-
-
Notifications
You must be signed in to change notification settings - Fork 18
/
visualize.py
557 lines (464 loc) · 22.4 KB
/
visualize.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
import argparse
import chess
import features
import model as M
import numpy as np
import torch
import matplotlib.pyplot as plt
from matplotlib.gridspec import GridSpec
from serialize import NNUEReader
class NNUEVisualizer():
def __init__(self, model, ref_model, args):
self.model = model
self.ref_model = ref_model
self.args = args
import matplotlib as mpl
self.dpi = 100
mpl.rcParams["figure.figsize"] = (
self.args.default_width//self.dpi, self.args.default_height//self.dpi)
mpl.rcParams["figure.dpi"] = self.dpi
def _process_fig(self, name, fig=None):
if self.args.save_dir:
from os.path import join
destname = join(
self.args.save_dir, "{}{}.jpg".format("" if self.args.label is None else self.args.label + "_", name))
print("Saving {}".format(destname))
if fig is not None:
fig.savefig(destname)
else:
plt.savefig(destname)
def plot_input_weights(self):
# Coalesce weights and transform them to Numpy domain.
weights = M.coalesce_ft_weights(self.model, self.model.input)
weights = weights[:, :M.L1]
weights = weights.flatten().numpy()
if self.args.ref_model:
ref_weights = M.coalesce_ft_weights(
self.ref_model, self.ref_model.input)
ref_weights = ref_weights[:, :M.L1]
ref_weights = ref_weights.flatten().numpy()
weights -= ref_weights
hd = M.L1 # Number of input neurons.
self.M = hd
# Preferred ratio of number of input neurons per row/col.
preferred_ratio = 8
# Number of input neurons per row.
# Find a factor of hd such that the aspect ratio
# is as close to the preferred ratio as possible.
factor, smallest_diff = 0, hd
for n in range(1, hd+1):
if hd % n == 0:
ratio = hd / (n*n)
diff = abs(preferred_ratio-ratio)
if diff < smallest_diff:
factor = n
smallest_diff = diff
numx = hd // factor
if self.args.sort_input_neurons:
# Sort input neurons by the L1-norm of their associated weights.
neuron_weights_norm = np.zeros(hd)
for i in range(hd):
neuron_weights_norm[i] = np.sum(np.abs(weights[i::hd]))
self.sorted_input_neurons = np.flip(
np.argsort(neuron_weights_norm))
else:
self.sorted_input_neurons = np.arange(hd, dtype=int)
# Derived/fixed constants.
numy = hd//numx
widthx = 128
widthy = 368
totalx = numx * widthx
totaly = numy * widthy
totaldim = totalx*totaly
if not self.args.no_input_weights:
default_order = self.args.input_weights_order == "piece-centric-flipped-king"
# Calculate masks for first input neuron.
img_mask = []
weights_mask = []
for j in range(0, weights.size, hd):
# Calculate piece and king placement.
pi = (j // hd) % 704
ki = (j // hd) // 704
if pi // 64 == 10 and ki != pi % 64:
pi += 64
piece = pi // 64
rank = (pi % 64) // 8
if ((rank == 0 or rank == 7) and (piece == 0 or piece == 1)):
# Ignore unused weights for pawns on first/last rank.
continue
kipos = [ki % 8, ki // 8]
pipos = [pi % 8, rank]
if default_order:
