debug

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

Author: kmehant

scripts/benchmarks/scenarios-moe.yaml

@@ -60,16 +60,23 @@ scenarios:
                 - 'ibm-granite/granite-3.0-3b-a800m-instruct'
                 - 'ibm-research/moe-7b-1b-active-shared-experts'
 
-    -   name: accelerated-moe-granite4
+    -   name: accelerated-moe-granite-4
         framework_config:
-            - moe-scattermoe-granite-ep8
-            - moe-scattermoe-granite-ep8-foak
-            - moe-scattermoe-granite-ep8-padding-free-foak
+            - # without acceleration
+            - moe-scattermoe-granite-ep1
+            - moe-scattermoe-granite-ep2
+            - moe-scattermoe-granite-ep4
+            - moe-scattermoe-granite-ep1-padding-free
+            - moe-scattermoe-granite-ep1-padding-free-foak
+            - moe-scattermoe-granite-ep2-padding-free
+            - moe-scattermoe-granite-ep2-padding-free-foak
+            - moe-scattermoe-granite-ep4-padding-free
+            - moe-scattermoe-granite-ep4-padding-free-foak
         arguments:
             learning_rate: 5e-5
             torch_dtype: bfloat16
             gradient_accumulation_steps: null
-            per_device_train_batch_size: 8
+            per_device_train_batch_size: 4
             logging_steps: 1
             packing: False
             adam_epsilon: 1e-8