rxnn 0.2.64__tar.gz → 0.2.66__tar.gz

{rxnn-0.2.64 → rxnn-0.2.66}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: rxnn
-Version: 0.2.64
+Version: 0.2.66
 Summary: RxNN: Reactive Neural Networks Platform
 License: Apache-2.0
 Keywords: deep-learning,ai,machine-learning

{rxnn-0.2.64 → rxnn-0.2.66}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "rxnn"
-version = "0.2.64"
+version = "0.2.66"
 description = "RxNN: Reactive Neural Networks Platform"
 
 license = "Apache-2.0"

{rxnn-0.2.64 → rxnn-0.2.66}/src/rxnn/rxt/models.py

@@ -268,6 +268,7 @@ class RxTAlphaMemoryAttention(nn.Module, PyTorchModelHubMixin, license="apache-2
             residual_per_slot_gate: bool = False,
             residual_init_gate: float = 0.0,
             use_dynamic_residual_gate: bool = False,
+            use_tanh_residual_gate: bool = False,
             debug_mode: bool = False,
             debug_interval: int = 10,
             **kwargs,
@@ -300,7 +301,7 @@ class RxTAlphaMemoryAttention(nn.Module, PyTorchModelHubMixin, license="apache-2
             stm, attention_layers, memory_norm_layers,
             use_gated_residual=use_gated_residual, per_slot_gate=residual_per_slot_gate,
             init_gate=residual_init_gate, use_dynamic_gate=use_dynamic_residual_gate,
-            debug_mode=debug_mode, debug_interval=debug_interval,
+            use_tanh_gate=use_tanh_residual_gate, debug_mode=debug_mode, debug_interval=debug_interval,
         )
 
     def freeze(self):

{rxnn-0.2.64 → rxnn-0.2.66}/src/rxnn/training/utils.py

@@ -151,5 +151,9 @@ def get_gradient_norms(model: nn.Module):
         param_norm = p.grad.data.norm(2)
         total_norm += param_norm.item() ** 2
     total_norm = total_norm ** 0.5
-    mean_norm = total_norm / len(grad_params)
+    params_len = len(grad_params)
+    if params_len != 0:
+        mean_norm = total_norm / params_len
+    else:
+        mean_norm = 0.0
     return total_norm, mean_norm