[test] reduce batch size for mnist training (#984)

The `test_mnist_training` is failing on n300 machines, due to a bug in ttnn. Reduce the batch size so that the test can pass. Filed an issue in tt-metal: tenstorrent/tt-metal#16348
tenstorrent · Dec 30, 2024 · 1c0421e · 1c0421e
1 parent ece7caa
commit 1c0421e
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/forge/test/mlir/mnist/training/test_training.py b/forge/test/mlir/mnist/training/test_training.py
@@ -29,7 +29,7 @@ def test_mnist_training():
 
     # Set training hyperparameters
     num_epochs = 3
-    batch_size = 2048
+    batch_size = 1024
     learning_rate = 0.001
 
     # Load dataset