disable ap

aws · Feb 14, 2025 · 50bfe82 · 50bfe82
1 parent 136918c
commit 50bfe82
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 2 deletions.
diff --git a/tensorflow/training/buildspec-2-18-sm.yml b/tensorflow/training/buildspec-2-18-sm.yml
@@ -5,7 +5,7 @@ framework: &FRAMEWORK tensorflow
 version: &VERSION 2.18.0
 short_version: &SHORT_VERSION "2.18"
 arch_type: x86
-autopatch_build: "True"
+# autopatch_build: "True"
 
 repository_info:
   training_repository: &TRAINING_REPOSITORY

diff --git a/tensorflow/training/docker/2.18/py3/cu125/Dockerfile.gpu b/tensorflow/training/docker/2.18/py3/cu125/Dockerfile.gpu
@@ -396,6 +396,7 @@ RUN rm -rf /tmp/*
 # Copy workaround script for incorrect hostname
 COPY start_cuda_compat.sh /usr/local/bin/start_cuda_compat.sh
 COPY dockerd-entrypoint.py /usr/local/bin/dockerd-entrypoint.py
+RUN chmod +x /usr/local/bin/start_cuda_compat.sh
 RUN chmod +x /usr/local/bin/dockerd-entrypoint.sh
 
 RUN HOME_DIR=/root \

diff --git a/tensorflow/training/docker/build_artifacts/dockerd-entrypoint.py b/tensorflow/training/docker/build_artifacts/dockerd-entrypoint.py
@@ -19,7 +19,10 @@
 import subprocess
 
 # run compat mounting by default
-subprocess.run(["bash", "-m", "/usr/local/bin/start_cuda_compat.sh"])
+try:
+    subprocess.run(["bash", "-m", "/usr/local/bin/start_cuda_compat.sh"])
+except Exception as e:
+    print(f"Error running script: {e}")
 
 if not os.path.exists("/opt/ml/input/config"):
     subprocess.call(["python", "/usr/local/bin/deep_learning_container.py", "&>/dev/null", "&"])