-
Notifications
You must be signed in to change notification settings - Fork 1
/
model.js
1810 lines (1509 loc) · 51 KB
/
model.js
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
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
"use strict";
async function except (errname, e) {
$(".overlay").remove();
await write_descriptions();
await enable_everything();
if(Object.keys(e).includes("message")) {
e = e.message;
}
wrn(errname + ": " + e + ". Resetting model.");
console.trace();
await write_error(e);
throw new Error(e);
}
async function get_model_config_hash () {
var arr = [];
$("#layers_container").find("input, checkbox, select").each(function (i, x) {
if($(x).attr("type") == "checkbox") {
arr.push($(x).is(":checked"));
} else {
arr.push($(x).val());
}
});
var str = arr.join(";;;;;;;;;");
var res = await md5(str);
return res;
}
async function _create_model () {
var _create_model_uuid = uuidv4();
while (create_model_queue.length) {
await delay(50);
}
create_model_queue.push(_create_model_uuid);
if(has_missing_values) {
l(language[lang]["not_creating_model_because_values_are_missing"]);
return model;
}
try {
try {
if(global_model_data) {
var model_data_tensors = find_tensors_with_is_disposed_internal(global_model_data);
for (var i = 0; i < model_data_tensors.length; i++) {
await dispose(model_data_tensors[i]);
}
}
if(model && Object.keys(model).includes("layers") && model.layers.length) {
for (var i = 0; i < model.layers.length; i++) {
await dispose(model.layers[i].bias);
await dispose(model.layers[i].kernel);
}
await dispose(model);
}
} catch (e) {
if(Object.keys(e).includes("message")) {
e = e.message;
}
err(e);
}
[model, global_model_data] = await create_model(model);
if(can_be_shown_in_latex()) {
$("#math_mode_settings").show();
} else {
$("#math_mode_settings").hide();
}
} catch (e) {
if(Object.keys(e).includes("message")) {
e = e.message;
}
create_model_queue = create_model_queue.filter(function(e) { return e !== _create_model_uuid; });
if(("" + e).includes("undefined has no properties")) {
wrn("[create_model] Trying to work on undefined model. This may be the case when this function is called, but the model is currently being rebuilt.");
return;
} else if(("" + e).includes("Input 0 is incompatible with layer")) {
throw new Error("[create_model] " + e);
} else if(("" + e).includes("BaseConv expects config.kernelSize to be number")) {
throw new Error("[create_model] " + e);
} else if(("" + e).includes("targetShape is undefined")) {
wrn("[create_model] " + e);
} else if(("" + e).includes("ReferenceError")) {
wrn("[create_model] " + e);
} else if(("" + e).includes("The channel dimension of the input should be defined")) {
wrn("[create_model] " + e);
} else if(("" + e).includes("model is undefined")) {
wrn("[create_model] Currently, the model is undefined. This may be fatal, but may also not be");
} else if(("" + e).includes("model.layers[i] is undefined")) {
wrn("[create_model] " + e);
} else if(("" + e).includes("Inputs to DepthwiseConv2D should have rank") || ("" + e).includes("Inputs to SeparableConv2D should have rank")) {
wrn("[create_model] " + e);
} else if(("" + e).includes("Cannot read properties of undefined (reading 'layers')")) {
wrn("[create_model] " + e);
return;
} else if(("" + e).includes("Cannot read properties of undefined")) {
wrn("[create_model] " + e);
return;
} else if(("" + e).includes("identifier starts immediately after numeric literal")) {
wrn("[create_model] " + e);
return;
} else if(
("" + e).includes("Convolution layer expected config.filters to be a 'number' > 0 but got undefined") ||
("" + e).includes("The kernelSize argument must be an integer or tuple of 2 integers") ||
("" + e).includes("The strides argument must be an integer or tuple of 2 integers") ||
