deep-learning-containers icon indicating copy to clipboard operation
deep-learning-containers copied to clipboard

Local build failing to load artifacts with "The specified bucket does not exist"

Open mattreyuk opened this issue 1 year ago • 0 comments

I'd like to build a custom version of Tensorflow 2.3 inference

but when I run docker build locally, it cannot access the s3 locations in

RUN curl https://tensorflow-aws.s3-us-west-2.amazonaws.com/MKL-Libraries/libiomp5.so -o /usr/local/lib/libiomp5.so
RUN curl https://tensorflow-aws.s3-us-west-2.amazonaws.com/MKL-Libraries/libmklml_intel.so -o /usr/local/lib/libmklml_intel.so

RUN curl $TFS_URL -o /usr/bin/tensorflow_model_server \
 && chmod 555 /usr/bin/tensorflow_model_server

when I look at the file contents in my built image they contain: <Error><Code>NoSuchBucket</Code><Message>The specified bucket does not exist</Message><BucketName>tensorflow-aws</BucketName> <Error><Code>NoSuchBucket</Code><Message>The specified bucket does not exist</Message><BucketName>aws-tensorflow-binaries</BucketName>

I've tried running the build with a valid role that has S3 access and the AWS_REGION=us-west-2

mattreyuk avatar Apr 15 '23 00:04 mattreyuk