Market data need to be downloaded from the following link:
https://lobsterdata.com/info/DataSamples.php
To install the FINN environment follow these installation steps https://finn.readthedocs.io/en/latest/getting_started.html
To set the environment for the tools and the Docker files, modify the following script with your own environment variables
source FINN_setup.sh
source run_docker.sh
Two files control the hardware generation step in FINN. The two files are located in the configuration directory called configuration_files
- dataflow_build_config.json (where the hardware target platform or FPGA device can be configure)
- folding_config.json (configuration file to set the PE and SMID values for each layer)
Note: Multiple dataflow_build_config are provided to show different possible configuration
To run the training, generate the model and analyse the report use the following scripts
python run_batch.py
results.csv file will be created and appended according to the suite test results.