debug

Signed-off-by: Mehant Kammakomati <mehant.kammakomati2@ibm.com>

Author: kmehant

scripts/benchmarks/scenarios-moe.yaml

@@ -76,7 +76,7 @@ scenarios:
             learning_rate: 5e-5
             torch_dtype: bfloat16
             gradient_accumulation_steps: null
-            per_device_train_batch_size: 4
+            per_device_train_batch_size: 2
             logging_steps: 1
             packing: False
             adam_epsilon: 1e-8