sagemaker use reserved instances (#258)

* sagemaker use reserved instances * lint
mlfoundations · Apr 26, 2024 · 083fa31 · 083fa31
1 parent 5e4e7f9
commit 083fa31
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/sagemaker_train/launch_sagemaker_train.py b/sagemaker_train/launch_sagemaker_train.py
@@ -189,6 +189,7 @@ def get_job_name(base):
         max_wait=5 * 24 * 60 * 60 if args.spot_instance else None,
         input_mode="FastFile",
         # environment={"TORCH_DISTRIBUTED_DEBUG": "DETAIL", "TORCH_CPP_LOG_LEVEL": "INFO"},
+        environment={"SM_USE_RESERVED_CAPACITY": "1"},
         keep_alive_period_in_seconds=30 * 60 if not args.spot_instance else None,  # 30 minutes
     )