keep debug log, rebuild sm image with released base dlc

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_inference = false
+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 = true
+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 = ""
@@ -122,7 +122,7 @@ use_scheduler = false
 dlc-pr-base = ""
 
 # Standard Framework Training
-dlc-pr-pytorch-training = "pytorch/training/buildspec-2-8-sm.yml"
+dlc-pr-pytorch-training = "pytorch/training/buildspec-2-8-sm.yml" 
 dlc-pr-tensorflow-2-training = ""
 dlc-pr-autogluon-training = ""
 

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

@@ -11,8 +11,7 @@ ARG GDRCOPY_VERSION=2.5
 ARG TE_VERSION=2.5
 ARG FLASH_ATTN_VERSION=2.8.2
 
-# FROM public.ecr.aws/deep-learning-containers/base:12.9.1-gpu-py312-ubuntu22.04-ec2 AS base_image
-FROM 669063966089.dkr.ecr.us-west-2.amazonaws.com/pr-base:12.9.1-gpu-py312-cu129-ubuntu22.04-ec2-pr-5186-2025-08-19-23-05-36 AS base_image
+FROM public.ecr.aws/deep-learning-containers/base:12.9.1-gpu-py312-ubuntu22.04-ec2 AS base_image
 # base has EFA, PYTHON and CUDA 12.9
 
 # This arg required to stop docker build waiting for region configuration while installing tz data from ubuntu 20