# Piece centric, but with flipped king position.
# Same order as used by https://github.com/hxim/Stockfish-Evaluation-Guide.
# See also https://github.com/glinscott/nnue-pytorch/issues/42#issuecomment-753604393.
inpos = [(7-kipos[0])+pipos[0]*8,
kipos[1]+(7-pipos[1])*8]
d = - 8 if piece < 2 else 48 + (piece // 2 - 1) * 64
else:
# King centric.
inpos = [8*kipos[0]+pipos[0],
8*(7-kipos[1])+(7-pipos[1])]
d = -2*(7-kipos[1]) - 1 if piece < 2 else 48 + \
(piece // 2 - 1) * 64
jhd = j % hd
x = inpos[0] + widthx * (jhd % numx) + (piece % 2)*64
y = inpos[1] + d + widthy * (jhd // numx)
ii = x + y * totalx
img_mask.append(ii)
weights_mask.append(j)
img_mask = np.array(img_mask, dtype=int)
weights_mask = np.array(weights_mask, dtype=int)
# Fill image for all input neurons.
img = np.zeros(totaldim)
for k in range(hd):
offset_x = k % numx
offset_y = k // numx
img[img_mask + offset_x*widthx + totalx*widthy *
offset_y] = weights[weights_mask + self.sorted_input_neurons[k]]
if self.args.input_weights_auto_scale:
vmin = None
vmax = None
else:
vmin = self.args.input_weights_vmin
vmax = self.args.input_weights_vmax
extra_info = ""
if self.args.sort_input_neurons:
extra_info += "sorted"
if not default_order:
extra_info += ", " + self.args.input_weights_order
else:
if not default_order:
extra_info += self.args.input_weights_order
if len(extra_info) > 0:
extra_info = "; " + extra_info
if self.args.input_weights_auto_scale or self.args.input_weights_vmin < 0:
title_template = "input weights [{LABEL}" + extra_info + "]"
hist_title_template = "input weights histogram [{LABEL}]"
cmap = 'coolwarm'
else:
img = np.abs(img)
title_template = "abs(input weights) [{LABEL}" + \
extra_info + "]"
hist_title_template = "abs(input weights) histogram [{LABEL}]"
cmap = 'viridis'
# Input weights.
scalex = (numx / numy) / preferred_ratio
plt.figure(figsize=((scalex*self.args.default_width) //
self.dpi, self.args.default_height//self.dpi))
plt.matshow(img.reshape((totaldim//totalx, totalx)),
fignum=0, vmin=vmin, vmax=vmax, cmap=cmap)
plt.colorbar(fraction=0.046, pad=0.04)
line_options = {'color': 'black', 'linewidth': 0.5}
for i in range(1, numx):
plt.axvline(x=widthx*i-0.5, **line_options)
for j in range(1, numy):
plt.axhline(y=widthy*j-0.5, **line_options)
plt.xlim([0, totalx])
plt.ylim([totaly, 0])
plt.xticks(ticks=widthx*np.arange(1, numx) - 0.5)
plt.yticks(ticks=widthy*np.arange(1, numy) - 0.5)
plt.axis('off')
plt.title(title_template.format(LABEL=self.args.label))
plt.tight_layout()
def format_coord(x, y):
x, y = int(round(x)), int(round(y))
x_ = x % widthx
y_ = y % widthy
piece_type = (y_+16)//64
piece_name = "{} {}".format(
"white" if x_ // (widthx//2) == 0 else "black", chess.piece_name(piece_type+1))
x_ = x_ % (widthx//2)
y_ = (y_+16) % 64 if y_ >= 48 else y_+8
if default_order:
# Piece centric, flipped king.
piece_square_name = chess.square_name(x_//8 + 8*(7-y_//8))
king_square_name = chess.square_name(
7-(x_ % 8) + 8*(y_ % 8))
else:
# King centric.
if piece_type == 0:
piece_square_name = chess.square_name(
x_ % 8 + 8*(6-((y_-8) % 6)))
king_square_name = chess.square_name(
x_//8 + 8*(7-(y_-8)//6))
else:
piece_square_name = chess.square_name(
x_ % 8 + 8*(7-(y_ % 8)))
king_square_name = chess.square_name(
x_//8 + 8*(7-y_//8))
neuron_id = int(numx * (y // widthy) + x // widthx)
if self.args.sort_input_neurons:
neuron_label = "sorted neuron {} (original {})".format(
neuron_id, self.sorted_input_neurons[neuron_id])
else:
neuron_label = "neuron {}".format(neuron_id)
return "{}, {} on {}, white king on {}".format(neuron_label, piece_name, piece_square_name, king_square_name)
ax = plt.gca()
ax.format_coord = format_coord
self._process_fig("input-weights")
if not self.args.no_hist:
