import torch
from transformers import AutoTokenizer, GPT2Model
import torch.nn as nn

class ChessMoveClassifier(nn.Module):
    def __init__(self, model_name, num_labels=4096):
        super().__init__()
        self.base_model = GPT2Model.from_pretrained(model_name)
        self.dropout = nn.Dropout(0.1)
        self.classifier = nn.Linear(self.base_model.config.n_embd, num_labels)

    def forward(self, input_ids, attention_mask=None, **kwargs):
        outputs = self.base_model(input_ids=input_ids, attention_mask=attention_mask)
        hidden_state = outputs.last_hidden_state[:, -1, :]
        logits = self.classifier(self.dropout(hidden_state))
        return {"logits": logits}

def model_fn(model_dir):
    model = ChessMoveClassifier(model_name="austindavis/ChessGPT_d12")
    model.load_state_dict(torch.load(f"{model_dir}/model.pt", map_location="cpu"))
    model.eval()
    return model