CODEWITHSUNDEEP
dockernvidiatritonserver
Atharav Jadhav
Atharav Jadhav

Reputation: 25

Customizing deployment with Model Analyzer in NVIDIA Triton Server

I am following the tutorial from NVIDIA Triton Server and am currently on the 3rd step to getting to know deployments of ML models. The step involves installing the Model Analyser Module and there is an associated command with it:

model-analyzer profile --model-repository /workspace/model_repository --profile-models text_recognition --triton-launch-mode=local --output-model-repository-path /workspace/output/ -f perf.yaml --override-output-model-repository --latency-budget 10 --run-config-search-mode quick

And I really don't know where is the perf.yaml file located. Here is the link for the tutorial: https://github.com/triton-inference-server/tutorials/tree/main/Conceptual_Guide/Part_3-optimizing_triton_configuration

I have tried executing that command both inside and outside the server container and inside the SDK container as well. Still no avail.

Any help regarding this would really be helpful.

Stating my question: "Where is the perf.yaml file located?"

Upvotes: 0

Views: 155

Answers (1)

Mahmoud Khaled
Mahmoud Khaled

Reputation: 1

According to the docs https://github.com/triton-inference-server/model_analyzer/blob/main/docs/config.md#cli-and-yaml-config-options

-f <path-to-configs>

This is used to path a yaml file of the configuration needed by the model analyser for the profiling. Check the tutorial attached, as it has configuration examples.

Upvotes: 0

Related Questions