asigalov61
commited on
Commit
•
c1fc4e4
1
Parent(s):
40bf57f
Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,6 @@
|
|
|
|
1 |
# https://huggingface.co/spaces/asigalov61/Chords-Progressions-Generator
|
|
|
2 |
|
3 |
import os
|
4 |
import time as reqtime
|
@@ -7,15 +9,19 @@ from pytz import timezone
|
|
7 |
|
8 |
import gradio as gr
|
9 |
|
|
|
|
|
|
|
10 |
import random
|
|
|
11 |
|
12 |
import TMIDIX
|
13 |
|
14 |
-
|
15 |
-
|
16 |
# =================================================================================================
|
17 |
|
18 |
-
def
|
19 |
|
20 |
print('=' * 70)
|
21 |
print('Req start time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT)))
|
@@ -33,105 +39,9 @@ def ClassifyMIDI(input_midi, input_sampling_resolution):
|
|
33 |
print('Input MIDI file name:', fn)
|
34 |
|
35 |
print('=' * 70)
|
36 |
-
print('Loading MIDI file...')
|
37 |
-
|
38 |
-
midi_name = fn
|
39 |
-
|
40 |
-
raw_score = TMIDIX.midi2single_track_ms_score(open(input_midi.name, 'rb').read())
|
41 |
-
|
42 |
-
escore_notes = TMIDIX.advanced_score_processor(raw_score, return_enhanced_score_notes=True)[0]
|
43 |
-
|
44 |
-
#===============================================================================
|
45 |
-
# Augmented enhanced score notes
|
46 |
-
|
47 |
-
escore_notes = TMIDIX.augment_enhanced_score_notes(escore_notes, timings_divider=32)
|
48 |
-
|
49 |
-
escore_notes = [e for e in escore_notes if e[6] < 80 or e[6] == 128]
|
50 |
-
|
51 |
-
#=======================================================
|
52 |
-
# Augmentation
|
53 |
-
|
54 |
-
#=======================================================
|
55 |
-
# FINAL PROCESSING
|
56 |
-
|
57 |
-
melody_chords = []
|
58 |
-
|
59 |
-
#=======================================================
|
60 |
-
# MAIN PROCESSING CYCLE
|
61 |
-
#=======================================================
|
62 |
-
|
63 |
-
pe = escore_notes[0]
|
64 |
-
|
65 |
-
pitches = []
|
66 |
-
|
67 |
-
notes_counter = 0
|
68 |
-
|
69 |
-
for e in escore_notes:
|
70 |
-
|
71 |
-
#=======================================================
|
72 |
-
# Timings...
|
73 |
-
|
74 |
-
delta_time = max(0, min(127, e[1]-pe[1]))
|
75 |
-
|
76 |
-
if delta_time != 0:
|
77 |
-
pitches = []
|
78 |
-
|
79 |
-
# Durations and channels
|
80 |
-
|
81 |
-
dur = max(1, min(127, e[2]))
|
82 |
-
|
83 |
-
# Patches
|
84 |
-
pat = max(0, min(128, e[6]))
|
85 |
-
|
86 |
-
# Pitches
|
87 |
-
|
88 |
-
if pat == 128:
|
89 |
-
ptc = max(1, min(127, e[4]))+128
|
90 |
-
else:
|
91 |
-
ptc = max(1, min(127, e[4]))
|
92 |
-
|
93 |
-
#=======================================================
|
94 |
-
# FINAL NOTE SEQ
|
95 |
-
|
96 |
-
# Writing final note synchronously
|
97 |
-
|
98 |
-
if ptc not in pitches:
|
99 |
-
melody_chords.extend([delta_time, dur+128, ptc+256])
|
100 |
-
pitches.append(ptc)
|
101 |
-
notes_counter += 1
|
102 |
-
|
103 |
-
pe = e
|
104 |
-
|
105 |
-
#==============================================================
|
106 |
-
|
107 |
-
print('Done!')
|
108 |
-
print('=' * 70)
|
109 |
-
|
110 |
-
print('Sampling score...')
