torch.manual_seed(123)
model = GPTModel(GPT_CONFIG_124M)

#setting output layer to use same weights as embedding layer. 
# as it is same dictionary. 
model.out_head.weight = model.tok_emb.weight


tied_param_count = sum(p.numel() for p in model.parameters())
print(f"Total number of unique parameters: {tied_param_count}")

        __
        __