model_architecture.py

from dataclasses import dataclass
import torch
import torch.nn as nn

@dataclass
class GPT2Config:
    vocab_size: int = 50304
    hidden_size: int = 1024  # GPT-2 Medium
    num_layers: int = 24     # GPT-2 Medium (for ~450M total)
    num_heads: int = 16
    intermediate_size: int = 4096
    max_position_embeddings: int = 1024
    rms_norm_eps: float = 1e-6
    dropout: float = 0.1
    use_bias: bool = False

class RMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.eps = eps  

    def forward(self, x):
        variance = x.pow(2).mean(-1, keepdim=True)
        x = x * torch.rsqrt(variance + self.eps)
        return self.weight * x  


class RotaryEmbedding(nn.Module):
    def __init__(self, dim, max_position_embeddings=2048, base=10000):
        super().__init__()
        self.dim = dim
        self.max_position_embeddings = max_position_embeddings
        self.base = base
        inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float() / self.dim))
        self.register_buffer("inv_freq", inv_freq)

    def forward(self, x, seq_len):
        t = torch.arange(seq_len, device=self.inv_freq.device).type_as(self.inv_freq)
        freqs = torch.einsum("i,j->ij", t, self.inv_freq)
        emb = torch.cat((freqs, freqs), dim=-1)
        cos = emb.cos()[None, None, :, :]
        sin = emb.sin()[None, None, :, :]
        return cos, sin

def rotate_half(x):
    x1 = x[..., : x.shape[-1] // 2]
    x2 = x[..., x.shape[-1] // 2 :]
    return torch.cat((-x2, x1), dim=-1)

def apply_rotary_pos_emb(q, k, cos, sin):
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed

class GPT2Attention(nn.Module):
    def __init__(self, config: GPT2Config):
        super().__init__()
        self.config = config
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_heads
        self.head_dim = self.hidden_size // self.num_heads
        self.scale = self.head_dim ** -0.5

        assert self.hidden_size % self.num_heads == 0, "hidden_size must be divisible by num_heads"

        self.q_proj = nn.Linear(self.hidden_size, self.hidden_size, bias=config.use_bias)
        self.k_proj = nn.Linear(self.hidden_size, self.hidden_size, bias=config.use_bias)
        self.v_proj = nn.Linear(self.hidden_size, self.hidden_size, bias=config.use_bias)
        self.o_proj = nn.Linear(self.hidden_size, self.hidden_size, bias=config.use_bias)
        self.rotary_emb = RotaryEmbedding(self.head_dim, max_position_embeddings=config.max_position_embeddings)
        self.dropout = nn.Dropout(config.dropout)

    def forward(self, hidden_states, attention_mask=None):
        batch_size, seq_len, _ = hidden_states.shape

        q = self.q_proj(hidden_states)
        k = self.k_proj(hidden_states)
        v = self.v_proj(hidden_states)

        q = q.view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
        k = k.view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
        v = v.view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)

        cos, sin = self.rotary_emb(None, seq_len)
        q, k = apply_rotary_pos_emb(q, k, cos, sin)

        attn_weights = torch.matmul(q, k.transpose(-2, -1)) * self.scale

    
        if attention_mask is not None:
            attn_weights = attn_weights + attention_mask

        attn_weights = nn.functional.softmax(attn_weights, dim=-1)
        attn_weights = self.dropout(attn_weights)
        
        attn_output = torch.matmul(attn_weights, v)
        attn_output = attn_output.transpose(1, 2).contiguous().view(batch_size, seq_len, self.hidden_size)
        attn_output = self.o_proj(attn_output)
        attn_output = self.dropout(attn_output)

        return attn_output

class GPT2FFN(nn.Module):
    def __init__(self, config: GPT2Config):
        super().__init__()
        self.gate_proj = nn.Linear(config.hidden_size, config.intermediate_size, bias=config.use_bias)
        self.up_proj = nn.Linear(config.hidden_size, config.intermediate_size, bias=config.use_bias)
        self.down_proj = nn.Linear(config.intermediate_size, config.hidden_size, bias=config.use_bias)
        self.dropout = nn.Dropout(config.dropout)
        self.act = nn.SiLU()

    def forward(self, x):
        gate = self.act(self.gate_proj(x))
        up = self.up_proj(x)
        feed_forward_hidden = gate * up
        feed_forward_hidden = self.dropout(feed_forward_hidden)
        output = self.down_proj(feed_forward_hidden)
        return output

class GPT2Block(nn.Module):
    def __init__(self, config: GPT2Config):
        super().__init__()
        self.attn_norm = RMSNorm(config.hidden_size, config.rms_norm_eps)
        self.ffn_norm = RMSNorm(config.hidden_size, config.rms_norm_eps)
        self.attention = GPT2Attention(config)
        self.ffn = GPT2FFN(config)
        self.dropout = nn.Dropout(config.dropout)

    def forward(self, hidden_states, attention_mask=None):
        residual = hidden_states
        hidden_states = self.attn_norm(hidden_states)
        attn_output = self.attention(hidden_states, attention_mask)
        attn_output = self.dropout(attn_output)
        hidden_states = residual + attn_output

        residual = hidden_states
        hidden_states = self.ffn_norm(hidden_states)
        ffn_output = self.ffn(hidden_states)
        ffn_output = self.dropout(ffn_output)
        hidden_states = residual + ffn_output

        return hidden_states

class GPT2Model(nn.Module):
    def __init__(self, config: GPT2Config):
        super().__init__()
        self.config = config
        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
        self.layers = nn.ModuleList([GPT2Block(config) for _ in range(config.num_layers)])
        self.norm = RMSNorm(config.hidden_size, config.rms_norm_eps)
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
        self.apply(self._init_weights)

        self.lm_head.weight = self.embed_tokens.weight  

    def _init_weights(self, module):
        if isinstance(module, nn.Linear):
            torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)
            if module.bias is not None:
                torch.nn.init.zeros_(module.bias)
        elif isinstance(module, nn.Embedding):
            torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)

    def forward(self, input_ids, attention_mask=None):
        batch_size, seq_len = input_ids.shape
        hidden_states = self.embed_tokens(input_ids)

        if attention_mask is None:
            causal_mask = torch.full((seq_len, seq_len), float('-inf'), device=input_ids.device)
            causal_mask = torch.triu(causal_mask, diagonal=1)
            attention_mask = causal_mask[None, None, :, :]

        for layer in self.layers:
            hidden_states = layer(hidden_states, attention_mask)

        hidden_states = self.norm(hidden_states)
        logits = self.lm_head(hidden_states)

        return logits