Update.
authorFrançois Fleuret <francois@fleuret.org>
Fri, 7 Jul 2023 15:48:30 +0000 (17:48 +0200)
committerFrançois Fleuret <francois@fleuret.org>
Fri, 7 Jul 2023 15:48:30 +0000 (17:48 +0200)
mygpt.py

index c93010a..8cd0152 100755 (executable)
--- a/mygpt.py
+++ b/mygpt.py
@@ -62,9 +62,7 @@ class CacheWrapper(nn.Module):
         else:
             self.cache_y[:, bs.first : bs.first + bs.nb] = self.f(bs.slice())
 
-        bs.x = self.cache_y
-
-        return bs
+        return BracketedSequence(self.cache_y, bs.first, bs.nb)
 
 
 ##############################
@@ -76,8 +74,7 @@ class WithResidual(nn.Module):
         self.f = f[0] if len(f) == 1 else nn.Sequential(*f)
 
     def forward(self, bs):
-        bs.x = bs.x + self.f(bs).x
-        return bs
+        return BracketedSequence(bs.x + self.f(bs).x, bs.first, bs.nb)
 
 
 ##############################
@@ -108,9 +105,7 @@ class AddPositionalEncoding(nn.Module):
             bs.slice() + self.pe[bs.first : bs.first + bs.nb]
         )
 
-        bs.x = self.cache_y
-
-        return bs
+        return BracketedSequence(self.cache_y, bs.first, bs.nb)
 
 
 ##############################
@@ -125,6 +120,7 @@ class QKVAttention(nn.Module):
         def randw(*d):
             return nn.Parameter(torch.randn(*d) / math.sqrt(d[-1]))
 
+        assert causal, "TODO: Switch off the cache when non-causal!!!"
         self.causal = causal
         self.attention_dropout = attention_dropout
 
@@ -148,6 +144,7 @@ class QKVAttention(nn.Module):
         q = torch.einsum(
             "ntc,hdc->nhtd", x_q[:, bs_q.first : bs_q.first + bs_q.nb], self.w_q
         )
+
         self.cache_k[:, :, bs_q.first : bs_q.first + bs_q.nb] = torch.einsum(
             "ntc,hdc->nhtd", x_q[:, bs_q.first : bs_q.first + bs_q.nb], self.w_k
         )
@@ -181,9 +178,7 @@ class QKVAttention(nn.Module):
 
         self.cache_y[:, bs_q.first : bs_q.first + bs_q.nb] = y @ self.w_o
 
-        bs_q.x = self.cache_y
-
-        return bs_q
+        return BracketedSequence(self.cache_y, bs_q.first, bs_q.nb)
 
 
 ##############################
@@ -252,7 +247,7 @@ class MyGPT(nn.Module):
                     m.weight.fill_(1.0)
 
     def forward(self, bs):
-        bs.x = F.pad(bs.x, (1, -1))
+        bs = BracketedSequence(F.pad(bs.x, (1, -1)), bs.first, bs.nb)
         bs = self.embedding(bs)
         bs = self.trunk(bs)
         bs = self.readout(bs)
@@ -288,27 +283,27 @@ class MyGPT(nn.Module):
 if __name__ == "__main__":
     print("Basic check.")
 
-    vocabulary_size = 10
-    x = torch.randint(vocabulary_size, (9, 7))
+    vocabulary_size = 3
+    x = torch.randint(vocabulary_size, (1, 5))
 
     model = MyGPT(
         vocabulary_size=vocabulary_size,
-        dim_model=18,
-        dim_keys=50,
-        dim_hidden=100,
+        dim_model=4,
+        dim_keys=2,
+        dim_hidden=2,
         nb_heads=2,
         nb_blocks=1,
         dropout=0.1,
+        causal=True,
     )
 
     model.eval()
 
     y1 = model(BracketedSequence(x)).x
-
     y2 = torch.randn_like(y1)
     for s in range(x.size(1)):
         z = model(BracketedSequence(x, s, 1))
-        y2[:, s] = z.x[:, s]
+        y2[:, s] = z.slice()
 
     print(f"error={((y1 - y2).norm() / (y1.norm() + y2.norm())).item()}")