Upload 5 files

codaBlock.py

@@ -0,0 +1,14 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from typing import Optional, Tuple
+
+# Final Projection Block
+class CodaBlock(nn.Module):
+    def __init__(self, d_model: int, vocab_size: int):
+        super().__init__()
+        self.norm = nn.LayerNorm(d_model)
+        self.output_proj = nn.Linear(d_model, vocab_size)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        return self.output_proj(self.norm(x))

latent_Recurrent.py

@@ -0,0 +1,22 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from typing import Optional, Tuple
+from Model.prelude_Block import PreludeBlock
+from Model.recurrent_Block import RecurrentBlock
+from Model.codaBlock import CodaBlock
+
+# Full Latent Recurrent Depth Model
+class LatentRecurrentDepthLM(nn.Module):
+    def __init__(self, vocab_size: int, d_model: int, num_heads: int, dropout: float = 0.1):
+        super().__init__()
+        self.prelude = PreludeBlock(vocab_size, d_model, num_heads, dropout)
+        self.recurrent = RecurrentBlock(d_model, num_heads, dropout)
+        self.coda = CodaBlock(d_model, vocab_size)
+
+    def forward(self, x: torch.Tensor, num_iterations: int, mask: Optional[torch.Tensor] = None) -> torch.Tensor:
+        hidden = self.prelude(x, mask)
+        recurrent_state = torch.zeros_like(hidden)
+        for _ in range(num_iterations):
+            hidden, recurrent_state = self.recurrent(hidden, recurrent_state, mask)
+        return self.coda(hidden)

multi_head_Attention.py

@@ -0,0 +1,44 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from typing import Optional, Tuple
+
+# Multi-Head Attention Mechanism
+class MultiHeadAttention(nn.Module):
+    def __init__(self, d_model: int, num_heads: int, dropout: float = 0.1):
+        super().__init__()
+        assert d_model % num_heads == 0
+
+        self.d_model = d_model
+        self.num_heads = num_heads
+        self.head_dim = d_model // num_heads
+
+        self.q_proj = nn.Linear(d_model, d_model)
+        self.k_proj = nn.Linear(d_model, d_model)
+        self.v_proj = nn.Linear(d_model, d_model)
+        self.o_proj = nn.Linear(d_model, d_model)
+
+        self.dropout = nn.Dropout(dropout)
+
+    def forward(self, x: torch.Tensor, mask: Optional[torch.Tensor] = None) -> torch.Tensor:
+        batch_size, seq_len, d_model = x.shape
+        
+        # Project and reshape for multi-head attention
+        q = self.q_proj(x).reshape(batch_size, seq_len, self.num_heads, self.head_dim)
+        k = self.k_proj(x).reshape(batch_size, seq_len, self.num_heads, self.head_dim)
+        v = self.v_proj(x).reshape(batch_size, seq_len, self.num_heads, self.head_dim)
+
+        # Transpose for attention computation
+        q, k, v = q.transpose(1, 2), k.transpose(1, 2), v.transpose(1, 2)
+
+        # Compute attention scores
+        scores = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.head_dim)
+        if mask is not None:
+            scores = scores.masked_fill(mask == 0, float('-inf'))
+        
+        attn_weights = F.softmax(scores, dim=-1)
+        attn_weights = self.dropout(attn_weights)
+        
+        # Apply attention to values
+        out = torch.matmul(attn_weights, v).transpose(1, 2).reshape(batch_size, seq_len, d_model)
+        return self.o_proj(out)

prelude_Block.py

@@ -0,0 +1,28 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from typing import Optional, Tuple
+from Model.multi_head_Attention import MultiHeadAttention
+
+
+# Prelude Block (Initial Processing)
+class PreludeBlock(nn.Module):
+    def __init__(self, vocab_size: int, d_model: int, num_heads: int, dropout: float = 0.1):
+        super().__init__()
+        self.token_embedding = nn.Embedding(vocab_size, d_model)
+        self.pos_encoding = nn.Parameter(torch.zeros(1, 1024, d_model))
+        self.attention = MultiHeadAttention(d_model, num_heads, dropout)
+        self.norm1, self.norm2 = nn.LayerNorm(d_model), nn.LayerNorm(d_model)
+        self.feed_forward = nn.Sequential(
+            nn.Linear(d_model, 4 * d_model),
+            nn.GELU(),
+            nn.Linear(4 * d_model, d_model),
+            nn.Dropout(dropout)
+        )
+
+    def forward(self, x: torch.Tensor, mask: Optional[torch.Tensor] = None) -> torch.Tensor:
+        seq_len = x.size(1)
+        x = self.token_embedding(x) + self.pos_encoding[:, :seq_len, :]
+        attended = self.attention(self.norm1(x), mask)
+        x = x + attended
+        return x + self.feed_forward(self.norm2(x))

recurrent_Block.py

@@ -0,0 +1,25 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from typing import Optional, Tuple
+from Model.multi_head_Attention import MultiHeadAttention
+
+# Recurrent Block (Processing Over Time)
+class RecurrentBlock(nn.Module):
+    def __init__(self, d_model: int, num_heads: int, dropout: float = 0.1):
+        super().__init__()
+        self.attention = MultiHeadAttention(d_model, num_heads, dropout)
+        self.norm1, self.norm2 = nn.LayerNorm(d_model), nn.LayerNorm(d_model)
+        self.feed_forward = nn.Sequential(
+            nn.Linear(d_model, 4 * d_model),
+            nn.GELU(),
+            nn.Linear(4 * d_model, d_model),
+            nn.Dropout(dropout)
+        )
+        self.state_proj = nn.Linear(d_model, d_model)
+
+    def forward(self, x: torch.Tensor, recurrent_state: torch.Tensor, mask: Optional[torch.Tensor] = None) -> Tuple[torch.Tensor, torch.Tensor]:
+        recurrent_state = self.state_proj(recurrent_state)
+        x = x + recurrent_state
+        attended = self.attention(self.norm1(x), mask)
+        return x + attended + self.feed_forward(self.norm2(x)), x