Update retrain_from_feedback.py

retrain_from_feedback.py

@@ -1,41 +1,70 @@
-import csv
+# retrain_from_feedback.py
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import pandas as pd
+from transformers import AutoTokenizer
+from evo_architecture import mutate_genome, log_genome, default_config
+from evo_model import EvoTransformerV22
 import os
-from datetime import datetime
-from retrain_from_feedback import train_evo
-
-# 🔁 Main entry point for feedback-triggered retraining
-def retrain_from_feedback(feedback_log):
-    # ✅ Check if feedback is present
-    if not feedback_log:
-        return "⚠️ No feedback data to retrain from."
-
-    # 📝 Write feedback to CSV
-    try:
-        os.makedirs("feedback", exist_ok=True)
-        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-        filepath = f"feedback/feedback_log.csv"  # also usable for loading
-        with open(filepath, "w", newline="", encoding="utf-8") as f:
-            writer = csv.writer(f)
-            writer.writerow([
-                "question", "option1", "option2", "answer",
-                "confidence", "reasoning", "context",
-                "user_preference", "evo_was_correct", "evo_output"
-            ])
-            for row in feedback_log:
-                question, option1, option2, answer, confidence, reasoning, context = row
-
-                # Simulate Evo being preferred (you can modify this logic later)
-                writer.writerow([
-                    question, option1, option2, answer,
-                    confidence, reasoning, context,
-                    "evo", "yes", answer
-                ])
-    except Exception as e:
-        return f"❌ Failed to save feedback: {str(e)}"
-
-    # 🔁 Trigger training
-    try:
-        train_evo()  # This uses the latest feedback_log.csv
-        return "✅ Evo retrained and weights saved."
-    except Exception as e:
-        return f"❌ Evo training failed: {str(e)}"
+
+MODEL_PATH = "evo_hellaswag.pt"
+CSV_PATH = "feedback_log.csv"
+
+def train_evo():
+    if not os.path.exists(CSV_PATH):
+        print("⚠️ No feedback_log.csv file found.")
+        return
+
+    df = pd.read_csv(CSV_PATH)
+    if df.empty:
+        print("⚠️ feedback_log.csv is empty.")
+        return
+
+    # Step 1: Evolve new architecture
+    base_config = default_config()
+    evolved_config = mutate_genome(base_config)
+    print("🧬 New mutated config:", evolved_config)
+
+    # Step 2: Initialize model with evolved config
+    model = EvoTransformerV22(
+        num_layers=evolved_config["num_layers"],
+        num_heads=evolved_config["num_heads"],
+        ffn_dim=evolved_config["ffn_dim"],
+        memory_enabled=evolved_config["memory_enabled"]
+    )
+
+    tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
+    optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)
+    model.train()
+
+    # Step 3: Train on feedback
+    total_loss = 0.0
+    for _, row in df.iterrows():
+        question = row["question"]
+        opt1 = row["option1"]
+        opt2 = row["option2"]
+        answer = row["answer"]
+
+        label = torch.tensor([1.0 if answer.strip() == opt2.strip() else 0.0])
+
+        input_text = f"{question} [SEP] {opt2 if label.item() == 1 else opt1}"
+        encoded = tokenizer(input_text, return_tensors="pt", padding="max_length", truncation=True, max_length=128)
+
+        logits = model(encoded["input_ids"])
+        loss = F.binary_cross_entropy_with_logits(logits.squeeze(), label)
+        loss.backward()
+        optimizer.step()
+        optimizer.zero_grad()
+        total_loss += loss.item()
+
+    # Step 4: Save new model weights
+    torch.save(model.state_dict(), MODEL_PATH)
+    print("✅ Evo model retrained and saved.")
+
+    # Step 5: Log genome and score (loss as proxy)
+    avg_loss = total_loss / len(df)
+    score = 1.0 - avg_loss  # Use (1 - loss) as crude fitness
+    log_genome(evolved_config, performance=round(score, 4))
+    print("🧬 Genome logged with score:", round(score, 4))