# Input weights histogram.
plt.figure()
plt.hist(img, log=True, bins=(
np.arange(int(np.min(img)*127)-1, int(np.max(img)*127)+3)-0.5)/127)
plt.title(hist_title_template.format(LABEL=self.args.label))
plt.tight_layout()
self._process_fig("input-weights-histogram")
def plot_fc_weights(self):
if not self.args.no_fc_weights:
num_buckets = self.model.feature_set.num_ls_buckets
fig, axs = plt.subplots(3, num_buckets, dpi=self.dpi)
extra_info = ""
if self.args.sort_input_neurons:
extra_info += "; sorted input neurons"
title_template = "weights [{LABEL}" + extra_info + "]"
fig.suptitle(title_template.format(LABEL=self.args.label))
if self.args.ref_model:
ref_layers = list(self.ref_model.layer_stacks.get_coalesced_layer_stacks())
def get_l1_weights(bucket_id, l1):
l1_weights_ = l1.weight.data.numpy()
if self.args.ref_model:
l1_weights_ -= ref_layers[bucket_id][0].weight.data.numpy()
N = l1_weights_.size // (2*self.M)
l1_weights = np.zeros((2*N, self.M))
for i in range(N):
l1_weights[2*i] = l1_weights_[i][self.sorted_input_neurons]
l1_weights[2*i+1] = l1_weights_[i][self.M +
self.sorted_input_neurons]
return l1_weights, N
def get_l2_weights(bucket_id, l2):
l2_weights = l2.weight.data.numpy()
if self.args.ref_model:
l2_weights -= ref_layers[bucket_id][1].weight.data.numpy()
return l2_weights
if self.args.fc_weights_auto_scale:
vmin = None
vmax = None
else:
vmin = self.args.fc_weights_vmin
vmax = self.args.fc_weights_vmax
if self.args.fc_weights_auto_scale or self.args.fc_weights_vmin < 0:
plot_abs = False
cmap = 'coolwarm'
else:
plot_abs = True
cmap = 'viridis'
line_options = {'color': 'gray', 'linewidth': 0.5}
for bucket_id, (l1, l2, output) in enumerate(self.model.layer_stacks.get_coalesced_layer_stacks()):
l1_weights, N = get_l1_weights(bucket_id, l1)
l2_weights = get_l2_weights(bucket_id, l2)
output_weights = output.weight.data.numpy()
if self.args.ref_model:
output_weights -= ref_layers[bucket_id][2].weight.data.numpy()
ax = axs[0, bucket_id]
im = ax.matshow(np.abs(l1_weights) if plot_abs else l1_weights,
vmin=vmin, vmax=vmax, cmap=cmap)
for j in range(1, N):
ax.axhline(y=2*j-0.5, **line_options)
ax = axs[1, bucket_id]
im = ax.matshow(np.abs(l2_weights) if plot_abs else l2_weights,
vmin=None if vmin == float("-inf") else vmin,
vmax=vmax, cmap=cmap)
ax = axs[2, bucket_id]
im = ax.matshow(np.abs(output_weights) if plot_abs else output_weights,
vmin=vmin, vmax=vmax, cmap=cmap)
row_names = ['bucket {}'.format(i) for i in range(num_buckets)]
col_names = ['l1', 'l2', 'output']
for i in range(3):
for j in range(num_buckets):
ax = axs[i, j]
ax.set_xticks([])
ax.set_yticks([])
if i == 0 and row_names[j]:
ax.set_xlabel(row_names[j])
ax.xaxis.set_label_position('top')
if j == 0 and col_names[i]:
ax.set_ylabel(col_names[i])
fig.colorbar(im, fraction=0.046, pad=0.04, ax=axs[i, :].ravel().tolist())
self._process_fig("fc-weights", fig)
if not self.args.no_hist:
fig, axs = plt.subplots(num_buckets, 1, sharex=True, dpi=self.dpi)
title_template = "L1 weights histogram [{LABEL}]"
fig.suptitle(title_template.format(LABEL=self.args.label))
for bucket_id, (l1, l2, output) in enumerate(self.model.layer_stacks.get_coalesced_layer_stacks()):
