add README and license

Cargo.toml

@@ -4,6 +4,7 @@ version = "2.1.0"
 edition = "2021"
 description = "Converts marlinflow JSON networks into binary format, for use in chess engines."
 authors = ["Cosmo Bobak"]
+license = "MIT"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 Cosmo Bobak
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,4 +1,64 @@
 # nnue-jsontobin
 
 A little utility for converting from [Marlinflow](https://github.com/dsekercioglu/marlinflow)'s NNUE JSON format into a binary NNUE for embedding into chess engine executables.
-Currently supports relative networks and HalfKA networks.
+Currently supports relative networks and HalfKA networks.
+
+## Usage
+
+Most typical usage is to convert a JSON file into a unified binary file, with an 8-bit output weight array.
+This is achieved by running the following command:
+
+```
+nnue-jsontobin INPUT.json --unified OUTPUT.bin
+```
+
+This will produce a binary file with the following structure:
+
+in C/C++:
+```cpp
+struct NetworkWeights {
+    std::array<std::int16_t, 768 * NEURONS * BUCKETS> feature_weights;
+    std::array<std::int16_t, NEURONS>                 feature_biases;
+    std::array<std::int8_t , NEURONS * 2>             output_weights;
+    std::int16_t                                      output_bias;
+}
+```
+or, in Rust:
+```rust
+#[repr(C)]
+struct NetworkWeights {
+    feature_weights: [i16; 768 * NEURONS * BUCKETS],
+    feature_biases:  [i16; NEURONS],
+    output_weights:  [i8; NEURONS * 2],
+    output_bias:     i16,
+}
+```
+
+Alternatively, you can produce split binary files, one for each field of the above struct.
+This is achieved by running the following command:
+
+```
+nnue-jsontobin INPUT.json --split OUTPUT_DIR
+```
+
+The output directory will contain the following files:
+```
+feature_weights.bin
+feature_biases.bin
+output_weights.bin
+output_bias.bin
+```
+
+For backwards compatibility, you can also produce a binary file with a 16-bit output weight array, by adding the `--big-out` flag.
+
+`nnue-jsontobin` performs quantisation of the network weights during the conversion process, and by default uses 255 and 64 as the quantisation factors.
+
+These factors are configurable with the `--qa` and `--qb` flags, respectively.
+
+## Building
+
+`nnue-jsontobin` is written in Rust, and can be built with `cargo build --release`.
+
+## License
+
+`nnue-jsontobin` is licensed under the MIT license.