|
111 |
-
|
112 |
-
chunk_size = 1020
|
113 |
-
|
114 |
-
score = melody_chords
|
115 |
-
|
116 |
-
input_data = []
|
117 |
-
|
118 |
-
for i in range(0, len(score)-chunk_size, chunk_size // input_sampling_resolution):
|
119 |
-
schunk = score[i:i+chunk_size]
|
120 |
-
|
121 |
-
if len(schunk) == chunk_size:
|
122 |
-
|
123 |
-
td = [937]
|
124 |
-
|
125 |
-
td.extend(schunk)
|
126 |
|
127 |
-
td.extend([938])
|
128 |
-
|
129 |
-
input_data.append(td)
|
130 |
-
|
131 |
print('Done!')
|
132 |
print('=' * 70)
|
133 |
-
|
134 |
-
#==============================================================
|
135 |
|
136 |
classification_summary_string = '=' * 70
|
137 |
classification_summary_string += '\n'
|
@@ -153,79 +63,11 @@ def ClassifyMIDI(input_midi, input_sampling_resolution):
|
|
153 |
classification_summary_string += '=' * 70
|
154 |
classification_summary_string += '\n'
|
155 |
|
156 |
-
print('Loading model...')
|
157 |
-
|
158 |
-
SEQ_LEN = 1026
|
159 |
-
PAD_IDX = 940
|
160 |
-
DEVICE = 'cuda' # 'cuda'
|
161 |
-
|
162 |
-
# instantiate the model
|
163 |
-
|
164 |
-
model = TransformerWrapper(
|
165 |
-
num_tokens = PAD_IDX+1,
|
166 |
-
max_seq_len = SEQ_LEN,
|
167 |
-
attn_layers = Decoder(dim = 1024, depth = 24, heads = 32, attn_flash = True)
|
168 |
-
)
|
169 |
-
|
170 |
-
model = AutoregressiveWrapper(model, ignore_index=PAD_IDX, pad_value=PAD_IDX)
|
171 |
-
|
172 |
-
model = torch.nn.DataParallel(model)
|
173 |
-
|
174 |
-
model.to(DEVICE)
|
175 |
-
|
176 |
-
print('=' * 70)
|
177 |
-
|
178 |
-
print('Loading model checkpoint...')
|
179 |
-
|
180 |
-
model.load_state_dict(
|
181 |
-
torch.load('Ultimate_MIDI_Classifier_Trained_Model_29886_steps_0.556_loss_0.8339_acc.pth',
|
182 |
-
map_location=DEVICE))
|
183 |
-
print('=' * 70)
|
184 |
-
|
185 |
-
if DEVICE == 'cpu':
|
186 |
-
dtype = torch.bfloat16
|
187 |
-
else:
|
188 |
-
dtype = torch.bfloat16
|
189 |
-
|
190 |
-
ctx = torch.amp.autocast(device_type=DEVICE, dtype=dtype)
|
191 |
-
|
192 |
-
print('Done!')
|
193 |
-
print('=' * 70)
|
194 |
-
|
195 |
#==================================================================
|
196 |
|
197 |
print('=' * 70)
|
198 |
print('Ultimate MIDI Classifier')
|
199 |
print('=' * 70)
|
200 |
-
print('Classifying...')
|
201 |
-
|
202 |
-
torch.cuda.empty_cache()
|
203 |
-
|
204 |
-
model.eval()
|
205 |
-
|
206 |
-
artist_results = []
|
207 |
-
song_results = []
|
208 |
-
|
209 |
-
results = []
|
210 |
-
|
211 |
-
for input in input_data:
|
212 |
-
|
213 |
-
x = torch.tensor(input[:1022], dtype=torch.long, device='cuda')
|
214 |
-
|
215 |
-
with ctx:
|
216 |
-
out = model.module.generate(x,
|
217 |
-
2,
|
218 |
-
filter_logits_fn=top_k,
|
219 |
-
filter_kwargs={'k': 1},
|
220 |
-
temperature=0.9,
|
221 |
-
return_prime=False,
|
222 |
-
verbose=False)
|
223 |
-
|
224 |
-
result = tuple(out[0].tolist())
|
225 |
-
|
226 |
-
results.append(result)
|
227 |
-
|
228 |
-
final_result = mode(results)
|
229 |
|
230 |
print('=' * 70)
|
231 |
print('Done!')
