Spaces:
Paused
Paused
speed up rendering audio
Browse files- app.py +41 -9
- midi_synthesizer.py +6 -2
app.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
import spaces
|
2 |
import random
|
3 |
import argparse
|
@@ -240,7 +242,8 @@ def run(model_name, tab, mid_seq, continuation_state, continuation_select, instr
|
|
240 |
|
241 |
def finish_run(model_name, mid_seq):
|
242 |
if mid_seq is None:
|
243 |
-
|
|
|
244 |
tokenizer = models[model_name].tokenizer
|
245 |
outputs = []
|
246 |
end_msgs = [create_msg("progress", [0, 0])]
|
@@ -249,16 +252,36 @@ def finish_run(model_name, mid_seq):
|
|
249 |
for i in range(OUTPUT_BATCH_SIZE):
|
250 |
events = [tokenizer.tokens2event(tokens) for tokens in mid_seq[i]]
|
251 |
mid = tokenizer.detokenize(mid_seq[i])
|
252 |
-
audio = synthesizer.synthesis(MIDI.score2opus(mid))
|
253 |
with open(f"outputs/output{i + 1}.mid", 'wb') as f:
|
254 |
f.write(MIDI.score2midi(mid))
|
255 |
-
outputs
|
256 |
end_msgs += [create_msg("visualizer_clear", [i, tokenizer.version]),
|
257 |
create_msg("visualizer_append", [i, events]),
|
258 |
create_msg("visualizer_end", i)]
|
259 |
return *outputs, send_msgs(end_msgs)
|
260 |
|
261 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
262 |
def undo_continuation(model_name, mid_seq, continuation_state):
|
263 |
if mid_seq is None or len(continuation_state) < 2:
|
264 |
return mid_seq, continuation_state, send_msgs([])
|
@@ -324,6 +347,7 @@ if __name__ == "__main__":
|
|
324 |
parser.add_argument("--max-gen", type=int, default=1024, help="max")
|
325 |
opt = parser.parse_args()
|
326 |
soundfont_path = hf_hub_download_retry(repo_id="skytnt/midi-model", filename="soundfont.sf2")
|
|
|
327 |
synthesizer = MidiSynthesizer(soundfont_path)
|
328 |
models_info = {
|
329 |
"generic pretrain model (tv2o-medium) by skytnt": ["skytnt/midi-model-tv2o-medium", "", "tv2o-medium"],
|
@@ -442,20 +466,23 @@ if __name__ == "__main__":
|
|
442 |
input_top_p = gr.Slider(label="top p", minimum=0.1, maximum=1, step=0.01, value=0.98)
|
443 |
input_top_k = gr.Slider(label="top k", minimum=1, maximum=128, step=1, value=30)
|
444 |
input_allow_cc = gr.Checkbox(label="allow midi cc event", value=True)
|
|
|
445 |
example3 = gr.Examples([[1, 0.95, 128], [1, 0.98, 20], [1, 0.98, 12]],
|
446 |
[input_temp, input_top_p, input_top_k])
|
447 |
run_btn = gr.Button("generate", variant="primary")
|
448 |
# stop_btn = gr.Button("stop and output")
|
449 |
output_midi_seq = gr.State()
|
450 |
output_continuation_state = gr.State([0])
|
451 |
-
|
|
|
452 |
with gr.Tabs(elem_id="output_tabs"):
|
453 |
for i in range(OUTPUT_BATCH_SIZE):
|
454 |
with gr.TabItem(f"output {i + 1}") as tab1:
|
455 |
output_midi_visualizer = gr.HTML(elem_id=f"midi_visualizer_container_{i}")
|
456 |
output_audio = gr.Audio(label="output audio", format="mp3", elem_id=f"midi_audio_{i}")
|
457 |
output_midi = gr.File(label="output midi", file_types=[".mid"])
|
458 |
-
|
|
|
459 |
run_event = run_btn.click(run, [input_model, tab_select, output_midi_seq, output_continuation_state,
|
460 |
input_continuation_select, input_instruments, input_drum_kit, input_bpm,
|
461 |
input_time_sig, input_key_sig, input_midi, input_midi_events,
|
@@ -465,12 +492,17 @@ if __name__ == "__main__":
|
|
465 |
input_top_k, input_allow_cc],
|
466 |