("" + e).includes("Expected units to be a positive integer, but got undefined") ||
("" + e).includes("have a defined dimension but the layer received an input with shape")
) {
wrn("[create_model] " + e);
return;
} else if (("" + e).includes("Improper config format")) {
err(`[create_model] ${e}`);
return;
} else {
await except("ERROR1", "" + e);
if(mode == "beginner") {
Swal.fire({
icon: "error",
title: "Oops [4]...",
text: "" + e
});
} else {
l("ERROR: " + e);
}
}
}
create_model_queue = create_model_queue.filter(function(e) { return e !== _create_model_uuid; });
if(!disable_layer_debuggers && model) {
add_layer_debuggers();
}
//add_optimizer_debugger();
}
async function _get_recreate_model(current_status_hash, model_config_hash, new_model_config_hash) {
var recreate_model = false;
if(model_config_hash != new_model_config_hash || current_status_hash != await get_current_status_hash()) {
recreate_model = true;
}
if(model_is_trained) {
if(model_config_hash == new_model_config_hash) {
recreate_model = false;
} else {
recreate_model = true;
if(recreate_model) {
model_is_trained = false;
}
}
}
return recreate_model;
}
function find_tensors_with_is_disposed_internal(obj, tensorList = []) {
if (typeof obj === "object") {
if (obj.isDisposedInternal !== undefined) {
tensorList.push(obj);
}
for (const key in obj) {
find_tensors_with_is_disposed_internal(obj[key], tensorList);
}
}
return tensorList;
}
async function compile_model (recursion_level=0) {
if(recursion_level > 3) {
err("recursion level for compile_model too high");
return;
}
assert(get_number_of_layers() >= 1, "Need at least 1 layer.");
var new_model_config_hash = await get_model_config_hash();
assert(typeof(new_model_config_hash) == "string", "new model config has is not a string");
var recreate_model = await _get_recreate_model(current_status_hash, model_config_hash, new_model_config_hash);
if(!model) {
if(finished_loading) {
wrn("model not given");
}
if(global_model_data) {
var model_data_tensors = find_tensors_with_is_disposed_internal(global_model_data);
for (var i = 0; i < model_data_tensors.length; i++) {
await dispose(model_data_tensors[i]);
}
}
try {
[model, global_model_data] = await create_model(model, await get_model_structure());
} catch (e) {
throw new Error(e);
}
}
if(recreate_model) {
model_is_trained = false;
reset_summary();
await _create_model();
await last_shape_layer_warning();
}
if(!model) {
wrn("[compile_model] No model to compile!");
return;
}
while (create_model_queue.length || !model) {
await delay(10);
}
if (!global_model_data) {
global_model_data = await get_model_data();
}
if (!global_model_data) {
wrn("global_model_data is empty!");
}
try {
model.compile(global_model_data);
model_config_hash = new_model_config_hash;
} catch (e) {
if(Object.keys(e).includes("message")) {
e = e.message;
}
if(("" + e).includes("model is empty")) {
set_model_layer_warning(0, "" + e);
for (var i = 0; i < $("#layer_setting").length; i++) {
set_layer_background(i, "red");
}
} else if (("" + e).includes("model is empty")) {
err("[compile_model] " + e);
return;
} else if (("" + e).includes("e is null")) {
err("[compile_model] " + e);
await delay(1000);
return await compile_model(recursion_level + 1);
} else if (("" + e).includes("model.compile is not a function")) {
err("[compile_model] " + e);
return;
} else {
if(e) {
err("" + e);
} else {
await except("ERROR2", "Unknown error");
}
return;
}
}
for (var i = 0; i < $("#layer_setting").length; i++) {
set_layer_background(i, "");
}
try {
$("#outputShape").val(JSON.stringify(model.outputShape));
} catch (e) {
if(("" + e).includes("model is undefined")) {
wrn("[compile_model] model is undefined while compile_model");
} else {
throw new Error(e);
}
}
write_model_summary_wait();
}
function get_weight_type_name_from_option_name (on) {
if(typeof(on) != "string") {
wrn(`[get_weight_type_name_from_option_name] get_weight_type_name_from_option_name(on = ${on}), typeof(on) = ${typeof(on)}`);
return;
}
if(on.match(/_/)) {
for (var i = 0; i < valid_initializer_types.length; i++) {
var v = valid_initializer_types[i];
var re = new RegExp("^" + v + "(?:_.*)?$");
if(on.match(re)) {
return v;
}
}
} else {
return on;
}
return on;
}
function get_data_for_layer (type, i, first_layer) {
assert(typeof(type) == "string", type + " is not a string but " + typeof(type));
assert(typeof(i) == "number", i + " is not a number but " + typeof(i));
assert(typeof(first_layer) == "boolean", first_layer + " is not a boolean but " + typeof(first_layer));
var data = {
"name": type + "_" + (i + 1)
};
if(i == 0 || first_layer) {
data["inputShape"] = get_input_shape();
}
for (var j = 0; j < layer_options[type]["options"].length; j++) {
var option_name = layer_options[type]["options"][j];
assert(typeof(option_name) == "string", option_name + " is not string but " + typeof(option_name));
if(["pool_size", "kernel_size", "strides"].includes(option_name)) {
if(type.endsWith("1d")) {
data[get_js_name(option_name)] = [parse_int(get_item_value(i, option_name + "_x"))];
} else if(type.endsWith("2d")) {
data[get_js_name(option_name)] = [parse_int(get_item_value(i, option_name + "_x")), parse_int(get_item_value(i, option_name + "_y"))];
} else if(type.endsWith("3d")) {
data[get_js_name(option_name)] = [parse_int(get_item_value(i, option_name + "_x")), parse_int(get_item_value(i, option_name + "_y")), parse_int(get_item_value(i, option_name + "_z"))];
} else if(type.endsWith("2dTranspose")) {
data[get_js_name(option_name)] = [parse_int(get_item_value(i, option_name + "_x")), parse_int(get_item_value(i, option_name + "_y"))];
} else {
alert("Unknown layer type: " + type);
}
} else if(["trainable", "use_bias"].includes(option_name) ) {
try {
data[get_js_name(option_name)] = get_item_value(i, option_name);
} catch (e) {
if(Object.keys(e).includes("message")) {
e = e.message;
}
if(("" + e).includes("identifier starts immediately after numeric literal")) {
err("" + e);
} else {
throw new Error(e);
}
}
} else if(["size", "dilation_rate"].includes(option_name)) {
var dil_rate = get_item_value(i, option_name);
dil_rate = dil_rate.replace(/[^0-9,]/g, "");
var code_str = "[" + dil_rate + "]";
data[get_js_name(option_name)] = eval(code_str);
} else if(option_name == "rate") {
data["rate"] = parse_float(get_item_value(i, "dropout"));
} else if(["epsilon", "momentum", "dropout_rate"].includes(option_name)) {
data[get_js_name(option_name)] = parse_float(get_item_value(i, option_name));
} else if(option_name == "activation" && $($($($(".layer_setting")[i]).find("." + option_name)[0])).val() == "None") {
// Do nothing if activation = None
data["activation"] = null;
} else if (valid_initializer_types.includes(get_key_name_camel_case(get_weight_type_name_from_option_name(option_name))) && option_name.includes("nitializer")) {
var weight_type = get_weight_type_name_from_option_name(option_name);
var initializer_name = get_item_value(i, weight_type + "_initializer");
if(initializer_name) {
var initializer_config = get_layer_initializer_config(i, weight_type);
var initializer_config_string = JSON.stringify(initializer_config);
data[get_key_name_camel_case(weight_type) + "Initializer"] = {"name": initializer_name, "config": initializer_config};
}
} else if (valid_initializer_types.includes(get_key_name_camel_case(get_weight_type_name_from_option_name(option_name))) && option_name.includes("egularizer")) {
var weight_type = get_weight_type_name_from_option_name(option_name);
var regularizer_name = get_item_value(i, weight_type + "_regularizer");
if(regularizer_name) {
var regularizer_config = get_layer_regularizer_config(i, weight_type);
var regularizer_config_string = JSON.stringify(regularizer_config);
data[get_key_name_camel_case(weight_type) + "Regularizer"] = {"name": regularizer_name, "config": regularizer_config};
}
} else {
var elem = $($($(".layer_setting")[i]).find("." + option_name)[0]);
var value = $(elem).val();
if($(elem).is(":checkbox")) {
data[get_js_name(option_name)] = value == "on" ? true : false;
} else {
if(value == "") {
if(!option_name.includes("constraint")) {
wrn("[get_data_for_layer] Something may be wrong here! Value for '" + option_name.toString() + "' is ''");
}
} else {
data[get_js_name(option_name)] = is_numeric(value) ? parse_float(value) : value;
}
}
}
}
delete data["visualize"];
return data;
}
async function get_model_structure(is_fake_model = 0) {
var new_current_status_hash = "";
var first_layer = true; // seperate from i because first layer may be input layer (which is not a "real" layer)
var structure = [];
var num_of_layers = get_number_of_layers();
assert(num_of_layers >= 1, "number of layers must be at least 1 or more");
for (var i = 0; i < num_of_layers; i++) {
var layer_type = $($($(".layer_setting")[i]).find(".layer_type")[0]);
var type = $(layer_type).val();
if(typeof(type) !== "undefined" && type) {
var data = get_data_for_layer(type, i, first_layer);
try {
var layer_info = {
"type": type,
"data": data
};
structure.push(layer_info);
first_layer = false;
} catch (e) {
wrn("[get_model_structure] Failed to add layer type ", type, ": ", e);
header("DATA:");
log(data);
$($(".warning_container")[i]).show();
$($(".warning_layer")[i]).html(e);
}
traindebug("tf.layers." + type + "(", data, ")");
} else {
if(finished_loading) {
wrn(`get_model_structure is empty for layer ${i}`);
}
}
}
await write_descriptions();
layer_structure_cache = JSON.stringify(structure);
return structure;
}
function is_number_array (value) {
if(typeof(value) == "object") {
for (var i = 0; i < value.length; i++) {
if(typeof(value[i]) != "number") {
return false;
}
}
return true;
}
return false;
}
function is_valid_parameter (keyname, value, layer) {
assert(typeof(keyname) == "string", "keyname " + keyname + " is not a string but " + typeof(keyname));
assert(["string", "number", "boolean", "object"].includes(typeof(value)), value + " is not a string/number/boolean but " + typeof(value));
assert(typeof(layer) == "number", layer + " is not a number but " + typeof(layer));
if(
(["units", "filters"].includes(keyname) && typeof(value) == "number") ||
(["kernelRegularizer", "biasRegularizer", "activityRegularizer", "kernelInitializer", "biasInitializer", "gammaInitializer", "gammaRegularizer", "betaInitializer"].includes(keyname) && (typeof(value) == "object") || ["zeros", "ones"].includes(value)) ||
(["unitForgetBias", "center", "scale", "unroll", "trainable", "useBias", "stateful", "returnSequences", "returnState", "goBackwards"].includes(keyname) && typeof(value) == "boolean") ||
(["name", "betaConstraint", "gammaConstraint"].includes(keyname) && typeof(value) == "string") ||
(["recurrentInitializer", "depthwiseInitializer", "pointwiseInitializer", "movingMeanInitializer", "movingVarianceInitializer", "betaInitializer", "gammaInitializer"].includes(keyname) && ["constant", "glorotNormal", "glorotUniform", "heNormal", "heUniform", "identity", "leCunNormal", "leCunUniform", "ones", "orthogonal", "randomNormal", "randomUniform", "truncatedNormal", "varianceScaling", "zeros", "string", "l1", "l2", "l1l2"].includes(value)) ||
(keyname == "dtype" && ["float32", "int32", "bool", "complex64", "string"].includes(value)) ||
(keyname == "padding" && ["valid", "same", "causal"].includes(value)) ||
(["activation", "recurrentActivation"].includes(keyname) && ["LeakyReLU", "elu", "hardSigmoid", "linear", "relu", "relu6", "selu", "sigmoid", "softmax", "softplus", "softsign", "tanh", "swish", "mish"].includes(value)) ||
(["kernelSize", "poolSize", "strides", "dilationRate", "size"].includes(keyname) && (is_number_array(value) || typeof(value) == "number")) ||
(keyname == "implementation" && [1, 2].includes(value)) ||
(keyname == "biasConstraint" && ["maxNorm", "minNorm"].includes(value)) ||
(keyname == "interpolation" && ["nearest", "bilinear"].includes(value)) ||
(keyname == "inputShape" && layer == 0 && (typeof(value) == "object" || is_number_array(value))) ||
(keyname == "targetShape" && is_number_array(value)) ||
(["alpha", "stddev", "depthMultiplier"].includes(keyname) && typeof(value) == "number") ||
(keyname == "axis" && typeof(value) == "number" && parse_int(value) == value) ||
(["recurrentDropout", "dropout", "rate", "dropout_rate"].includes(keyname) && typeof(value) == "number" && value >= 0 && value <= 1) ||
(["epsilon"].includes(keyname) && typeof(value) == "number" && value >= 0) ||
(["theta"].includes(keyname) && typeof(value) == "number") ||
(["maxValue", "momentum"].includes(keyname) && typeof(value) == "number") ||
(["seed"].includes(keyname) && typeof(value) == "number") ||
(["cell"].includes(keyname) && typeof(value).includes("object"))
) {
return true;
}
//log("keyname: ", keyname, "value: ", value, "layer:", layer);
return false;
}
function get_key_name_camel_case(keyname) {
assert(typeof(keyname) == "string", `keyname "${keyname}" is not a string, but ${typeof(keyname)}`);
var letters = keyname.split("");
var results = [];
var next_is_camel_case = false;
for (var i = 0; i < letters.length; i++) {
if(letters[i] == "_") {
next_is_camel_case = true;
} else {
if(next_is_camel_case) {
results.push(letters[i].toUpperCase());
next_is_camel_case = false;
} else {
results.push(letters[i]);
}
}
}
return results.join("");
}
function remove_empty(obj) {
var res = Object.fromEntries(Object.entries(obj).filter(([_, v]) => v != null));
return res;
}
async function get_html_from_model () {
var html = "";
html += "<html>" + "\n";
html += " <head>\n";
html += " <meta charset='UTF-8'>\n";
html += " <title>Example Network</title>\n";
html += ` <script src='https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@${tf.version["tfjs-core"]}/dist/tf.min.js'></script>\n`;
html += ` <script src='https://code.jquery.com/jquery-${$().jquery}.js'></script>\n`;
html += " <!--<link href='main.css' rel='stylesheet' />-->\n";
html += " </head>\n";
html += " <body>\n";
html += " <script type='text/javascript'>\n";
html += " var model;\n";
html += " var labels = ['" + labels.join("', '") + "'];\n";
html += " var divide_by = " + $("#divide_by").val() + ";\n";
html += " async function load_model () {\n";
html += " model = await tf.loadLayersModel('./model.json');\n";
html += " }\n";
var input_shape_is_image_val = await input_shape_is_image();
if(input_shape_is_image_val) {
html += " var load_file = (function(event) {\n";
html += " var output = document.getElementById('output');\n";
html += " $('#output').removeAttr('src');\n";
html += " output.src = URL.createObjectURL(event.target.files[0]);\n";
html += " output.onload = async function() {\n";
html += " await load_model();\n";
html += " URL.revokeObjectURL(output.src);\n";
html += " var img = $('#output')[0];\n";
html += " img.height = model.layers[0].input.shape[1];\n";
html += " img.width = model.layers[0].input.shape[2];\n";
html += " var tensor = tf.browser.fromPixels(img);\n";
html += " tensor = tf.divNoNan(tensor, divide_by);\n";
html += " var results_tensor = await model.predict(tensor.expandDims());\n";
html += " var results = results_tensor.dataSync();\n";
html += " var html = '<pre>';\n";
html += " for (var i = 0; i < results.length; i++) {\n";
html += " var label = labels[i % labels.length];\n";
html += " html += label + ': ' + results[i] + \"\\n\";\n";
html += " }\n";
html += " html += '</pre>';\n";
html += " $('#results').html(html);\n";
html += " $('#results_container').show();\n";
html += " };\n";
html += " $('#output').show();\n";
html += " });\n";
} else {
html += " async function predict() {\n";
html += " await load_model();\n";
html += " var input = $('#inputtensor').val()\n";
html += " var tensor = tf.tensor(eval(input));\n";
html += " tensor = tf.divNoNan(tensor, divide_by);\n";
html += " var prediction_tensor = await model.predict(tensor);\n";
html += " var results = await prediction_tensor.dataSync();\n";
html += " var html = '<pre>';\n";
html += " for (var i = 0; i < results.length; i++) {\n";
html += " var label = labels[i % labels.length];\n";
html += " if(label) {\n";
html += " html += label + ': ' + results[i] + \"\\n\";\n";
html += " } else {\n";
html += " html += results[i] + '\\n';\n";
html += " }\n";
html += " }\n";
html += " html += '</pre>';\n";
html += " $('#results').html(html);\n";
html += " }\n";
}
html += " </script>\n";
if(input_shape_is_image_val) {
html += " <input type='file' id='upload_img' onchange='load_file(event)' />\n";
html += " <div id='results_container' style='display: none'>\n";
html += " <img id='output' />\n";
html += " <div id='results'></div>\n";
html += " </div>\n";
} else {
html += " <textarea style='width: 500px; height: 200px;' id='inputtensor'></textarea><br>\n";
html += " <button onclick='predict()'>Predict</button>\n";
html += " <div id='results'></div>\n";
html += " <script>\n";
html += " async function write_placeholder() {\n";
html += " await load_model();\n";
html += " var shape = model.layers[0].input.shape;\n";
html += " shape.shift();\n";
html += " $('#inputtensor').attr('placeholder', 'Shape: [[' + shape.join(', ') + ']]');\n";
html += " }\n";
html += " write_placeholder();\n";
html += " </script>\n";
}
html += " </body>\n";
html += "</html>" + "\n";
return html;
}
function check_initializers (data, has_keys) {
valid_initializer_types.forEach((init_or_regularizer_type) => {
["Regularizer", "Initializer"].forEach((regularizer_or_init) => {
var keyname = get_key_name_camel_case(init_or_regularizer_type + regularizer_or_init);
if(regularizer_or_init == "Initializer") {
if(has_keys.includes(keyname)) {
var original_name = data[keyname]["name"];
if(typeof(original_name) == "string") {
var options_stringified = JSON.stringify(data[keyname]["config"]);
if(original_name) {
var execute_this_string = `tf.initializers.${original_name}(${options_stringified})`;
try {
data[keyname] = eval(execute_this_string);
} catch (e) {
err("Error: ", e, "execute_this_string:", execute_this_string);
console.trace();
}
} else {
data[keyname] = null;
}
//} else {
// log("original_name (A):", original_name);
}
}
} else if(regularizer_or_init == "Regularizer") {
if(has_keys.includes(keyname)) {
var original_name = data[keyname]["name"];
assert(typeof(original_name) == "string", "original_name is not string (B)");
var options_stringified = JSON.stringify(data[keyname]["config"]);
if(typeof(original_name) == "string") {
if(original_name && original_name != "none") {
try {
data[keyname] = eval(`tf.regularizers.${original_name}(${options_stringified})`);
} catch (e) {
err(e);
console.trace();
}
} else {
data[keyname] = null;
}
//} else {
// log("original_name (B):", original_name);
}
}
} else {
log("Invalid regularizer_or_init: " + regularizer_or_init);
}
});
});
return data;
}
function _check_data (data, type) {
if(!data) {
err("Data is undefined");
return;
}
var has_keys = Object.keys(data);
try {
if(has_keys.includes("dropout_rate") && type == "dropout") {
var tmp = data["dropout_rate"];
delete data["dropout_rate"];
data["rate"] = tmp;
has_keys = Object.keys(data);
}
} catch (e) {
err(e);
}
try {
if(typeof(data) == "object" && ["lstm", "gru", "simpleRNN"].includes(type) && has_keys.includes("rate")) {
var tmp = data["rate"];
delete data["rate"];
data["dropout"] = tmp;
has_keys = Object.keys(data);
}
} catch (e) {
err(e);
}
try {
if(typeof(data) == "object" && "targetShape" in data && ["string", "number"].includes(typeof(data["targetShape"]))) {
data["targetShape"] = eval("[" + data["targetShape"] + "]");
}
} catch (e) {
err(e);
}
try {
if(typeof(data) == "object" && "size" in data && typeof(data["size"]) == "string") {
data["size"] = eval("[" + data["size"] + "]");
}
} catch (e) {
err(e);
}
try {
if(typeof(data) == "object" && "dilationRate" in data && data["dilationRate"].length == 0) {
data["dilationRate"] = null;
}
} catch (e) {
err(e);
}
try {
if(typeof(data) == "object" && "units" in data && typeof(data["units"]) == "undefined") {
if(finished_loading) {
wrn("[_check_data] units was not defined. Using 2 as default");
}
data["units"] = 2;
}
} catch (e) {
err(e);
}
try {
["strides", "kernelSize"].forEach(function (correction_name) {
if(typeof(data) == "object" && correction_name in data && (isNaN(data[correction_name][0]) || typeof(data[correction_name][0]) == "undefined")) {
data[correction_name] = [];
for (var k = 0; k < data[correction_name].length; k++) {
data[correction_name][k] = 1;
}
}
});
} catch (e) {
err(e);
}
try {
data = check_initializers(data, has_keys);
} catch (e) {
err(e);
}
try {
if(type == "rnn") {
// never worked...
var lstm_cells = [];
for (var index = 0; index < data["units"]; index++) {
lstm_cells.push(tf.layers.RNNCell({units: data["units"]}));
}
data["cell"] = lstm_cells;
log(data);
}
} catch (e) {
err(e);
}
try {
data = remove_empty(data);
} catch (e) {
err(e);
}
return data;
}
async function _add_layer_to_model (type, data, fake_model_structure, i, new_model, model_uuid) {
try {
if(layer_options[type]["custom"]) {
if(i == 0) {
data["inputShape"] = get_input_shape();
} else {
delete data["inputShape"];
}
var model_add_code = `new_model.add(new ${type}(${JSON.stringify(data)}))`;
eval(model_add_code);
} else {
//console.log("adding ", tf.layers[type], ", data: ", data);
var new_layer = tf.layers[type](data);
new_model.add(new_layer);
var added_layer = new_model.layers[new_model.layers.length - 1];
if(added_layer["bias"]) {
_custom_tensors["" + added_layer.bias.id] = ["UUID:" + model_uuid, added_layer.bias, "[bias in _add_layer_to_model]"];
_clean_custom_tensors();
}
if(added_layer["kernel"]) {
_custom_tensors["" + added_layer.kernel.id] = ["UUID:" + model_uuid, added_layer.kernel, "[kernel in _add_layer_to_model]"];
_clean_custom_tensors();
}
if(new_model && new_model.layers) {
var new_output_shape = new_model.layers[new_model.layers.length - 1].getOutputAt(0).shape;
if(new_output_shape) {
for (j in new_output_shape) {
if(new_output_shape[j] === 0) {
if(new_output_shape.shape) {
log("New output-shape:", new_output_shape.shape);
}
throw new Error("Input shape contains 0 at layer " + j);
}
}
try {
var new_output_shape = new_model.layers[new_model.layers.length - 1].getOutputAt(1);
throw new Error(`Layer ${i} has more than one output head!`);
} catch (e) {
if(("" + e).includes("Has Multi-Output")) {
throw new Error(e);
}
}
}
}
}
set_layer_background(i, "");
} catch (e) {
if(Object.keys(e).includes("message")) {
e = e.message;
}
/*
err("!!!!!!!!!!!!!!!!");
err("" + e);
*/
if(!fake_model_structure && !("" + e).includes("nodeIndex is not a number")) { // "nodeIndex is not a number" means the model has only one output node, which is good
if(
("" + e).includes("Negative dimension size caused by adding layer") ||
("" + e).includes("Has Multi-Output") ||
("" + e).includes("Input shape contains 0") ||
("" + e).includes("is incompatible with layer") ||
("" + e).includes("targetShape is undefined") ||
("" + e).includes("is not fully defined") ||
("" + e).includes("The dilationRate argument must be an integer") ||
("" + e).includes("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument")
) {
set_layer_background(i, "red");
set_model_layer_warning(i, "" + e);
} else {
set_model_layer_warning(i, "" + e);
l("ERROR: " + e);
console.log("ORIGINAL e: ", e);
log(type);
log(data);
throw new Error(e);
}
await dispose(new_model);
}
return false;
}
return new_model;
}
function _set_layer_gui (data, fake_model_structure, i) {
assert(typeof(data) == "object", "data is not an object");
assert(typeof(i) == "number", "i is not a number");
var data_keys = Object.keys(data);
for (var k = 0; k < data_keys.length; k++) {
var this_key = data_keys[k];
var layer_setting = $($(".layer_setting")[i]);
var current_setting = layer_setting.find("." + js_names_to_python_names[this_key]);
if(!fake_model_structure && !is_valid_parameter(this_key, data[this_key], i)) {
header("=================");
log(`INVALID PARAMETER FOR LAYER ${i}: ` + this_key + ": ", data[this_key], " (" + typeof(data[this_key]) + ")");
header("<<<<<<<<<<<<<<<<<");
current_setting.css("background-color", "red");
} else {
current_setting.css("background-color", "");
}
}
}
async function create_model (old_model, fake_model_structure, force) {
if(has_missing_values) {
l("Not creating model because some values are missing (create model)");
if(old_model) {
return old_model;
}
err("No model found, but has missing values");
return [old_model, null];
}
var new_layers_container_md5 = await get_layers_container_md5();
if(!layers_container_md5 || force) {
layers_container_md5 = new_layers_container_md5;
}
var new_current_status_hash = await get_current_status_hash(fake_model_structure ? 0 : 1);
if(!force && disable_show_python_and_create_model) {
return [old_model, null];
}
current_status_hash = new_current_status_hash;
$(".warning_container").hide();
var model_structure = fake_model_structure;
if(model_structure === undefined) {
model_structure = await get_model_structure();
}
assert(typeof(model_structure) == "object", "model_structure is not an object");
var model_uuid = "";
if(fake_model_structure) {
model_uuid = "FAKE_MODEL";
} else {
model_uuid = uuidv4();
}
var new_model;
try {
new_model = await _add_layers_to_model(model_structure, fake_model_structure, i, model_uuid);
await dispose_old_model_tensors(model_uuid);
} catch (e) {
if(("" + e).includes("Negative dimension size caused by adding layer")) {
wrn(`[create_model] Trying to add the layer ${i} failed, probably because the input size is too small or there are too many stacked layers.`);
} else if(("" + e).includes("Input shape contains 0")) {
wrn("[create_model] " + e);
} else if(("" + e).includes("is not fully defined")) {
wrn("[create_model] " + e);
return;
} else if(("" + e).includes("Input 0 is incompatible with layer")) {
wrn("[create_model] Model could not be created because of problems with the input layer.");
return;
} else {
throw new Error("[create_model] " +e);
}
return;
}
$(".warning_container").html("").hide();
if(model && model.layers && model.layers.length) {
if(i in model.layers) {
try {
var ok = 1;
try {
model.layers[i].input.shape;
ok = 0;
} catch (er) { // ignore delibaretly, when it fails, its ok
wrn("" + er);
}
if(!ok) {
throw new Error(`model.layers[${i}] is a multibound head`);
}
} catch(e) {
err(e);
wrn("Model has multi-node inputs. It should not have!!! Continuing anyway, but please, debug this!!!");
}
}
}