x-transformers 1.27.20__py3-none-any.whl → 1.27.22__py3-none-any.whl

x_transformers/xval.py

@@ -260,10 +260,19 @@ class XValAutoregressiveWrapper(nn.Module):
         inp, target = x[:, :-1], x[:, 1:]
         x_num_inp, x_num_target = x_num[:, :-1], x_num[:, 1:]
 
+        # ignore index
+
+        target_mask = target != self.ignore_index
+
+        # key padding mask
+
         mask = kwargs.get('mask', None)
-        if exists(mask) and mask.shape[1] == x.shape[1]:
-            mask = mask[:, :-1]
-            kwargs['mask'] = mask
+        if exists(mask):
+            target_mask &= mask
+
+            if mask.shape[1] == x.shape[1]:
+                mask = mask[:, :-1]
+                kwargs['mask'] = mask
 
         logits, numerical_pred = self.net(inp, x_num_inp, **kwargs)
 
@@ -271,17 +280,23 @@ class XValAutoregressiveWrapper(nn.Module):
 
         cross_entropy_loss = F.cross_entropy(logits, target, reduction = 'none', ignore_index = self.ignore_index)
 
-        target_mask = target != self.ignore_index
+        # protect against nan in `x_num` input tensor
+
+        target_is_number_mask = target == self.net.numerical_token_id
+        x_num_target = x_num_target.masked_fill(~target_is_number_mask, 0.)
+
+        # numerical mse loss
 
         numerical_mse_loss = F.mse_loss(numerical_pred, x_num_target, reduction = 'none')
 
         numerical_mse_loss = numerical_mse_loss * target_mask
+        numerical_mse_loss = numerical_mse_loss.masked_fill(~target_is_number_mask, 0.)
 
-        loss = cross_entropy_loss + numerical_mse_loss * self.numerical_loss_weight
+        # combine losses
 
-        if exists(mask):
-            loss = loss[mask]
+        loss = cross_entropy_loss + numerical_mse_loss * self.numerical_loss_weight
 
+        loss = loss[target_mask]
         loss = loss.mean()
 
         if not return_loss_breakdown:

{x_transformers-1.27.20.dist-info → x_transformers-1.27.22.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: x-transformers
-Version: 1.27.20
+Version: 1.27.22
 Summary: X-Transformers - Pytorch
 Home-page: https://github.com/lucidrains/x-transformers
 Author: Phil Wang

{x_transformers-1.27.20.dist-info → x_transformers-1.27.22.dist-info}/RECORD

@@ -6,9 +6,9 @@ x_transformers/dpo.py,sha256=LjvWgCkqTl-UuehrzQ8nkX5guLr4whYwsmm7SKSwdls,3450
 x_transformers/nonautoregressive_wrapper.py,sha256=ys_p8obc7lTeeodCqvkRKxOXQ1C9T3j5Jwr-JbVgnXk,10432
 x_transformers/x_transformers.py,sha256=kQhRUMGDsinzkdYcOfE1GriJ057j7D4xSjbH79FFRSE,63574
 x_transformers/xl_autoregressive_wrapper.py,sha256=DCx4n0_c1tFai4nOqaWVnqx2p9eutsZsDMiMP1ckxNU,4117
-x_transformers/xval.py,sha256=ulEPep6i5Hl7H-H9vGfdsmHdprUmK8ajB306jViyV2c,8147
-x_transformers-1.27.20.dist-info/LICENSE,sha256=As9u198X-U-vph5noInuUfqsAG2zX_oXPHDmdjwlPPY,1066
-x_transformers-1.27.20.dist-info/METADATA,sha256=KfY4iK3HVlgVafNYc7GDUarZLvEacMDqzMACQBAZndU,662
-x_transformers-1.27.20.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
-x_transformers-1.27.20.dist-info/top_level.txt,sha256=hO6KGpFuGucRNEtRfme4A_rGcM53AKwGP7RVlRIxS5Q,15
-x_transformers-1.27.20.dist-info/RECORD,,
+x_transformers/xval.py,sha256=EN3hxxleTRGYeAz6i4x3U_PrOm9TjxMF3eDhMKGx59E,8575
+x_transformers-1.27.22.dist-info/LICENSE,sha256=As9u198X-U-vph5noInuUfqsAG2zX_oXPHDmdjwlPPY,1066
+x_transformers-1.27.22.dist-info/METADATA,sha256=RTbXIIpFRnve8FVp8vLQ4LE-9x59IV6ADnu34gGAZXA,662
+x_transformers-1.27.22.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
+x_transformers-1.27.22.dist-info/top_level.txt,sha256=hO6KGpFuGucRNEtRfme4A_rGcM53AKwGP7RVlRIxS5Q,15
+x_transformers-1.27.22.dist-info/RECORD,,