|
@@ -302,93 +144,39 @@ if __name__ == "__main__":
|
|
302 |
print('App start time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT)))
|
303 |
print('=' * 70)
|
304 |
|
305 |
-
#===============================================================================
|
306 |
-
# Helper functions
|
307 |
-
#===============================================================================
|
308 |
-
|
309 |
-
def str_strip_song(string):
|
310 |
-
if string is not None:
|
311 |
-
string = string.replace('-', ' ').replace('_', ' ').replace('=', ' ')
|
312 |
-
str1 = re.compile('[^a-zA-Z ]').sub('', string)
|
313 |
-
return re.sub(' +', ' ', str1).strip().title()
|
314 |
-
else:
|
315 |
-
return ''
|
316 |
-
|
317 |
-
def str_strip_artist(string):
|
318 |
-
if string is not None:
|
319 |
-
string = string.replace('-', ' ').replace('_', ' ').replace('=', ' ')
|
320 |
-
str1 = re.compile('[^0-9a-zA-Z ]').sub('', string)
|
321 |
-
return re.sub(' +', ' ', str1).strip().title()
|
322 |
-
else:
|
323 |
-
return ''
|
324 |
-
|
325 |
-
def song_artist_to_song_artist_tokens(file_name):
|
326 |
-
idx = classifier_labels.index(file_name)
|
327 |
-
|
328 |
-
tok1 = idx // 424
|
329 |
-
tok2 = idx % 424
|
330 |
-
|
331 |
-
return [tok1, tok2]
|
332 |
-
|
333 |
-
def song_artist_tokens_to_song_artist(file_name_tokens):
|
334 |
-
|
335 |
-
tok1 = file_name_tokens[0]
|
336 |
-
tok2 = file_name_tokens[1]
|
337 |
-
|
338 |
-
idx = (tok1 * 424) + tok2
|
339 |
-
|
340 |
-
return classifier_labels[idx]
|
341 |
-
|
342 |
-
#===============================================================================
|
343 |
-
|
344 |
print('=' * 70)
|
345 |
print('Loading Ultimate MIDI Classifier labels...')
|
346 |
print('=' * 70)
|
347 |
-
|
348 |
-
|
349 |
-
genre_labels = TMIDIX.Tegridy_Any_Pickle_File_Reader('Ultimate_MIDI_Classifier_Music_Genre_Labels')
|
350 |
-
genre_labels_fnames = [f[0] for f in genre_labels]
|
351 |
print('=' * 70)
|
352 |
print('Done!')
|
353 |
print('=' * 70)
|
354 |
|
355 |
app = gr.Blocks()
|
356 |
with app:
|
357 |
-
gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>
|
358 |
-
gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>
|
359 |
gr.Markdown(
|
360 |
-
"![Visitors](https://api.visitorbadge.io/api/visitors?path=asigalov61.
|
361 |
-
"This is a demo for
|
362 |
-
"Check out [
|
363 |
"[Open In Colab]"
|
364 |
-
"(https://colab.research.google.com/github/asigalov61/
|
365 |
-
" for all options, faster execution and endless
|
366 |
)
|
367 |
|
368 |
-
gr.Markdown("##
|
369 |
-
|
370 |
-
|
371 |
-
input_midi = gr.File(label="Input MIDI", file_types=[".midi", ".mid", ".kar"])
|
372 |
input_sampling_resolution = gr.Slider(1, 5, value=2, step=1, label="Classification sampling resolution")
|
373 |
|
374 |
run_btn = gr.Button("classify", variant="primary")
|
375 |
|
376 |
-
gr.Markdown("##
|
377 |
|
378 |
output_midi_cls_summary = gr.Textbox(label="MIDI classification results")
|
379 |
|
380 |
run_event = run_btn.click(ClassifyMIDI, [input_midi, input_sampling_resolution],
|
381 |
[output_midi_cls_summary])
|
382 |
-
|
383 |
-
[["Honesty.kar", 2],
|
384 |
-
["House Of The Rising Sun.mid", 2],
|
385 |
-
["Nothing Else Matters.kar", 2],
|
386 |
-
["Sharing The Night Together.kar", 2]
|
387 |
-
],
|
388 |
-
[input_midi, input_sampling_resolution],
|
389 |
-
[output_midi_cls_summary],
|
390 |
-
ClassifyMIDI,
|
391 |
-
cache_examples=True,
|
392 |
-
)
|
393 |
-
|
394 |
app.queue().launch()
|
|
|
1 |
+
# =================================================================================================
|
2 |
# https://huggingface.co/spaces/asigalov61/Chords-Progressions-Generator
|
3 |
+
# =================================================================================================
|
4 |
|
5 |
import os
|
6 |
import time as reqtime
|
|
|
9 |
|
10 |
import gradio as gr
|
11 |
|
12 |
+
import numpy as np
|
13 |
+
|
14 |
+
import os
|
15 |
import random
|
16 |
+
from collections import Counter
|
17 |
|
18 |
import TMIDIX
|
19 |
|
20 |
+
from midi_to_colab_audio import midi_to_colab_audio
|
21 |
+
|
22 |
# =================================================================================================
|
23 |
|
24 |
+
def Generate_Chords_Progression(input_midi, input_sampling_resolution):
|
25 |
|
26 |
print('=' * 70)
|
27 |
print('Req start time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT)))
|
|
|
39 |
print('Input MIDI file name:', fn)
|
40 |
|
41 |
print('=' * 70)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
|
|
|
|
|
|
|
|
|
43 |
print('Done!')
|
44 |
print('=' * 70)
|
|
|
|
|
45 |
|
46 |
classification_summary_string = '=' * 70
|
47 |
classification_summary_string += '\n'
|
|
|
63 |
classification_summary_string += '=' * 70
|
64 |
classification_summary_string += '\n'
|
65 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
#==================================================================
|
67 |
|
68 |
print('=' * 70)
|
69 |
print('Ultimate MIDI Classifier')
|
70 |
print('=' * 70)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
|
72 |
print('=' * 70)
|
73 |
print('Done!')
|
|
|
144 |
print('App start time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT)))
|
145 |
print('=' * 70)
|
146 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
147 |
print('=' * 70)
|
148 |
print('Loading Ultimate MIDI Classifier labels...')
|
149 |
print('=' * 70)
|
150 |
+
good_chords_chunks = TMIDIX.Tegridy_Any_Pickle_File_Reader('pitches_chords_progressions_5_3_15')
|
151 |
+
|
|
|
|
|
152 |
print('=' * 70)
|
153 |
print('Done!')
|
154 |
print('=' * 70)
|
155 |
|
156 |
app = gr.Blocks()
|
157 |
with app:
|
158 |
+
gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>Chords Progressions Generator</h1>")
|
159 |
+
gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>Generate unique chords progressions</h1>")
|
160 |
gr.Markdown(
|
161 |
+
"![Visitors](https://api.visitorbadge.io/api/visitors?path=asigalov61.Chords-Progressions-Generator&style=flat)\n\n"
|
162 |
+
"This is a demo for Tegridy MIDI Dataset\n\n"
|
163 |
+
"Check out [Tegridy MIDI Dataset](https://github.com/asigalov61/Tegridy-MIDI-Dataset) on GitHub!\n\n"
|
164 |
"[Open In Colab]"
|
165 |
+
"(https://colab.research.google.com/github/asigalov61/Tegridy-MIDI-Dataset/blob/master/Chords-Progressions/Pitches_Chords_Progressions_Generator.ipynb)"
|
166 |
+
" for all options, faster execution and endless generation"
|
167 |
)
|
168 |
|
169 |
+
gr.Markdown("## Select generation options")
|
170 |
+
|
|
|
|
|
171 |
input_sampling_resolution = gr.Slider(1, 5, value=2, step=1, label="Classification sampling resolution")
|
172 |
|
173 |
run_btn = gr.Button("classify", variant="primary")
|
174 |
|
175 |
+
gr.Markdown("## Generation results")
|
176 |
|
177 |
output_midi_cls_summary = gr.Textbox(label="MIDI classification results")
|
178 |
|
179 |
run_event = run_btn.click(ClassifyMIDI, [input_midi, input_sampling_resolution],
|
180 |
[output_midi_cls_summary])
|
181 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
182 |
app.queue().launch()
|