Missing Tensorflow 2.9 inference image

[plumdog]

Describe the bug

Tensorflow 2.9 inference image is reported as not available by this library, but does exist in registries.

To reproduce

>>> import sagemaker
>>> sagemaker.__version__
'2.99.0'
>>> sagemaker.image_uris.retrieve('tensorflow', 'eu-west-1', image_scope='inference', version='2.8', instance_type='ml.m5.large')
'763104351884.dkr.ecr.eu-west-1.amazonaws.com/tensorflow-inference:2.8-cpu'
>>> sagemaker.image_uris.retrieve('tensorflow', 'eu-west-1', image_scope='inference', version='2.9', instance_type='ml.m5.large')
Traceback (most recent call last):
...
    raise ValueError(
ValueError: Unsupported tensorflow version: 2.9. You may need to upgrade your SDK version (pip install -U sagemaker) for newer tensorflow versions. Supported tensorflow version(s): 1.10.0, 1.11.0, 1.12.0, 1.13.0, 1.14.0, 1.15.0, 1.15.2, 1.15.3, 1.15.4, 1.15.5, 1.4.1, 1.5.0, 1.6.0, 1.7.0, 1.8.0, 1.9.0, 2.0.0, 2.0.1, 2.0.2, 2.0.3, 2.0.4, 2.1.0, 2.1.1, 2.1.2, 2.1.3, 2.2.0, 2.2.1, 2.2.2, 2.3.0, 2.3.1, 2.3.2, 2.4.1, 2.4.3, 2.5.1, 2.6.0, 2.6.3, 2.7.0, 2.8.0, 1.10, 1.11, 1.12, 1.13, 1.14, 1.15, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.1, 2.2, 2.3, 2.4, 2.5, 2.6, 2.7, 2.8.

But then, pulling images directly with Docker:

$ aws ecr get-login-password --region eu-west-1 | docker login --username AWS --password-stdin "763104351884.dkr.ecr.eu-west-1.amazonaws.com"
Login Succeeded
$ docker pull 763104351884.dkr.ecr.eu-west-1.amazonaws.com/tensorflow-inference:2.8.0-cpu                                                   
2.8.0-cpu: Pulling from tensorflow-inference
d5fd17ec1767: Downloading [=========>                                         ]  5.316MB/28.57MB
055d1b7bd083: Downloading [=>                                                 ]  4.855MB/214MB
501cf980eba1: Downloading [===>                                               ]  4.855MB/65.85MB
6eee2e438f6b: Pulling fs layer 
9b8cedc52042: Waiting 
c973b6111e73: Pulling fs layer 
8eda15d07ac8: Waiting 
d24fcb50554b: Waiting 
48896ee9d13a: Waiting 
c1c14bb3e866: Waiting 
29ea66b2c7c4: Pulling fs layer 
f27ea8b7dd4d: Pulling fs layer 
08b10d0bd12c: Waiting 
4fccd6276502: Waiting 
2a86a3199c71: Waiting 
37d47cae2d7e: Pulling fs layer 
af78b1cccabf: Waiting 
dcae54aa69ab: Waiting 
^C
$ docker pull 763104351884.dkr.ecr.eu-west-1.amazonaws.com/tensorflow-inference:2.9.0-cpu
2.9.0-cpu: Pulling from tensorflow-inference
d7bfe07ed847: Already exists 
ef6b4bb7a4ac: Downloading [=>                                                 ]  8.084MB/208.5MB
e365b2d13a81: Downloading [====>                                              ]  5.934MB/65.85MB
dc8a8d8bbf88: Download complete 
0dc26fa0c010: Downloading [================>                                  ]  10.03MB/30.2MB
63525e769512: Waiting 
5f7f441d968a: Waiting 
45742bfa386b: Waiting 
f20501310d53: Waiting 
7d3ab0cf075a: Waiting 
9bb4f6e81a92: Waiting 
b36eab1c1cbb: Waiting 
11ebb67b1885: Waiting 
75fc33f4c92d: Waiting 
31f4f6c8ee90: Waiting 
07a43246c874: Waiting 
57a569513401: Waiting 
572596e2e57a: Waiting

So the images exist in the registries and I am able to pull them, but this library doesn't know about them. This tracks with the current state of https://github.com/aws/sagemaker-python-sdk/blob/b4f05b86a90f4ae202ad7f9b048922ab490731fe/src/sagemaker/image_uri_config/tensorflow.json, which lists 2.9 as a training image, but not as an inference image.

Expected behavior

This library to know about the Tensorflow 2.9 inference image and for sagemaker.image_uris.retrieve('tensorflow', 'eu-west-1', image_scope='inference', version='2.9', instance_type='ml.m5.large') to return '763104351884.dkr.ecr.eu-west-1.amazonaws.com/tensorflow-inference:2.9-cpu'

Screenshots or logs
See replication steps above.

System information
A description of your system. Please provide:

• SageMaker Python SDK version: 2.99.0
• Framework name (eg. PyTorch) or algorithm (eg. KMeans): Tensorflow
• Framework version: 2.9
• Python version: n/a
• CPU or GPU: n/a
• Custom Docker image (Y/N): N

Additional context
Add any other context about the problem here.

[arjkesh]

Hi @plumdog - thank you for reporting this issue. The sagemaker python SDK is unfortunately not updated with the TF2.9 images, though a PR is open to remedy the issue. We hope to get this in as soon as possible. In the meantime, if you are blocked on deploying a TF 2.9 model, you can use the "image_uri" parameter in the sagemaker model class to deploy TF2.9 models, and you can reference the same image URI you posted in this issue.

763104351884.dkr.ecr.eu-west-1.amazonaws.com/tensorflow-inference:2.9-cpu

Please let us know if there is any further info we can provide you in the short term - we will update this issue immediately after merging the TF2.9 image URI config changes.

[martinRenou]

Closing as fixed by #3465