[output_midi_seq, output_continuation_state, input_seed, js_msg],
|
467 |
concurrency_limit=10, queue=True)
|
468 |
-
run_event.then(fn=finish_run,
|
469 |
-
|
470 |
-
|
471 |
-
|
|
|
|
|
|
|
|
|
472 |
# stop_btn.click(None, [], [], cancels=run_event,
|
473 |
# queue=False)
|
474 |
undo_btn.click(undo_continuation, [input_model, output_midi_seq, output_continuation_state],
|
475 |
[output_midi_seq, output_continuation_state, js_msg], queue=False)
|
476 |
app.queue().launch(server_port=opt.port, share=opt.share, inbrowser=True)
|
|
|
|
1 |
+
from concurrent.futures import ThreadPoolExecutor
|
2 |
+
|
3 |
import spaces
|
4 |
import random
|
5 |
import argparse
|
|
|
242 |
|
243 |
def finish_run(model_name, mid_seq):
|
244 |
if mid_seq is None:
|
245 |
+
outputs = [None] * OUTPUT_BATCH_SIZE
|
246 |
+
return *outputs, []
|
247 |
tokenizer = models[model_name].tokenizer
|
248 |
outputs = []
|
249 |
end_msgs = [create_msg("progress", [0, 0])]
|
|
|
252 |
for i in range(OUTPUT_BATCH_SIZE):
|
253 |
events = [tokenizer.tokens2event(tokens) for tokens in mid_seq[i]]
|
254 |
mid = tokenizer.detokenize(mid_seq[i])
|
|
|
255 |
with open(f"outputs/output{i + 1}.mid", 'wb') as f:
|
256 |
f.write(MIDI.score2midi(mid))
|
257 |
+
outputs.append(f"outputs/output{i + 1}.mid")
|
258 |
end_msgs += [create_msg("visualizer_clear", [i, tokenizer.version]),
|
259 |
create_msg("visualizer_append", [i, events]),
|
260 |
create_msg("visualizer_end", i)]
|
261 |
return *outputs, send_msgs(end_msgs)
|
262 |
|
263 |
|
264 |
+
def synthesis_task(mid):
|
265 |
+
return synthesizer.synthesis(MIDI.score2opus(mid))
|
266 |
+
|
267 |
+
def render_audio(model_name, mid_seq, should_render_audio):
|
268 |
+
if (not should_render_audio) or mid_seq is None:
|
269 |
+
outputs = [None] * OUTPUT_BATCH_SIZE
|
270 |
+
return tuple(outputs)
|
271 |
+
tokenizer = models[model_name].tokenizer
|
272 |
+
outputs = []
|
273 |
+
if not os.path.exists("outputs"):
|
274 |
+
os.mkdir("outputs")
|
275 |
+
audio_futures = []
|
276 |
+
for i in range(OUTPUT_BATCH_SIZE):
|
277 |
+
mid = tokenizer.detokenize(mid_seq[i])
|
278 |
+
audio_future = thread_pool.submit(synthesis_task, mid)
|
279 |
+
audio_futures.append(audio_future)
|
280 |
+
for future in audio_futures:
|
281 |
+
outputs.append((44100, future.result()))
|
282 |
+
return tuple(outputs)
|
283 |
+
|
284 |
+
|
285 |
def undo_continuation(model_name, mid_seq, continuation_state):
|
286 |
if mid_seq is None or len(continuation_state) < 2:
|
287 |
return mid_seq, continuation_state, send_msgs([])
|
|
|
347 |
parser.add_argument("--max-gen", type=int, default=1024, help="max")
|
348 |
opt = parser.parse_args()
|
349 |
soundfont_path = hf_hub_download_retry(repo_id="skytnt/midi-model", filename="soundfont.sf2")
|
350 |
+
thread_pool = ThreadPoolExecutor(max_workers=OUTPUT_BATCH_SIZE)
|
351 |
synthesizer = MidiSynthesizer(soundfont_path)
|
352 |
models_info = {
|
353 |
"generic pretrain model (tv2o-medium) by skytnt": ["skytnt/midi-model-tv2o-medium", "", "tv2o-medium"],
|
|
|
466 |
input_top_p = gr.Slider(label="top p", minimum=0.1, maximum=1, step=0.01, value=0.98)
|
467 |
input_top_k = gr.Slider(label="top k", minimum=1, maximum=128, step=1, value=30)
|
468 |
input_allow_cc = gr.Checkbox(label="allow midi cc event", value=True)
|
469 |
+
