tensorflow serving with mino s3
521bibi opened this issue · comments
I started a minio S3 with docker locally
Then,
docker run --rm -p 8501:8501 --name tfs-s3 -e AWS_ACCESS_KEY_ID=minioadmin -e AWS_SECRET_ACCESS_KEY=minioadmin -e S3_ENDPOINT=http://127.0.0.1:9000 -e AWS_REGION=us-east-1 -t tensorflow/serving:2.4.1 --model_config_file=s3://models/models.config
error：
Tried tensorflow/serving version 2.4, 2.5, 2.6, all the same. Maybe my problem is with S3_ENDPOINT=http://127.0.0.1:9000? Or am I missing a step?
Hope someone can help me, thanks a lot
As per this comment from a similar issue, tensorflow_io should be imported separately to your code to have s3 access. Please confirm if you have already imported the same.
Also you can change the S3_ENDPOINT from loop back address to the actual endpoint IP to see if helps.
Thank you!
Adding tf_io will not work. When using tensorflow, import tf_io. It is useful to use the tf_io function, but I am using tensorflow serving now, and I want to use the official image directly.
The second solution is effective, I changed S3_ENDPOINT=http://127.0.0.1:9000 to other IP of my computer and it worked
Thank you very much
tensorflow serving with s3 has been solved, the next step I will try to deploy tensorflow serving with hdfs.