import torch
import pandas as pd
from torch.utils.data import Dataset, DataLoader
from transformers import AutoTokenizer
from evo_model import EvoTransformer
import torch.nn as nn
import torch.optim as optim

class FeedbackDataset(Dataset):
    def __init__(self, csv_file):
        self.data = pd.read_csv(csv_file).dropna()
        self.tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")

    def __len__(self):
        return len(self.data)

    def __getitem__(self, idx):
        row = self.data.iloc[idx]
        prompt = row['prompt']
        context = row['context']
        label = int(row['label'])
        text = f"{prompt} {context}"
        encoded = self.tokenizer(text, truncation=True, padding='max_length', max_length=128, return_tensors="pt")
        return encoded['input_ids'].squeeze(0), torch.tensor(label)

def fine_tune_on_feedback():
    csv_file = "feedback_log.csv"
    dataset = FeedbackDataset(csv_file)
    dataloader = DataLoader(dataset, batch_size=8, shuffle=True)

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model = EvoTransformer().to(device)
    model.load_state_dict(torch.load("evo_hellaswag.pt", map_location=device))
    model.train()

    criterion = nn.CrossEntropyLoss()
    optimizer = optim.Adam(model.parameters(), lr=2e-5)

    for epoch in range(2):
        for input_ids, labels in dataloader:
            input_ids = input_ids.to(device)
            labels = labels.to(device)

            outputs = model(input_ids)
            loss = criterion(outputs, labels)

            optimizer.zero_grad()
            loss.backward()
            optimizer.step()

    torch.save(model.state_dict(), "evo_hellaswag.pt")
    print("✅ Evo retrained and saved.")