Instructions

This section provides a Rec SDK TensorFlow-based little-demo sample to describe the files and key APIs required for model training using tf.Session. little-demo is only a code example and describes the logic for API calls. It does not contain specific models or implement specific functions.

It is for reference only and cannot be used to adapt to your own model. little-demo is stored here.

Multi-round evaluation is not supported in the train_and_evaluate scenario.

**Table 1** little-demo description
File	Description
config.py	Model-related configuration
dataset.py	Dataset preprocessing
deterministic_loss	Loss sample for deterministic computing
main.py	Model training entry
model.py	Model construction
op_impl_mode.ini	Operator configuration file
optimizer.py	Optimizer
random_data_generator.py	Random dataset generation
README.md	Description of demo model running
run_deterministic.sh	Script for deterministic computing
run_mode.py	Encapsulated training and inference processes
run.sh	Model training startup script
utils.py	Accuracy detection tool

Parent topic: Quick Start