Have a bitcoin_logger instance running for a while. eg:
target/release/bitcoin-logger --zmq-address tcp://192.168.1.167:28332 --save-path /mnt/ssd/bitcoin_log --rpc-address http://127.0.0.1:8332 --cookie-path /mnt/ssd/bitcoin/.cookie
Make a csv with bitcoin-csv binary in the bitcoin_logger. eg:
./target/release/bitcoin-csv --dataset-file /mnt/bigssd/bitcoin/dataset --load-path /mnt/big/bitcoin_log/
Use tensorflow python program at https://colab.research.google.com/drive/1js7MCPkggQGvFXeMijPZy4G2cWparzlZ
Create a virtualenv with needed requirements
source venv/bin/activate
python model.py
python model_with_hurry.pu
There are two models because one is done for hurry tx: confirming in 1 or 2 blocks, and the other model for tx confirming from 3 to 1008 blocks
Copy the resulting dirs, like 20210221-220251 into this repo, under models dir.
update build.rs pointing to the new dirs in default_models var.
update test test_vector poiting to the new dirs
cargo test