transition

modules/blocks.py

@@ -12,7 +12,7 @@ from einops import rearrange
 import sys
 sys.path.append(str(Path(__file__).parent))
 
-from .util import AutoPad, Interpolate, ShapeMixin, F_x, Flatten, ResidualBlock, PreNorm
+from .util import AutoPad, Interpolate, ShapeMixin, F_x, Flatten
 
 DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
@@ -85,7 +85,6 @@ class ConvModule(ShapeMixin, nn.Module):
         else:
             pass
 
-
     def forward(self, x):
         tensor = self.norm(x)
         tensor = self.conv(tensor)
@@ -100,12 +99,13 @@ class PreInitializedConvModule(ShapeMixin, nn.Module):
     def __init__(self, in_shape, weight_matrix):
         super(PreInitializedConvModule, self).__init__()
         self.in_shape = in_shape
+        self.weight_matrix = weight_matrix
         raise NotImplementedError
         # ToDo Get the weight_matrix shape and init a conv_module of similar size,
         #      override the weights then.
 
     def forward(self, x):
-
+        x = torch.matmul(x, self.weight_matrix)  # ToDo: This is an Placeholder
         return x
 
 
@@ -214,8 +214,9 @@ class RecurrentModule(ShapeMixin, nn.Module):
         tensor = self.rnn(x)
         return tensor
 
+
 class FeedForward(nn.Module):
-    def __init__(self, dim, hidden_dim, dropout = 0.):
+    def __init__(self, dim, hidden_dim, dropout=0.):
         super().__init__()
         self.net = nn.Sequential(
             nn.Linear(dim, hidden_dim),
@@ -224,31 +225,35 @@ class FeedForward(nn.Module):
             nn.Linear(hidden_dim, dim),
             nn.Dropout(dropout)
         )
+
     def forward(self, x):
         return self.net(x)
 
+
 class Attention(nn.Module):
-    def __init__(self, dim, heads = 8, dropout = 0.):
+    def __init__(self, dim, heads=8, dropout=0.):
         super().__init__()
         self.heads = heads
-        self.scale = dim ** -0.5
+        self.scale = dim / heads ** -0.5
 
-        self.to_qkv = nn.Linear(dim, dim * 3, bias = False)
+        self.to_qkv = nn.Linear(dim, dim * 3, bias=False)
         self.to_out = nn.Sequential(
             nn.Linear(dim, dim),
             nn.Dropout(dropout)
         )
 
-    def forward(self, x, mask = None):
+    def forward(self, x, mask=None, return_attn_weights=False):
+        # noinspection PyTupleAssignmentBalance
         b, n, _, h = *x.shape, self.heads
-        qkv = self.to_qkv(x).chunk(3, dim = -1)
-        q, k, v = [rearrange(t, 'b n (h d) -> b h n d', h = h) for t in qkv]
+
+        qkv = self.to_qkv(x).chunk(3, dim=-1)
+        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b h n d', h=h), qkv)
 
         dots = torch.einsum('bhid,bhjd->bhij', q, k) * self.scale
         mask_value = -torch.finfo(dots.dtype).max
 
         if mask is not None:
-            mask = F.pad(mask.flatten(1), [1, 0], value = True)
+            mask = F.pad(mask.flatten(1), (1, 0), value=True)
             assert mask.shape[-1] == dots.shape[-1], 'mask has incorrect dimensions'
             mask = mask[:, None, :] * mask[:, :, None]
             dots.masked_fill_(~mask, mask_value)
@@ -258,39 +263,47 @@ class Attention(nn.Module):
 
         out = torch.einsum('bhij,bhjd->bhid', attn, v)
         out = rearrange(out, 'b h n d -> b n (h d)')
-        out =  self.to_out(out)
-        return out
-
-class Transformer(nn.Module):
-    def __init__(self, dim, depth, heads, mlp_dim, dropout):
-        super().__init__()
-        self.layers = nn.ModuleList([])
-        for _ in range(depth):
-            self.layers.append(nn.ModuleList([
-                ResidualBlock(PreNorm(dim, Attention(dim, heads = heads, dropout = dropout))),
-                ResidualBlock(PreNorm(dim, FeedForward(dim, mlp_dim, dropout = dropout)))
-            ]))
-
-    def forward(self, x, mask = None, *_, **__):
-        for attn, ff in self.layers:
-            x = attn(x, mask = mask)
-            x = ff(x)
-        return x
+        out = self.to_out(out)
+        if return_attn_weights:
+            return out, attn
+        else:
+            return out
 
 
 class TransformerModule(ShapeMixin, nn.Module):
 
-    def __init__(self, in_shape, hidden_size, n_heads, num_layers=1, dropout=None, use_norm=False, activation='gelu'):
+    def __init__(self, in_shape, depth, heads, mlp_dim, dropout=None, use_norm=False, activation='gelu'):
         super(TransformerModule, self).__init__()
 
         self.in_shape = in_shape
 
         self.flat = Flatten(self.in_shape) if isinstance(self.in_shape, (tuple, list)) else F_x(in_shape)
 
-        self.transformer = Transformer(dim=self.flat.flat_shape, depth=num_layers, heads=n_heads,
-                                       mlp_dim=hidden_size, dropout=dropout)
+        self.layers = nn.ModuleList([])
+        self.embedding_dim = self.flat.flat_shape
+        self.norm = nn.LayerNorm(self.embedding_dim)
+        self.attns = nn.ModuleList([Attention(self.embedding_dim, heads=heads, dropout=dropout) for _ in range(depth)])
+        self.mlps = nn.ModuleList([FeedForward(self.embedding_dim, mlp_dim, dropout=dropout) for _ in range(depth)])
 
-    def forward(self, x, mask=None, key_padding_mask=None):
+    def forward(self, x, mask=None, return_attn_weights=False, **_):
         tensor = self.flat(x)
-        tensor = self.transformer(tensor, mask, key_padding_mask)
-        return tensor
+        attn_weights = list()
+
+        for attn, mlp in zip(self.attns, self.mlps):
+            # Attention
+            skip_connection = tensor.clone()
+            tensor = self.norm(tensor)
+            if return_attn_weights:
+                tensor, attn_weight = attn(tensor, mask=mask, return_attn_weights=return_attn_weights)
+                attn_weights.append(attn_weight)
+            else:
+                tensor = attn(tensor, mask=mask)
+            tensor = tensor + skip_connection
+
+            # MLP
+            skip_connection = tensor.clone()
+            tensor = self.norm(tensor)
+            tensor = mlp(tensor)
+            tensor = tensor + skip_connection
+
+        return (tensor, attn_weights) if return_attn_weights else tensor