rxnn 0.1.60__py3-none-any.whl → 0.1.61__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (5) hide show
  1. rxnn/transformers/models.py +4 -4
  2. {rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/METADATA +1 -1
  3. {rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/RECORD +5 -5
  4. {rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/LICENSE +0 -0
  5. {rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/WHEEL +0 -0
rxnn/transformers/models.py CHANGED
@@ -72,11 +72,11 @@ class ReactiveTransformerDecoder(ReactiveTransformerBase):
72
72
  # Process shared layers
73
73
  if self.shared_layers is not None:
74
74
  for i in range(self.num_shared_layers):
75
- layer_stm = self.stm(i)
75
+ layer_stm = self.stm(i).expand(x.size(0), -1, -1)
76
76
  x = self.shared_layers[i](x, layer_stm, mask=mask)
77
77
  # Process own layers
78
78
  for i in range(self.num_own_layers):
79
- layer_stm = self.stm(i)
79
+ layer_stm = self.stm(i).expand(x.size(0), -1, -1)
80
80
  x = self.layers[i](x, layer_stm, mask=mask)
81
81
  return self.head(x)
82
82
 
@@ -93,12 +93,12 @@ class ReactiveTransformerEncoder(ReactiveTransformerBase):
93
93
  # Process shared layers
94
94
  if self.shared_layers is not None:
95
95
  for i in range(self.num_shared_layers):
96
- layer_stm = self.stm(i)
96
+ layer_stm = self.stm(i).expand(x.size(0), -1, -1)
97
97
  x = self.shared_layers[i](x, layer_stm, mask=attention_mask)
98
98
  hidden_states.append(x)
99
99
  # Process own layers
100
100
  for i in range(self.num_own_layers):
101
- layer_stm = self.stm(i)
101
+ layer_stm = self.stm(i).expand(x.size(0), -1, -1)
102
102
  x = self.layers[i](x, layer_stm, mask=attention_mask)
103
103
  hidden_states.append(x)
104
104
  return x, torch.stack(hidden_states)
{rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/METADATA RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.3
2
2
  Name: rxnn
3
- Version: 0.1.60
3
+ Version: 0.1.61
4
4
  Summary: RxNN: Reactive Neural Networks Platform
5
5
  License: Apache-2.0
6
6
  Keywords: deep-learning,ai,machine-learning
{rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/RECORD RENAMED
@@ -20,12 +20,12 @@ rxnn/transformers/attention.py,sha256=2pJoAqRB6AKHtA2lj158NzbHzXr0dCaKj1RKjMfHsO
20
20
  rxnn/transformers/ff.py,sha256=jJnuBDsnnX5uYC_WZH8cXAYrMnz0P-iX7MwcPivjRtI,2533
21
21
  rxnn/transformers/layers.py,sha256=tSOwr_IwJcAjPLmyRJI-93wRjgHH6OPJj4QrdkEFYdc,7282
22
22
  rxnn/transformers/mask.py,sha256=J0cfLVLt3SzS2ra3KcY4khrkhI975Dw4CjpUi3Sn25s,419
23
- rxnn/transformers/models.py,sha256=QFzBrOR7tDp9d_T0HoIukBMfEbLxsCictV5p3e2ilxg,7552
23
+ rxnn/transformers/models.py,sha256=xbnn3FTNZFhaqq9A0XEM12ie_WL_58pPeq0qFXIgve0,7656
24
24
  rxnn/transformers/moe.py,sha256=j6jEx6Ip0zttlUZKKn82azxo95lkLZs-H2GLSMD88hY,5859
25
25
  rxnn/transformers/positional.py,sha256=2l38RS0Dini3f6Z3LUHr3XwWzg1UK7fO2C6wazWDAYU,4292
26
26
  rxnn/transformers/sampler.py,sha256=poWBpxg1iuK5gEJtxHkk5VVfS9V48hs2Olqdhy_Gw8c,6548
27
27
  rxnn/utils.py,sha256=d5U8i5ukovgDyqiycc2AoxObTz_eF_bgo2MKvdtJ98s,467
28
- rxnn-0.1.60.dist-info/LICENSE,sha256=C8coDFIUYuOcke4JLPwTqahQUCyXyGq6WOaigOkx8tY,11275
29
- rxnn-0.1.60.dist-info/METADATA,sha256=eTT7LMBzM_AiWqreCwois6pTV50-zb8tyMlgrz4Mq-g,16627
30
- rxnn-0.1.60.dist-info/WHEEL,sha256=fGIA9gx4Qxk2KDKeNJCbOEwSrmLtjWCwzBz351GyrPQ,88
31
- rxnn-0.1.60.dist-info/RECORD,,
28
+ rxnn-0.1.61.dist-info/LICENSE,sha256=C8coDFIUYuOcke4JLPwTqahQUCyXyGq6WOaigOkx8tY,11275
29
+ rxnn-0.1.61.dist-info/METADATA,sha256=Utq--dpip7-MXEYQLdu8-FyV6YFmFhRX9JpZpDpVBAY,16627
30
+ rxnn-0.1.61.dist-info/WHEEL,sha256=fGIA9gx4Qxk2KDKeNJCbOEwSrmLtjWCwzBz351GyrPQ,88
31
+ rxnn-0.1.61.dist-info/RECORD,,
{rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/LICENSE RENAMED
File without changes
{rxnn-0.1.60.dist-info → rxnn-0.1.61.dist-info}/WHEEL RENAMED
File without changes