rxnn 0.1.61__tar.gz → 0.1.63__tar.gz

{rxnn-0.1.61 → rxnn-0.1.63}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: rxnn
-Version: 0.1.61
+Version: 0.1.63
 Summary: RxNN: Reactive Neural Networks Platform
 License: Apache-2.0
 Keywords: deep-learning,ai,machine-learning
@@ -14,7 +14,6 @@ Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Requires-Dist: datasets (>=3.5.0,<4.0.0)
-Requires-Dist: flash-attention (>=1.0.0,<2.0.0)
 Requires-Dist: huggingface-hub (>=0.30.0,<0.31.0)
 Requires-Dist: tensorboard (>=2.19.0,<3.0.0)
 Requires-Dist: tokenizers (>=0.21.0,<0.22.0)

{rxnn-0.1.61 → rxnn-0.1.63}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "rxnn"
-version = "0.1.61"
+version = "0.1.63"
 description = "RxNN: Reactive Neural Networks Platform"
 
 license = "Apache-2.0"
@@ -30,4 +30,4 @@ datasets = "^3.5.0"
 tokenizers = "^0.21.0"
 huggingface-hub = "^0.30.0"
 tensorboard = "^2.19.0"
-flash-attention = "^1.0.0"
+

{rxnn-0.1.61 → rxnn-0.1.63}/src/rxnn/training/base.py

@@ -113,7 +113,9 @@ class BaseTrainer(ABC):
 
         self.model.train()
         for epoch in range(self.current_epoch, self.current_epoch + epochs):
-            if self.is_running:
+            if not self.is_running:
+                break
+            else:
                 self.current_epoch = epoch
                 self.epoch_steps = 0
                 if train_sampler is not None:
@@ -143,7 +145,9 @@ class BaseTrainer(ABC):
         self.optimizer_step_count = 0
 
         for batch_idx, batch in enumerate(dataloader):
-            if self.is_running:
+            if not self.is_running:
+                break
+            else:
                 for callback in self.callbacks:
                     callback.on_batch_start(self.model, batch_idx, batch)
                 if self.get_batch_size(batch) == batch_size:

{rxnn-0.1.61 → rxnn-0.1.63}/src/rxnn/transformers/attention.py

@@ -137,9 +137,6 @@ class MultiHeadAttention(nn.Module):
         b, t, d = query.size()
         q, k, v = self._forward_qkv(query, key, value, b, t, d)
         if not self.rel_embed:
-            print('q', q.size())
-            print('k', k.size())
-            print('v', v.size())
             q, k = self._apply_rope(q, k)
             attn_output = self._calculate_attention(q, k, v, b, t, d, mask=mask)
         else:

{rxnn-0.1.61 → rxnn-0.1.63}/src/rxnn/transformers/layers.py

@@ -86,7 +86,6 @@ class ReactiveTransformerLayer(nn.Module):
         residual = x
         if not self.use_post_norm:
             x = self.norm1(x)
-        print('self x', x.size())
         x = self.attention(x, x, x, mask=mask)
         x = residual + x
         if self.use_post_norm:
@@ -95,8 +94,6 @@ class ReactiveTransformerLayer(nn.Module):
         residual = x
         if not self.use_post_norm:
             x = self.norm2(x)
-        print('x', x.size())
-        print('STM', stm.size())
         x = self.memory_cross_attention(x, stm, stm)
         x = residual + x
         if self.use_post_norm: