projectlosangeles commited on
Commit
bf0b976
1 Parent(s): 0dce85b

Upload 11 files

Browse files
Master_MIDI_Dataset_GPU_Search_and_Filter.ipynb CHANGED
@@ -11,7 +11,7 @@
11
  "id": "SiTIpPjArIyr"
12
  },
13
  "source": [
14
- "# Master MIDI Dataset GPU Search and Filter (ver. 2.0)\n",
15
  "\n",
16
  "***\n",
17
  "\n",
@@ -119,10 +119,15 @@
119
  "import statistics\n",
120
  "import shutil\n",
121
  "\n",
 
 
 
122
  "import cupy as cp\n",
123
  "\n",
124
  "from huggingface_hub import hf_hub_download\n",
125
  "\n",
 
 
126
  "print('Loading TMIDIX module...')\n",
127
  "os.chdir('/content/Los-Angeles-MIDI-Dataset')\n",
128
  "\n",
@@ -257,17 +262,15 @@
257
  "random.shuffle(sigs_data)\n",
258
  "\n",
259
  "signatures_file_names = []\n",
260
- "sigs_matrixes = [ [0]*(len(TMIDIX.ALL_CHORDS)+128) for i in range(len(sigs_data))]\n",
261
  "\n",
262
  "idx = 0\n",
263
  "for s in tqdm(sigs_data):\n",
264
  "\n",
265
  " signatures_file_names.append(s[0])\n",
266
  "\n",
267
- " counts_sum = sum([c[1] for c in s[1]])\n",
268
- "\n",
269
  " for ss in s[1]:\n",
270
- " sigs_matrixes[idx][ss[0]] = ss[1] / counts_sum\n",
271
  "\n",
272
  " idx += 1\n",
273
  "\n",
@@ -305,10 +308,15 @@
305
  "\n",
306
  "#@markdown NOTE: You can stop the search at any time to render partial results\n",
307
  "\n",
308
- "number_of_top_matches_MIDIs_to_collect = 20 #@param {type:\"slider\", min:5, max:50, step:1}\n",
309
- "search_matching_type = \"ratios\" # @param [\"ratios\", \"distances\"]\n",
 
 
 
 
310
  "distances_norm_order = 3 # @param {type:\"slider\", min:1, max:10, step:1}\n",
311
- "maximum_match_ratio_to_search_for = 0.999 #@param {type:\"slider\", min:0, max:1, step:0.001}\n",
 
312
  "\n",
313
  "print('=' * 70)\n",
314
  "print('Master MIDI Dataset GPU Search and Filter')\n",
@@ -338,8 +346,8 @@
338
  "\n",
339
  " ###################\n",
340
  "\n",
341
- " if not os.path.exists('/content/Output-MIDI-Dataset'):\n",
342
- " os.makedirs('/content/Output-MIDI-Dataset')\n",
343
  "\n",
344
  " ###################\n",
345
  "\n",
@@ -368,6 +376,8 @@
368
  " e[1] = int(e[1] / 16)\n",
369
  " e[2] = int(e[2] / 16)\n",
370
  "\n",
 
 
371
  " src_sigs = []\n",
372
  "\n",
373
  " for i in range(-6, 6):\n",
@@ -380,10 +390,12 @@
380
  " cscore = TMIDIX.chordify_score([1000, escore_copy])\n",
381
  "\n",
382
  " sig = []\n",
 
383
  "\n",
384
  " for c in cscore:\n",
385
  "\n",
386
  " pitches = sorted(set([p[4] for p in c if p[3] != 9]))\n",
 
387
  "\n",
388
  " if pitches:\n",
389
  " if len(pitches) > 1:\n",
@@ -397,20 +409,28 @@
397
  "\n",
398
  " sig.append(sig_token)\n",
399
  "\n",
400
- " fsig = [list(v) for v in Counter(sig).most_common()]\n",
 
 
 
 
 
401
  "\n",
402
- " src_sig_mat = [0] * (len(TMIDIX.ALL_CHORDS)+128)\n",
403
  "\n",
404
- " counts_sum = sum([c[1] for c in fsig])\n",
405
  "\n",
406
  " for s in fsig:\n",
407
  "\n",
408
- " src_sig_mat[s[0]] = s[1] / counts_sum\n",
409
  "\n",
410
  " src_sigs.append(src_sig_mat)\n",
411
  "\n",
412
  " src_signatures = cp.stack(cp.array(src_sigs))\n",
413
  "\n",
 
 
 
414
  " #=======================================================\n",
415
  "\n",
416
  " print('Searching for matches...Please wait...')\n",
@@ -432,18 +452,49 @@
432
  "\n",
433
  " for target_sig in tqdm(src_signatures):\n",
434
  "\n",
435
- " if search_matching_type == 'ratios':\n",
 
 
 
 
436
  "\n",
437
- " ratios = cp.where(target_sig != 0, cp.divide(cp.minimum(signatures_data, target_sig), cp.maximum(signatures_data, target_sig)), 0)\n",
438
- " max_comp_lengths = cp.maximum(cp.repeat(cp.sum(target_sig != 0), signatures_data.shape[0]), cp.sum(signatures_data != 0, axis=1))\n",
439
  "\n",
440
- " results = cp.divide(cp.sum(ratios, axis=1), max_comp_lengths)\n",
441
  "\n",
442
- " elif search_matching_type == 'distances':\n",
 
 
 
443
  "\n",
444
  " distances = cp.power(cp.sum(cp.power(cp.abs(signatures_data - target_sig), distances_norm_order), axis=1), 1 / distances_norm_order)\n",
445
  "\n",
446
- " results = cp.max(distances) - distances\n",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
447
  "\n",
448
  " unique_means = cp.unique(results)\n",
449
  " sorted_means = cp.sort(unique_means)[::-1]\n",
@@ -462,7 +513,19 @@
462
  "\n",
463
  " tv_idx += 1\n",
464
  "\n",
465
- " filtered_results = sorted(zip(all_filtered_means, all_filtered_idxs, all_filtered_tvs), key=lambda x: x[0], reverse=True)[:filter_size]\n",
 
 
 
 
 
 
 
 
 
 
 
 
466
  "\n",
467
  " #=======================================================\n",
468
  "\n",
@@ -487,7 +550,7 @@
487
  " #=======================================================\n",
488
  "\n",
489
  " dir_str = str(fn1)\n",
490
- " copy_path = '/content/Output-MIDI-Dataset/'+dir_str\n",
491
  " if not os.path.exists(copy_path):\n",
492
  " os.mkdir(copy_path)\n",
493
  "\n",
@@ -534,6 +597,75 @@
534
  "execution_count": null,
535
  "outputs": []
536
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
537
  {
538
  "cell_type": "markdown",
539
  "metadata": {
 
11
  "id": "SiTIpPjArIyr"
12
  },
13
  "source": [
14
+ "# Master MIDI Dataset GPU Search and Filter (ver. 5.0)\n",
15
  "\n",
16
  "***\n",
17
  "\n",
 
119
  "import statistics\n",
120
  "import shutil\n",
121
  "\n",
122
+ "import locale\n",
123
+ "locale.getpreferredencoding = lambda: \"UTF-8\"\n",
124
+ "\n",
125
  "import cupy as cp\n",
126
  "\n",
127
  "from huggingface_hub import hf_hub_download\n",
128
  "\n",
129
+ "from google.colab import files\n",
130
+ "\n",
131
  "print('Loading TMIDIX module...')\n",
132
  "os.chdir('/content/Los-Angeles-MIDI-Dataset')\n",
133
  "\n",
 
262
  "random.shuffle(sigs_data)\n",
263
  "\n",
264
  "signatures_file_names = []\n",
265
+ "sigs_matrixes = [ [0]*(len(TMIDIX.ALL_CHORDS)+256) for i in range(len(sigs_data))]\n",
266
  "\n",
267
  "idx = 0\n",
268
  "for s in tqdm(sigs_data):\n",
269
  "\n",
270
  " signatures_file_names.append(s[0])\n",
271
  "\n",
 
 
272
  " for ss in s[1]:\n",
273
+ " sigs_matrixes[idx][ss[0]] = ss[1]\n",
274
  "\n",
275
  " idx += 1\n",
276
  "\n",
 
308
  "\n",
309
  "#@markdown NOTE: You can stop the search at any time to render partial results\n",
310
  "\n",
311
+ "number_of_top_matches_MIDIs_to_collect = 30 #@param {type:\"slider\", min:5, max:50, step:1}\n",
312
+ "search_matching_type = \"Ratios\" # @param [\"Ratios\", \"Distances\", \"Correlations\"]\n",
313
+ "maximum_match_ratio_to_search_for = 1 #@param {type:\"slider\", min:0, max:1, step:0.001}\n",
314
+ "match_results_weight = 2 # @param {type:\"slider\", min:0.1, max:3, step:0.1}\n",
315
+ "match_lengths_weight = 1 # @param {type:\"slider\", min:0.1, max:3, step:0.1}\n",
316
+ "match_counts_weight = 1 # @param {type:\"slider\", min:0.1, max:3, step:0.1}\n",
317
  "distances_norm_order = 3 # @param {type:\"slider\", min:1, max:10, step:1}\n",
318
+ "epsilon = 0.5 # @param {type:\"slider\", min:0.001, max:1, step:0.001}\n",
319
+ "match_drums = False # @param {type:\"boolean\"}\n",
320
  "\n",
321
  "print('=' * 70)\n",
322
  "print('Master MIDI Dataset GPU Search and Filter')\n",
 
346
  "\n",
347
  " ###################\n",
348
  "\n",
349
+ " if not os.path.exists('/content/Output-MIDI-Dataset/'+search_matching_type):\n",
350
+ " os.makedirs('/content/Output-MIDI-Dataset/'+search_matching_type)\n",
351
  "\n",
352
  " ###################\n",
353
  "\n",
 
376
  " e[1] = int(e[1] / 16)\n",
377
  " e[2] = int(e[2] / 16)\n",
378
  "\n",
379
+ " drums_offset = len(TMIDIX.ALL_CHORDS) + 128\n",
380
+ "\n",
381
  " src_sigs = []\n",
382
  "\n",
383
  " for i in range(-6, 6):\n",
 
390
  " cscore = TMIDIX.chordify_score([1000, escore_copy])\n",
391
  "\n",
392
  " sig = []\n",
393
+ " dsig = []\n",
394
  "\n",
395
  " for c in cscore:\n",
396
  "\n",
397
  " pitches = sorted(set([p[4] for p in c if p[3] != 9]))\n",
398
+ " drums = sorted(set([p[4] for p in c if p[3] == 9]))\n",
399
  "\n",
400
  " if pitches:\n",
401
  " if len(pitches) > 1:\n",
 
409
  "\n",
410
  " sig.append(sig_token)\n",
411
  "\n",
412
+ " if drums:\n",
413
+ " dsig.extend(drums)\n",
414
+ "\n",
415
+ " sig_p = dict.fromkeys(sig+dsig, 0)\n",
416
+ " for item in sig+dsig:\n",
417
+ " sig_p[item] += 1\n",
418
  "\n",
419
+ " fsig = [list(v) for v in sig_p.items()]\n",
420
  "\n",
421
+ " src_sig_mat = [0] * (len(TMIDIX.ALL_CHORDS)+256)\n",
422
  "\n",
423
  " for s in fsig:\n",
424
  "\n",
425
+ " src_sig_mat[s[0]] = s[1]\n",
426
  "\n",
427
  " src_sigs.append(src_sig_mat)\n",
428
  "\n",
429
  " src_signatures = cp.stack(cp.array(src_sigs))\n",
430
  "\n",
431
+ " if not match_drums:\n",
432
+ " src_signatures = cp.where(src_signatures < drums_offset, src_signatures, epsilon)\n",
433
+ "\n",
434
  " #=======================================================\n",
435
  "\n",
436
  " print('Searching for matches...Please wait...')\n",
 
452
  "\n",
453
  " for target_sig in tqdm(src_signatures):\n",
454
  "\n",
455
+ " if not match_drums:\n",
456
+ " target_sig = cp.where(target_sig < drums_offset, target_sig, epsilon)\n",
457
+ "\n",
458
+ " comps_lengths = cp.vstack((cp.repeat(cp.sum(target_sig != 0), signatures_data.shape[0]), cp.sum(signatures_data != 0, axis=1)))\n",
459
+ " comps_lengths_ratios = cp.divide(cp.min(comps_lengths, axis=0), cp.max(comps_lengths, axis=0))\n",
460
  "\n",
461
+ " comps_counts_sums = cp.vstack((cp.repeat(cp.sum(target_sig), signatures_data.shape[0]), cp.sum(signatures_data, axis=1)))\n",
462
+ " comps_counts_sums_ratios = cp.divide(cp.min(comps_counts_sums, axis=0), cp.max(comps_counts_sums, axis=0))\n",
463
  "\n",
464
+ " if search_matching_type == 'Ratios':\n",
465
  "\n",
466
+ " ratios = cp.where(target_sig != 0, cp.divide(cp.minimum(signatures_data, target_sig), cp.maximum(signatures_data, target_sig)), epsilon)\n",
467
+ " results = cp.mean(ratios, axis=1)\n",
468
+ "\n",
469
+ " elif search_matching_type == 'Distances':\n",
470
  "\n",
471
  " distances = cp.power(cp.sum(cp.power(cp.abs(signatures_data - target_sig), distances_norm_order), axis=1), 1 / distances_norm_order)\n",
472
  "\n",
473
+ " distances_mean = cp.mean(distances)\n",
474
+ " distances_std = cp.std(distances)\n",
475
+ "\n",
476
+ " results = 1 - cp.divide((distances - distances_mean), distances_std)\n",
477
+ "\n",
478
+ " elif search_matching_type == 'Correlations':\n",
479
+ "\n",
480
+ " main_array_mean = cp.mean(signatures_data, axis=1, keepdims=True)\n",
481
+ " main_array_std = cp.std(signatures_data, axis=1, keepdims=True)\n",
482
+ " target_array_mean = cp.mean(target_sig)\n",
483
+ " target_array_std = cp.std(target_sig)\n",
484
+ "\n",
485
+ " signatures_data_normalized = cp.where(main_array_std != 0, (signatures_data - main_array_mean) / main_array_std, epsilon)\n",
486
+ " target_sig_normalized = cp.where(target_array_std != 0, (target_sig - target_array_mean) / target_array_std, epsilon)\n",
487
+ "\n",
488
+ " correlations = cp.divide(cp.einsum('ij,j->i', signatures_data_normalized, target_sig_normalized), (signatures_data.shape[1] - 1))\n",
489
+ " scaled_correlations = cp.divide(correlations, cp.sqrt(cp.sum(correlations**2)))\n",
490
+ " exp = cp.exp(scaled_correlations - cp.max(scaled_correlations))\n",
491
+ " results = cp.multiply(cp.divide(exp, cp.sum(exp)), 1e5)\n",
492
+ "\n",
493
+ " results_weight = match_results_weight\n",
494
+ " comp_lengths_weight = match_lengths_weight\n",
495
+ " comp_counts_sums_weight = match_counts_weight\n",
496
+ "\n",
497
+ " results = cp.divide(cp.add(cp.add(results_weight, comp_lengths_weight), comp_counts_sums_weight), cp.add(cp.add(cp.divide(results_weight, cp.where(results !=0, results, epsilon)), cp.divide(comp_lengths_weight, cp.where(comps_lengths_ratios !=0, comps_lengths_ratios, epsilon))), cp.divide(comp_counts_sums_weight, cp.where(comps_counts_sums_ratios !=0, comps_counts_sums_ratios, epsilon))))\n",
498
  "\n",
499
  " unique_means = cp.unique(results)\n",
500
  " sorted_means = cp.sort(unique_means)[::-1]\n",
 
513
  "\n",
514
  " tv_idx += 1\n",
515
  "\n",
516
+ " f_results = sorted(zip(all_filtered_means, all_filtered_idxs, all_filtered_tvs), key=lambda x: x[0], reverse=True)\n",
517
+ "\n",
518
+ " triplet_dict = {}\n",
519
+ "\n",
520
+ " for triplet in f_results:\n",
521
+ "\n",
522
+ " if triplet[0] not in triplet_dict:\n",
523
+ " triplet_dict[triplet[0]] = triplet\n",
524
+ " else:\n",
525
+ " if triplet[2] == 0:\n",
526
+ " triplet_dict[triplet[0]] = triplet\n",
527
+ "\n",
528
+ " filtered_results = list(triplet_dict.values())[:filter_size]\n",
529
  "\n",
530
  " #=======================================================\n",
531
  "\n",
 
550
  " #=======================================================\n",
551
  "\n",
552
  " dir_str = str(fn1)\n",
553
+ " copy_path = '/content/Output-MIDI-Dataset/'+search_matching_type+'/'+dir_str\n",
554
  " if not os.path.exists(copy_path):\n",
555
  " os.mkdir(copy_path)\n",
556
  "\n",
 
597
  "execution_count": null,
598
  "outputs": []
599
  },
600
+ {
601
+ "cell_type": "markdown",
602
+ "source": [
603
+ "# (DOWNLOAD RESULTS)"
604
+ ],
605
+ "metadata": {
606
+ "id": "7Lyy0vjV0dlI"
607
+ }
608
+ },
609
+ {
610
+ "cell_type": "code",
611
+ "source": [
612
+ "#@title Zip and download all search results\n",
613
+ "\n",
614
+ "print('=' * 70)\n",
615
+ "\n",
616
+ "try:\n",
617
+ " os.remove('Master_MIDI_Dataset_Search_Results.zip')\n",
618
+ "except OSError:\n",
619
+ " pass\n",
620
+ "\n",
621
+ "print('Zipping... Please wait...')\n",
622
+ "print('=' * 70)\n",
623
+ "\n",
624
+ "%cd /content/Output-MIDI-Dataset/\n",
625
+ "!zip -r Master_MIDI_Dataset_Search_Results.zip *\n",
626
+ "%cd /content/\n",
627
+ "\n",
628
+ "print('=' * 70)\n",
629
+ "print('Done!')\n",
630
+ "print('=' * 70)\n",
631
+ "\n",
632
+ "print('Downloading final zip file...')\n",
633
+ "print('=' * 70)\n",
634
+ "\n",
635
+ "files.download('/content/Output-MIDI-Dataset/Master_MIDI_Dataset_Search_Results.zip')\n",
636
+ "\n",
637
+ "print('Done!')\n",
638
+ "print('=' * 70)"
639
+ ],
640
+ "metadata": {
641
+ "cellView": "form",
642
+ "id": "1psdj0RJ0aWH"
643
+ },
644
+ "execution_count": null,
645
+ "outputs": []
646
+ },
647
+ {
648
+ "cell_type": "code",
649
+ "source": [
650
+ "# @title Delete search results directory and files\n",
651
+ "\n",
652
+ "#@markdown WARNING: This can't be undone so make sure you downloaded the search results first\n",
653
+ "\n",
654
+ "print('=' * 70)\n",
655
+ "print('Deleting... Please wait...')\n",
656
+ "print('=' * 70)\n",
657
+ "\n",
658
+ "!rm -rf /content/Output-MIDI-Dataset\n",
659
+ "print('Done!')\n",
660
+ "print('=' * 70)"
661
+ ],
662
+ "metadata": {
663
+ "cellView": "form",
664
+ "id": "z3B-YHIz0jDt"
665
+ },
666
+ "execution_count": null,
667
+ "outputs": []
668
+ },
669
  {
670
  "cell_type": "markdown",
671
  "metadata": {
TMIDIX.py CHANGED
@@ -1461,6 +1461,7 @@ import tqdm
1461
 
1462
  from itertools import zip_longest
1463
  from itertools import groupby
 
1464
 
1465
  from operator import itemgetter
1466
 
@@ -4538,6 +4539,29 @@ def check_and_fix_tones_chord(tones_chord):
4538
 
4539
  ###################################################################################
4540
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4541
  # This is the end of the TMIDI X Python module
4542
 
4543
  ###################################################################################
 
1461
 
1462
  from itertools import zip_longest
1463
  from itertools import groupby
1464
+ from collections import Counter
1465
 
1466
  from operator import itemgetter
1467
 
 
4539
 
4540
  ###################################################################################
4541
 
4542
+ def create_similarity_matrix(list_of_values, matrix_length=0):
4543
+
4544
+ counts = Counter(list_of_values).items()
4545
+
4546
+ if matrix_length > 0:
4547
+ sim_matrix = [0] * max(matrix_length, len(list_of_values))
4548
+ else:
4549
+ sim_matrix = [0] * len(counts)
4550
+
4551
+ for c in counts:
4552
+ sim_matrix[c[0]] = c[1]
4553
+
4554
+ similarity_matrix = [[0] * len(sim_matrix) for _ in range(len(sim_matrix))]
4555
+
4556
+ for i in range(len(sim_matrix)):
4557
+ for j in range(len(sim_matrix)):
4558
+ if max(sim_matrix[i], sim_matrix[j]) != 0:
4559
+ similarity_matrix[i][j] = min(sim_matrix[i], sim_matrix[j]) / max(sim_matrix[i], sim_matrix[j])
4560
+
4561
+ return similarity_matrix, sim_matrix
4562
+
4563
+ ###################################################################################
4564
+
4565
  # This is the end of the TMIDI X Python module
4566
 
4567
  ###################################################################################
master_midi_dataset_gpu_search_and_filter.py CHANGED
@@ -6,7 +6,7 @@ Automatically generated by Colaboratory.
6
  Original file is located at
7
  https://colab.research.google.com/github/asigalov61/Los-Angeles-MIDI-Dataset/blob/main/Extras/Master_MIDI_Dataset_GPU_Search_and_Filter.ipynb
8
 
9
- # Master MIDI Dataset GPU Search and Filter (ver. 2.0)
10
 
11
  ***
12
 
@@ -49,10 +49,15 @@ import pprint
49
  import statistics
50
  import shutil
51
 
 
 
 
52
  import cupy as cp
53
 
54
  from huggingface_hub import hf_hub_download
55
 
 
 
56
  print('Loading TMIDIX module...')
57
  os.chdir('/content/Los-Angeles-MIDI-Dataset')
58
 
@@ -142,17 +147,15 @@ print('=' * 70)
142
  random.shuffle(sigs_data)
143
 
144
  signatures_file_names = []
145
- sigs_matrixes = [ [0]*(len(TMIDIX.ALL_CHORDS)+128) for i in range(len(sigs_data))]
146
 
147
  idx = 0
148
  for s in tqdm(sigs_data):
149
 
150
  signatures_file_names.append(s[0])
151
 
152
- counts_sum = sum([c[1] for c in s[1]])
153
-
154
  for ss in s[1]:
155
- sigs_matrixes[idx][ss[0]] = ss[1] / counts_sum
156
 
157
  idx += 1
158
 
@@ -174,10 +177,15 @@ print('=' * 70)
174
 
175
  #@markdown NOTE: You can stop the search at any time to render partial results
176
 
177
- number_of_top_matches_MIDIs_to_collect = 20 #@param {type:"slider", min:5, max:50, step:1}
178
- search_matching_type = "ratios" # @param ["ratios", "distances"]
 
 
 
 
179
  distances_norm_order = 3 # @param {type:"slider", min:1, max:10, step:1}
180
- maximum_match_ratio_to_search_for = 0.999 #@param {type:"slider", min:0, max:1, step:0.001}
 
181
 
182
  print('=' * 70)
183
  print('Master MIDI Dataset GPU Search and Filter')
@@ -207,8 +215,8 @@ if filez:
207
 
208
  ###################
209
 
210
- if not os.path.exists('/content/Output-MIDI-Dataset'):
211
- os.makedirs('/content/Output-MIDI-Dataset')
212
 
213
  ###################
214
 
@@ -237,6 +245,8 @@ if filez:
237
  e[1] = int(e[1] / 16)
238
  e[2] = int(e[2] / 16)
239
 
 
 
240
  src_sigs = []
241
 
242
  for i in range(-6, 6):
@@ -249,10 +259,12 @@ if filez:
249
  cscore = TMIDIX.chordify_score([1000, escore_copy])
250
 
251
  sig = []
 
252
 
253
  for c in cscore:
254
 
255
  pitches = sorted(set([p[4] for p in c if p[3] != 9]))
 
256
 
257
  if pitches:
258
  if len(pitches) > 1:
@@ -266,20 +278,28 @@ if filez:
266
 
267
  sig.append(sig_token)
268
 
269
- fsig = [list(v) for v in Counter(sig).most_common()]
 
 
 
 
 
270
 
271
- src_sig_mat = [0] * (len(TMIDIX.ALL_CHORDS)+128)
272
 
273
- counts_sum = sum([c[1] for c in fsig])
274
 
275
  for s in fsig:
276
 
277
- src_sig_mat[s[0]] = s[1] / counts_sum
278
 
279
  src_sigs.append(src_sig_mat)
280
 
281
  src_signatures = cp.stack(cp.array(src_sigs))
282
 
 
 
 
283
  #=======================================================
284
 
285
  print('Searching for matches...Please wait...')
@@ -301,18 +321,49 @@ if filez:
301
 
302
  for target_sig in tqdm(src_signatures):
303
 
304
- if search_matching_type == 'ratios':
 
305
 
306
- ratios = cp.where(target_sig != 0, cp.divide(cp.minimum(signatures_data, target_sig), cp.maximum(signatures_data, target_sig)), 0)
307
- max_comp_lengths = cp.maximum(cp.repeat(cp.sum(target_sig != 0), signatures_data.shape[0]), cp.sum(signatures_data != 0, axis=1))
308
 
309
- results = cp.divide(cp.sum(ratios, axis=1), max_comp_lengths)
 
310
 
311
- elif search_matching_type == 'distances':
 
 
 
 
 
312
 
313
  distances = cp.power(cp.sum(cp.power(cp.abs(signatures_data - target_sig), distances_norm_order), axis=1), 1 / distances_norm_order)
314
 
315
- results = cp.max(distances) - distances
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
316
 
317
  unique_means = cp.unique(results)
318
  sorted_means = cp.sort(unique_means)[::-1]
@@ -331,7 +382,19 @@ if filez:
331
 
332
  tv_idx += 1
333
 
334
- filtered_results = sorted(zip(all_filtered_means, all_filtered_idxs, all_filtered_tvs), key=lambda x: x[0], reverse=True)[:filter_size]
 
 
 
 
 
 
 
 
 
 
 
 
335
 
336
  #=======================================================
337
 
@@ -356,7 +419,7 @@ if filez:
356
  #=======================================================
357
 
358
  dir_str = str(fn1)
359
- copy_path = '/content/Output-MIDI-Dataset/'+dir_str
360
  if not os.path.exists(copy_path):
361
  os.mkdir(copy_path)
362
 
@@ -396,4 +459,47 @@ else:
396
  print('Could not find any MIDI files. Please check Dataset dir...')
397
  print('=' * 70)
398
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
399
  """# Congrats! You did it! :)"""
 
6
  Original file is located at
7
  https://colab.research.google.com/github/asigalov61/Los-Angeles-MIDI-Dataset/blob/main/Extras/Master_MIDI_Dataset_GPU_Search_and_Filter.ipynb
8
 
9
+ # Master MIDI Dataset GPU Search and Filter (ver. 5.0)
10
 
11
  ***
12
 
 
49
  import statistics
50
  import shutil
51
 
52
+ import locale
53
+ locale.getpreferredencoding = lambda: "UTF-8"
54
+
55
  import cupy as cp
56
 
57
  from huggingface_hub import hf_hub_download
58
 
59
+ from google.colab import files
60
+
61
  print('Loading TMIDIX module...')
62
  os.chdir('/content/Los-Angeles-MIDI-Dataset')
63
 
 
147
  random.shuffle(sigs_data)
148
 
149
  signatures_file_names = []
150
+ sigs_matrixes = [ [0]*(len(TMIDIX.ALL_CHORDS)+256) for i in range(len(sigs_data))]
151
 
152
  idx = 0
153
  for s in tqdm(sigs_data):
154
 
155
  signatures_file_names.append(s[0])
156
 
 
 
157
  for ss in s[1]:
158
+ sigs_matrixes[idx][ss[0]] = ss[1]
159
 
160
  idx += 1
161
 
 
177
 
178
  #@markdown NOTE: You can stop the search at any time to render partial results
179
 
180
+ number_of_top_matches_MIDIs_to_collect = 30 #@param {type:"slider", min:5, max:50, step:1}
181
+ search_matching_type = "Ratios" # @param ["Ratios", "Distances", "Correlations"]
182
+ maximum_match_ratio_to_search_for = 1 #@param {type:"slider", min:0, max:1, step:0.001}
183
+ match_results_weight = 2 # @param {type:"slider", min:0.1, max:3, step:0.1}
184
+ match_lengths_weight = 1 # @param {type:"slider", min:0.1, max:3, step:0.1}
185
+ match_counts_weight = 1 # @param {type:"slider", min:0.1, max:3, step:0.1}
186
  distances_norm_order = 3 # @param {type:"slider", min:1, max:10, step:1}
187
+ epsilon = 0.5 # @param {type:"slider", min:0.001, max:1, step:0.001}
188
+ match_drums = False # @param {type:"boolean"}
189
 
190
  print('=' * 70)
191
  print('Master MIDI Dataset GPU Search and Filter')
 
215
 
216
  ###################
217
 
218
+ if not os.path.exists('/content/Output-MIDI-Dataset/'+search_matching_type):
219
+ os.makedirs('/content/Output-MIDI-Dataset/'+search_matching_type)
220
 
221
  ###################
222
 
 
245
  e[1] = int(e[1] / 16)
246
  e[2] = int(e[2] / 16)
247
 
248
+ drums_offset = len(TMIDIX.ALL_CHORDS) + 128
249
+
250
  src_sigs = []
251
 
252
  for i in range(-6, 6):
 
259
  cscore = TMIDIX.chordify_score([1000, escore_copy])
260
 
261
  sig = []
262
+ dsig = []
263
 
264
  for c in cscore:
265
 
266
  pitches = sorted(set([p[4] for p in c if p[3] != 9]))
267
+ drums = sorted(set([p[4] for p in c if p[3] == 9]))
268
 
269
  if pitches:
270
  if len(pitches) > 1:
 
278
 
279
  sig.append(sig_token)
280
 
281
+ if drums:
282
+ dsig.extend(drums)
283
+
284
+ sig_p = dict.fromkeys(sig+dsig, 0)
285
+ for item in sig+dsig:
286
+ sig_p[item] += 1
287
 
288
+ fsig = [list(v) for v in sig_p.items()]
289
 
290
+ src_sig_mat = [0] * (len(TMIDIX.ALL_CHORDS)+256)
291
 
292
  for s in fsig:
293
 
294
+ src_sig_mat[s[0]] = s[1]
295
 
296
  src_sigs.append(src_sig_mat)
297
 
298
  src_signatures = cp.stack(cp.array(src_sigs))
299
 
300
+ if not match_drums:
301
+ src_signatures = cp.where(src_signatures < drums_offset, src_signatures, epsilon)
302
+
303
  #=======================================================
304
 
305
  print('Searching for matches...Please wait...')
 
321
 
322
  for target_sig in tqdm(src_signatures):
323
 
324
+ if not match_drums:
325
+ target_sig = cp.where(target_sig < drums_offset, target_sig, epsilon)
326
 
327
+ comps_lengths = cp.vstack((cp.repeat(cp.sum(target_sig != 0), signatures_data.shape[0]), cp.sum(signatures_data != 0, axis=1)))
328
+ comps_lengths_ratios = cp.divide(cp.min(comps_lengths, axis=0), cp.max(comps_lengths, axis=0))
329
 
330
+ comps_counts_sums = cp.vstack((cp.repeat(cp.sum(target_sig), signatures_data.shape[0]), cp.sum(signatures_data, axis=1)))
331
+ comps_counts_sums_ratios = cp.divide(cp.min(comps_counts_sums, axis=0), cp.max(comps_counts_sums, axis=0))
332
 
333
+ if search_matching_type == 'Ratios':
334
+
335
+ ratios = cp.where(target_sig != 0, cp.divide(cp.minimum(signatures_data, target_sig), cp.maximum(signatures_data, target_sig)), epsilon)
336
+ results = cp.mean(ratios, axis=1)
337
+
338
+ elif search_matching_type == 'Distances':
339
 
340
  distances = cp.power(cp.sum(cp.power(cp.abs(signatures_data - target_sig), distances_norm_order), axis=1), 1 / distances_norm_order)
341
 
342
+ distances_mean = cp.mean(distances)
343
+ distances_std = cp.std(distances)
344
+
345
+ results = 1 - cp.divide((distances - distances_mean), distances_std)
346
+
347
+ elif search_matching_type == 'Correlations':
348
+
349
+ main_array_mean = cp.mean(signatures_data, axis=1, keepdims=True)
350
+ main_array_std = cp.std(signatures_data, axis=1, keepdims=True)
351
+ target_array_mean = cp.mean(target_sig)
352
+ target_array_std = cp.std(target_sig)
353
+
354
+ signatures_data_normalized = cp.where(main_array_std != 0, (signatures_data - main_array_mean) / main_array_std, epsilon)
355
+ target_sig_normalized = cp.where(target_array_std != 0, (target_sig - target_array_mean) / target_array_std, epsilon)
356
+
357
+ correlations = cp.divide(cp.einsum('ij,j->i', signatures_data_normalized, target_sig_normalized), (signatures_data.shape[1] - 1))
358
+ scaled_correlations = cp.divide(correlations, cp.sqrt(cp.sum(correlations**2)))
359
+ exp = cp.exp(scaled_correlations - cp.max(scaled_correlations))
360
+ results = cp.multiply(cp.divide(exp, cp.sum(exp)), 1e5)
361
+
362
+ results_weight = match_results_weight
363
+ comp_lengths_weight = match_lengths_weight
364
+ comp_counts_sums_weight = match_counts_weight
365
+
366
+ results = cp.divide(cp.add(cp.add(results_weight, comp_lengths_weight), comp_counts_sums_weight), cp.add(cp.add(cp.divide(results_weight, cp.where(results !=0, results, epsilon)), cp.divide(comp_lengths_weight, cp.where(comps_lengths_ratios !=0, comps_lengths_ratios, epsilon))), cp.divide(comp_counts_sums_weight, cp.where(comps_counts_sums_ratios !=0, comps_counts_sums_ratios, epsilon))))
367
 
368
  unique_means = cp.unique(results)
369
  sorted_means = cp.sort(unique_means)[::-1]
 
382
 
383
  tv_idx += 1
384
 
385
+ f_results = sorted(zip(all_filtered_means, all_filtered_idxs, all_filtered_tvs), key=lambda x: x[0], reverse=True)
386
+
387
+ triplet_dict = {}
388
+
389
+ for triplet in f_results:
390
+
391
+ if triplet[0] not in triplet_dict:
392
+ triplet_dict[triplet[0]] = triplet
393
+ else:
394
+ if triplet[2] == 0:
395
+ triplet_dict[triplet[0]] = triplet
396
+
397
+ filtered_results = list(triplet_dict.values())[:filter_size]
398
 
399
  #=======================================================
400
 
 
419
  #=======================================================
420
 
421
  dir_str = str(fn1)
422
+ copy_path = '/content/Output-MIDI-Dataset/'+search_matching_type+'/'+dir_str
423
  if not os.path.exists(copy_path):
424
  os.mkdir(copy_path)
425
 
 
459
  print('Could not find any MIDI files. Please check Dataset dir...')
460
  print('=' * 70)
461
 
462
+ """# (DOWNLOAD RESULTS)"""
463
+
464
+ # Commented out IPython magic to ensure Python compatibility.
465
+ #@title Zip and download all search results
466
+
467
+ print('=' * 70)
468
+
469
+ try:
470
+ os.remove('Master_MIDI_Dataset_Search_Results.zip')
471
+ except OSError:
472
+ pass
473
+
474
+ print('Zipping... Please wait...')
475
+ print('=' * 70)
476
+
477
+ # %cd /content/Output-MIDI-Dataset/
478
+ !zip -r Master_MIDI_Dataset_Search_Results.zip *
479
+ # %cd /content/
480
+
481
+ print('=' * 70)
482
+ print('Done!')
483
+ print('=' * 70)
484
+
485
+ print('Downloading final zip file...')
486
+ print('=' * 70)
487
+
488
+ files.download('/content/Output-MIDI-Dataset/Master_MIDI_Dataset_Search_Results.zip')
489
+
490
+ print('Done!')
491
+ print('=' * 70)
492
+
493
+ # @title Delete search results directory and files
494
+
495
+ #@markdown WARNING: This can't be undone so make sure you downloaded the search results first
496
+
497
+ print('=' * 70)
498
+ print('Deleting... Please wait...')
499
+ print('=' * 70)
500
+
501
+ !rm -rf /content/Output-MIDI-Dataset
502
+ print('Done!')
503
+ print('=' * 70)
504
+
505
  """# Congrats! You did it! :)"""
midi_to_colab_audio.py ADDED
The diff for this file is too large to render. See raw diff