Configuring the Environment

Before performing model inference using TensorFlow Serving, configure the BIOS, create and activate the conda environment, and install the dependencies.

BIOS Configuration Description

Before the test, configure the BIOS environment. For details about the configuration items, description, and paths, see Table 1.

**Table 1** BIOS configuration description
Configuration Item	Configuration Description	Menu Path	Recommended Value
Power Policy	Configures the mode of the power policy item.	Advanced->Performance Config->Power Policy	Performance
SMT2	Sets the Simultaneous Multithreading Technology (SMT) mode.	Advanced> Power And Performance Configuration > CPU PM Control > SMT2	Enabled
Support Smmu	Sets the mode of the System Memory Management Unit (SMMU).	Advanced > MISC Config > Support Smmu	Disabled
CPU Prefetching Configuration	Sets the CPU prefetching mode.	Advanced> Power And Performance Configuration > CPU PM Control > CPU Prefetching Configuration	Enabled

Creating a conda Environment

Create a conda virtual environment for the inference phase.
1
conda create -n model_zoo_infer python=3.11
Activate the conda environment after it is created.
1
conda activate model_zoo_infer
Install the inference dependency.
1
pip install tensorflow==2.15.0

Parent topic: Model Inference on TensorFlow Serving