# L1 weights histogram.
ax = axs[bucket_id]
l1_weights, N = get_l1_weights(bucket_id, l1)
ax.hist(l1_weights.flatten(), log=True, bins=(
np.arange(int(np.min(l1_weights)*64)-1, int(np.max(l1_weights)*64)+3)-0.5)/64)
self._process_fig("l1-weights-histogram", fig)
fig, axs = plt.subplots(num_buckets, 1, sharex=True, dpi=self.dpi)
title_template = "L2 weights histogram [{LABEL}]"
fig.suptitle(title_template.format(LABEL=self.args.label))
for bucket_id, (l1, l2, output) in enumerate(self.model.layer_stacks.get_coalesced_layer_stacks()):
# L2 weights histogram.
ax = axs[bucket_id]
l2_weights = get_l2_weights(bucket_id, l2)
ax.hist(l2_weights.flatten(), log=True, bins=(
np.arange(int(np.min(l2_weights)*64)-1, int(np.max(l2_weights)*64)+3)-0.5)/64)
self._process_fig("l2-weights-histogram", fig)
def plot_fc_biases(self):
if not self.args.no_biases:
if self.args.ref_model:
ref_layers = list(self.ref_model.layer_stacks.get_coalesced_layer_stacks())
num_buckets = self.model.feature_set.num_ls_buckets
fig, axs = plt.subplots(3, num_buckets, dpi=self.dpi)
extra_info = ""
if self.args.sort_input_neurons:
extra_info += "; sorted input neurons"
title_template = "biases [{LABEL}" + extra_info + "]"
fig.suptitle(title_template.format(LABEL=self.args.label))
if self.args.fc_weights_auto_scale:
vmin = None
vmax = None
else:
vmin = self.args.fc_weights_vmin
vmax = self.args.fc_weights_vmax
if self.args.fc_weights_auto_scale or self.args.fc_weights_vmin < 0:
plot_abs = False
cmap = 'coolwarm'
else:
plot_abs = True
cmap = 'viridis'
for bucket_id, (l1, l2, output) in enumerate(self.model.layer_stacks.get_coalesced_layer_stacks()):
l1_biases = l1.bias.data.numpy()
l2_biases = l2.bias.data.numpy()
output_bias = output.bias.data.numpy()
if self.args.ref_model:
l1_biases -= ref_layers[bucket_id][0].bias.data.numpy()
l2_biases -= ref_layers[bucket_id][1].bias.data.numpy()
output_bias -= ref_layers[bucket_id][2].bias.data.numpy()
ax = axs[0, bucket_id]
im = ax.matshow(np.expand_dims(l1_biases, axis=0),
vmin=vmin, vmax=vmax, cmap=cmap)
ax = axs[1, bucket_id]
im = ax.matshow(np.expand_dims(l2_biases, axis=0),
vmin=vmin, vmax=vmax, cmap=cmap)
ax = axs[2, bucket_id]
im = ax.matshow(np.expand_dims(output_bias, axis=0),
vmin=vmin, vmax=vmax, cmap=cmap)
row_names = ['bucket {}'.format(i) for i in range(num_buckets)]
col_names = ['l1', 'l2', 'output']
for i in range(3):
for j in range(num_buckets):
ax = axs[i, j]
ax.set_xticks([])
ax.set_yticks([])
if i == 0 and row_names[j]:
ax.set_xlabel(row_names[j])
ax.xaxis.set_label_position('top')
if j == 0 and col_names[i]:
ax.set_ylabel(col_names[i])
fig.colorbar(im, fraction=0.046, pad=0.04, ax=axs[i, :].ravel().tolist())
self._process_fig("biases", fig)
def load_model(filename, feature_set):
if filename.endswith(".pt") or filename.endswith(".ckpt"):
if filename.endswith(".pt"):
model = torch.load(filename)
else:
model = M.NNUE.load_from_checkpoint(
filename, feature_set=feature_set)
model.eval()
elif filename.endswith(".nnue"):
with open(filename, 'rb') as f:
reader = NNUEReader(f, feature_set)
model = reader.model
else:
raise Exception("Invalid filetype: " + str(filename))
return model
def main():
parser = argparse.ArgumentParser(
description="Visualizes networks in ckpt, pt and nnue format.")
