Update pytorch.md

source/machine-learning/pytorch.md

@@ -6,3 +6,11 @@
   - convert tensor to standard Python number
   - only works for tensors with one element - otherwise an exception is raised
   - implicitly moves the value to the CPU
+
+## shape
+
+Example:
+
+```python
+batch_size, sequence_length = attention_mask.shape
+```