help="Commit was made")
+    parser.add_argument("--push", action="store_true", help="Code was pushed")
+    parser.add_argument("--errors", action="store_true", help="Errors were seen")
+    args = parser.parse_args()
+    
+    model, tokenizer = load_model(args.model)
+    
+    summary = generate_summary(
+        model,
+        tokenizer,
+        task=args.task,
+        files_changed=args.files,
+        tests_run=args.tests_run,
+        test_count=args.test_count,
+        commit=args.commit,
+        push=args.push,
+        errors_seen=args.errors,
+    )
+    
+    print("\nGenerated Summary:")
+    print("=" * 70)
+    print(summary)
+    print("=" * 70)
+
+
+if __name__ == "__main__":
+    main()