update comments and rebuild ec2 image

Author: jinyan-li1

dlc_developer_config.toml

@@ -41,12 +41,12 @@ build_frameworks = ["pytorch"]
 
 
 # By default we build both training and inference containers. Set true/false values to determine which to build.
-build_training = true 
+build_training = true
 build_inference = false
 
 # Set do_build to "false" to skip builds and test the latest image built by this PR
 # Note: at least one build is required to set do_build to "false"
-do_build = true
+do_build = true 
 
 [notify]
 ### Notify on test failures
@@ -57,24 +57,24 @@ notify_test_failures = false
 
 [test]
 ### On by default
-sanity_tests = true
-security_tests = true
-  safety_check_test = true
-  ecr_scan_allowlist_feature = true
-ecs_tests = true
-eks_tests = true
-ec2_tests = true
+sanity_tests = true 
+security_tests = true 
+  safety_check_test = true 
+  ecr_scan_allowlist_feature = true 
+ecs_tests = true 
+eks_tests = true 
+ec2_tests = true 
 # Set it to true if you are preparing a Benchmark related PR
-ec2_benchmark_tests = true 
+ec2_benchmark_tests = true
 
 ### Set ec2_tests_on_heavy_instances = true to be able to run any EC2 tests that use large/expensive instance types by
 ### default. If false, these types of tests will be skipped while other tests will run as usual.
 ### These tests are run in EC2 test jobs, so ec2_tests must be true if ec2_tests_on_heavy_instances is true.
 ### Off by default (set to false)
-ec2_tests_on_heavy_instances = true  
+ec2_tests_on_heavy_instances = true 
 ### SM specific tests
 ### On by default
-sagemaker_local_tests = true 
+sagemaker_local_tests = true
 ### Set enable_ipv6 = true to run tests with IPv6-enabled resources
 ### Off by default (set to false)
 enable_ipv6 = false
@@ -92,13 +92,13 @@ enable_ipv6 = false
 ipv6_vpc_name = ""
 
 # run standard sagemaker remote tests from test/sagemaker_tests
-sagemaker_remote_tests = true 
+sagemaker_remote_tests = true
 # run efa sagemaker tests
-sagemaker_efa_tests = false 
+sagemaker_efa_tests = true
 # run release_candidate_integration tests
-sagemaker_rc_tests = true
+sagemaker_rc_tests = true 
 # run sagemaker benchmark tests
-sagemaker_benchmark_tests = true
+sagemaker_benchmark_tests = true 
 
 # SM remote EFA test instance type
 sagemaker_remote_efa_instance_type = ""
@@ -119,7 +119,7 @@ use_scheduler = false
 ### TRAINING PR JOBS ###
 
 # Standard Framework Training
-dlc-pr-pytorch-training = "pytorch/training/buildspec-2-8-sm.yml" 
+dlc-pr-pytorch-training = "pytorch/training/buildspec-2-8-ec2.yml" 
 dlc-pr-tensorflow-2-training = ""
 dlc-pr-autogluon-training = ""
 

pytorch/training/buildspec.yml

@@ -1 +1 @@
-buildspec_pointer: buildspec-2-8-sm.yml
+buildspec_pointer: buildspec-2-8-ec2.yml

pytorch/training/docker/2.8/py3/cu129/Dockerfile.gpu

@@ -83,6 +83,7 @@ RUN apt-get update \
     check \
     llvm \
     xz-utils \
+    libsqlite3-dev \
  && rm -rf /var/lib/apt/lists/* \
  && apt-get clean
 

test/dlc_tests/container_tests/bin/efa/testEFA

@@ -89,7 +89,7 @@ check_efa_nccl_all_reduce(){
 
     RETURN_VAL=${PIPESTATUS[0]}
     # In case, if you would like see logs, uncomment below line
-    RESULT=$(cat ${TRAINING_LOG})
+    # RESULT=$(cat ${TRAINING_LOG})
 
     if [ ${RETURN_VAL} -eq 0 ]; then
         echo "***************************** check_efa_nccl_all_reduce passed *****************************"

test/dlc_tests/ec2/pytorch/training/common_cases.py

@@ -357,7 +357,7 @@ def pytorch_cudnn_match_gpu(pytorch_training, ec2_connection, region):
     """
     Test cuDNN Package
     PT 2.1 reintroduces a dependency on CUDNN to support NVDA TransformerEngine. This test is to ensure that torch CUDNN matches system CUDNN in the container.
-    Checks both /usr/include/ and /usr/include/x86_64-linux-gnu/ paths to support different cuDNN package installations.
+    Checks both /usr/include/ and /usr/local/cuda/include/ paths to support different cuDNN package installations.
     """
     container_name = "pytorch_cudnn"
     account_id = get_account_id_from_image_uri(pytorch_training)