tritonclient.grpc doesn't support timeout for other commands than infer.

[dogky123]

Description [ WARNING] client.py:207 - tritonclient.grpc doesn't support timeout for other commands than infer. Ignoring network_timeout: 60.0.

To reproduce ModelClient(url='grpc://172.19.16.45:8001',model_name='test')

Question How to ignore the warning or fix this warning

[piotrm-nvidia]

Thank you for your feature request and your interest in PyTriton. I appreciate your feedback and I will try to help you with your issue.

ModelClient uses tritonclient.grpc.InferenceServerClient to communicate with the Triton server over GRPC protocol. However, this client does not support timeout parameter for the whole connection and for the configuration and status methods. These methods are used by ModelClient to send configuration requests before any inference request is sent.

Therefore, ModelClient cannot enforce any timeout different than the default for GRPC connections when using GRPC protocol. You can use HTTP protocol instead, which does not have this limitation, if you want to avoid this warning.

The extension of tritonclient.grpc API with more timeouts is not planned soon, as far as I know.

We are considering two possible solutions for ModelClient to make this warning less intrusive:

• Replace logging.warning with warning.warn, which can be suppressed by the user.
• Downgrade logging level to logging.info, which is less alarming and more informative.

We would love to hear your opinion on these solutions. Please let us know what you think and if you have any other suggestions. Thank you for your cooperation and understanding.

[piotrm-nvidia]

See: https://github.com/triton-inference-server/pytriton/issues/38

[dogky123]

Thank you for your replying. I usually use warnings.filterwarnings to ignore the warning for other libs. It woule be better if i can apply it to pytriton.

[piotrm-nvidia]

Warnings support was added from release 0.4.2.