input_render_audio = gr.Checkbox(label="render audio after generation", value=True)
|
470 |
example3 = gr.Examples([[1, 0.95, 128], [1, 0.98, 20], [1, 0.98, 12]],
|
471 |
[input_temp, input_top_p, input_top_k])
|
472 |
run_btn = gr.Button("generate", variant="primary")
|
473 |
# stop_btn = gr.Button("stop and output")
|
474 |
output_midi_seq = gr.State()
|
475 |
output_continuation_state = gr.State([0])
|
476 |
+
midi_outputs = []
|
477 |
+
audio_outputs = []
|
478 |
with gr.Tabs(elem_id="output_tabs"):
|
479 |
for i in range(OUTPUT_BATCH_SIZE):
|
480 |
with gr.TabItem(f"output {i + 1}") as tab1:
|
481 |
output_midi_visualizer = gr.HTML(elem_id=f"midi_visualizer_container_{i}")
|
482 |
output_audio = gr.Audio(label="output audio", format="mp3", elem_id=f"midi_audio_{i}")
|
483 |
output_midi = gr.File(label="output midi", file_types=[".mid"])
|
484 |
+
midi_outputs.append(output_midi)
|
485 |
+
audio_outputs.append(output_audio)
|
486 |
run_event = run_btn.click(run, [input_model, tab_select, output_midi_seq, output_continuation_state,
|
487 |
input_continuation_select, input_instruments, input_drum_kit, input_bpm,
|
488 |
input_time_sig, input_key_sig, input_midi, input_midi_events,
|
|
|
492 |
input_top_k, input_allow_cc],
|
493 |
[output_midi_seq, output_continuation_state, input_seed, js_msg],
|
494 |
concurrency_limit=10, queue=True)
|
495 |
+
finish_run_event = run_event.then(fn=finish_run,
|
496 |
+
inputs=[input_model, output_midi_seq],
|
497 |
+
outputs=midi_outputs + [js_msg],
|
498 |
+
queue=False)
|
499 |
+
finish_run_event.then(fn=render_audio,
|
500 |
+
inputs=[input_model, output_midi_seq, input_render_audio],
|
501 |
+
outputs=audio_outputs,
|
502 |
+
queue=False)
|
503 |
# stop_btn.click(None, [], [], cancels=run_event,
|
504 |
# queue=False)
|
505 |
undo_btn.click(undo_continuation, [input_model, output_midi_seq, output_continuation_state],
|
506 |
[output_midi_seq, output_continuation_state, js_msg], queue=False)
|
507 |
app.queue().launch(server_port=opt.port, share=opt.share, inbrowser=True)
|
508 |
+
thread_pool.shutdown()
|
midi_synthesizer.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
import fluidsynth
|
2 |
import numpy as np
|
3 |
|
@@ -9,14 +11,16 @@ class MidiSynthesizer:
|
|
9 |
fl = fluidsynth.Synth(samplerate=float(sample_rate))
|
10 |
sfid = fl.sfload(soundfont_path)
|
11 |
self.devices = [[fl, sfid, False]]
|
|
|
12 |
|
13 |
def get_fluidsynth(self):
|
14 |
for device in self.devices:
|
15 |
if not device[2]:
|
16 |
device[2] = True
|
17 |
return device
|
18 |
-
|
19 |
-
|
|
|
20 |
device = [fl, sfid, True]
|
21 |
self.devices.append(device)
|
22 |
return device
|
|
|
1 |
+
from threading import Lock
|
2 |
+
|
3 |
import fluidsynth
|
4 |
import numpy as np
|
5 |
|
|
|
11 |
fl = fluidsynth.Synth(samplerate=float(sample_rate))
|
12 |
sfid = fl.sfload(soundfont_path)
|
13 |
self.devices = [[fl, sfid, False]]
|
14 |
+
self.file_lock = Lock()
|
15 |
|
16 |
def get_fluidsynth(self):
|
17 |
for device in self.devices:
|
18 |
if not device[2]:
|
19 |
device[2] = True
|
20 |
return device
|
21 |
+
with self.file_lock:
|
22 |
+
fl = fluidsynth.Synth(samplerate=float(self.sample_rate))
|
23 |
+
sfid = fl.sfload(self.soundfont_path)
|
24 |
device = [fl, sfid, True]
|
25 |
self.devices.append(device)
|
26 |
return device
|