Network visualiser

Bench: 2134743

scripts/network_to_img.py

@@ -0,0 +1,70 @@
+#  WhiteCore is a C++ chess engine
+#  Copyright (c) 2023 Balázs Szilágyi
+#
+#  This program is free software: you can redistribute it and/or modify
+#  it under the terms of the GNU General Public License as published by
+#  the Free Software Foundation, either version 3 of the License, or
+#  (at your option) any later version.
+#
+#  This program is distributed in the hope that it will be useful,
+#  but WITHOUT ANY WARRANTY; without even the implied warranty of
+#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+#  GNU General Public License for more details.
+#
+#  You should have received a copy of the GNU General Public License
+#  along with this program.  If not, see <https://www.gnu.org/licenses/>.
+#
+
+import argparse
+import struct
+import numpy as np
+import matplotlib.pyplot as plt
+
+
+col_labels = ["WK", "WP", "WN", "WB", "WR", "WQ",
+              "BK", "BP", "BN", "BB", "BR", "BQ"]
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('input_file', help='Path to the network file')
+    args = parser.parse_args()
+
+    with open(args.input_file, 'rb') as f:
+        magic_raw = f.read(4)
+        magic = struct.unpack('i', magic_raw)[0]
+        assert magic == -4
+
+        data = f.read(256 * 2)
+        biases = struct.unpack('256h', data)
+
+        data = f.read(768 * 256 * 2)
+        weights = struct.unpack('196608h', data)
+        weights = np.array(weights)
+        weights = weights.reshape((768, 256))
+        fig, axs = plt.subplots(8, 12, figsize=(6, 4))
+        plt.subplots_adjust(hspace=0.1, wspace=0.1)
+        for base in range(32):
+            for neuron in range(8):
+                numbers = weights[:, neuron + base * 8]
+                numbers = np.array(numbers).reshape((96, 8))
+
+                row_start = 0
+                row_end = 7
+                for i in range(12):
+                    x = numbers[row_start:row_end, :]
+                    img = np.interp(x, (-128, 128), (0, 255)).astype(np.uint8)
+                    axs[neuron][i].set_xticks([])
+                    axs[neuron][i].set_yticks([])
+                    axs[neuron][i].imshow(img, cmap='copper')
+                    row_start += 8
+                    row_end += 8
+
+            for i in range(12):
+                axs[7][i].set_xlabel(col_labels[i])
+
+            plt.savefig(f'img/network_{base}.png')
+
+
+if __name__ == '__main__':
+    main()