from transformers import PreTrainedModel, PretrainedConfig
import torch
import torch.nn as nn
import torch.nn.functional as F


class SASOKConfig(PretrainedConfig):
    model_type = "sasok"

    def __init__(self, vocab_size=50000, hidden_size=512, num_heads=8, num_layers=4, **kwargs):
        super().__init__(**kwargs)
        self.vocab_size = vocab_size
        self.hidden_size = hidden_size
        self.num_heads = num_heads
        self.num_layers = num_layers


class SASOKModel(PreTrainedModel):
    config_class = SASOKConfig

    def __init__(self, config):
        super().__init__(config)
        self.embedding = nn.Embedding(config.vocab_size, config.hidden_size)
        self.percept_bn = nn.BatchNorm1d(config.hidden_size)

        self.emotion_ln = nn.LayerNorm(config.hidden_size)
        self.attn = nn.MultiheadAttention(config.hidden_size, config.num_heads, batch_first=True)
        self.attn_ln = nn.LayerNorm(config.hidden_size)

        self.meta_stack = nn.ModuleList([
            nn.TransformerEncoderLayer(d_model=config.hidden_size, nhead=config.num_heads, norm_first=True)
            for _ in range(config.num_layers)
        ])

        self.final_ln = nn.LayerNorm(config.hidden_size)
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size)

        self.init_weights()

    def forward(self, input_ids, attention_mask=None, labels=None):
        x = self.embedding(input_ids)
        x = x.transpose(1, 2)
        x = self.percept_bn(x).transpose(1, 2)

        x = self.emotion_ln(x)
        x_ln = self.attn_ln(x)
        x, _ = self.attn(x_ln, x_ln, x_ln) + x

        for layer in self.meta_stack:
            x = layer(x)

        x = self.final_ln(x)
        logits = self.lm_head(x)

        loss = None
        if labels is not None:
            loss_fn = nn.CrossEntropyLoss()
            loss = loss_fn(logits.view(-1, logits.size(-1)), labels.view(-1))

        return {"loss": loss, "logits": logits} if loss is not None else {"logits": logits}