parser.add_argument(
"model", help="Source model (can be .ckpt, .pt or .nnue)")
parser.add_argument(
"--ref-model", type=str, required=False,
help="Visualize the difference between the given reference model (can be .ckpt, .pt or .nnue).")
parser.add_argument(
"--ref-features", type=str, required=False,
help="The reference feature set to use (default = same as source model).")
parser.add_argument(
"--input-weights-vmin", default=-1, type=float,
help="Minimum of color map range for input weights (absolute values are plotted if this is positive or zero).")
parser.add_argument(
"--input-weights-vmax", default=1, type=float,
help="Maximum of color map range for input weights.")
parser.add_argument(
"--input-weights-auto-scale", action="store_true",
help="Use auto-scale for the color map range for input weights. This ignores input-weights-vmin and input-weights-vmax.")
parser.add_argument(
"--input-weights-order", type=str, choices=["piece-centric-flipped-king", "king-centric"], default="piece-centric-flipped-king",
help="Order of the input weights for each input neuron.")
parser.add_argument(
"--sort-input-neurons", action="store_true",
help="Sort the neurons of the input layer by the L1-norm (sum of absolute values) of their weights.")
parser.add_argument(
"--fc-weights-vmin", default=-2, type=float,
help="Minimum of color map range for fully-connected layer weights (absolute values are plotted if this is positive or zero).")
parser.add_argument(
"--fc-weights-vmax", default=2, type=float,
help="Maximum of color map range for fully-connected layer weights.")
parser.add_argument(
"--fc-weights-auto-scale", action="store_true",
help="Use auto-scale for the color map range for fully-connected layer weights. This ignores fc-weights-vmin and fc-weights-vmax.")
parser.add_argument(
"--no-hist", action="store_true",
help="Don't generate any histograms.")
parser.add_argument(
"--no-biases", action="store_true",
help="Don't generate plots for biases.")
parser.add_argument(
"--no-input-weights", action="store_true",
help="Don't generate plots or histograms for input weights.")
parser.add_argument(
"--no-fc-weights", action="store_true",
help="Don't generate plots or histograms for fully-connected layer weights.")
parser.add_argument(
"--default-width", default=1600, type=int,
help="Default width of all plots (in pixels).")
parser.add_argument(
"--default-height", default=900, type=int,
help="Default height of all plots (in pixels).")
parser.add_argument(
"--save-dir", type=str, required=False,
help="Save the plots in this directory.")
parser.add_argument(
"--dont-show", action="store_true",
help="Don't show the plots.")
parser.add_argument(
"--label", type=str, required=False,
help="Override the label used in plot titles and as prefix of saved files.")
features.add_argparse_args(parser)
args = parser.parse_args()
supported_features = ('HalfKAv2', 'HalfKAv2^')
assert args.features in supported_features
feature_set = features.get_feature_set_from_name(args.features)
from os.path import basename
label = basename(args.model)
model = load_model(args.model, feature_set)
if args.ref_model:
if args.ref_features:
assert args.ref_features in supported_features
ref_feature_set = features.get_feature_set_from_name(
args.ref_features)
else:
ref_feature_set = feature_set
ref_model = load_model(args.ref_model, ref_feature_set)
print("Visualizing difference between {} and {}".format(
args.model, args.ref_model))
from os.path import basename
label = "diff " + label + "-" + basename(args.ref_model)
else:
ref_model = None
print("Visualizing {}".format(args.model))
if args.label is None:
args.label = label
visualizer = NNUEVisualizer(model, ref_model, args)
visualizer.plot_input_weights()
visualizer.plot_fc_weights()
visualizer.plot_fc_biases()
if not args.dont_show:
plt.show()
if __name__ == '__main__':
main()