Spaces:
Running
on
Zero
Running
on
Zero
batch
Browse files- .gitignore +1 -0
- app.py +159 -116
- javascript/app.js +59 -35
- midi_model.py +46 -25
.gitignore
CHANGED
@@ -151,3 +151,4 @@ cython_debug/
|
|
151 |
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
|
152 |
.idea/
|
153 |
output.mid
|
|
|
|
151 |
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
|
152 |
.idea/
|
153 |
output.mid
|
154 |
+
/outputs/
|
app.py
CHANGED
@@ -18,11 +18,12 @@ from midi_model import MIDIModel, MIDIModelConfig
|
|
18 |
from midi_synthesizer import MidiSynthesizer
|
19 |
|
20 |
MAX_SEED = np.iinfo(np.int32).max
|
|
|
21 |
in_space = os.getenv("SYSTEM") == "spaces"
|
22 |
|
23 |
|
24 |
@torch.inference_mode()
|
25 |
-
def generate(model: MIDIModel, prompt=None, max_len=512, temp=1.0, top_p=0.98, top_k=20,
|
26 |
disable_patch_change=False, disable_control_change=False, disable_channels=None, generator=None):
|
27 |
tokenizer = model.tokenizer
|
28 |
if disable_channels is not None:
|
@@ -33,49 +34,69 @@ def generate(model: MIDIModel, prompt=None, max_len=512, temp=1.0, top_p=0.98, t
|
|
33 |
if prompt is None:
|
34 |
input_tensor = torch.full((1, max_token_seq), tokenizer.pad_id, dtype=torch.long, device=model.device)
|
35 |
input_tensor[0, 0] = tokenizer.bos_id # bos
|
|
|
|
|
36 |
else:
|
37 |
-
prompt
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
if prompt.shape[-1] < max_token_seq:
|
39 |
-
prompt = np.pad(prompt, ((0, 0), (0, max_token_seq - prompt.shape[-1])),
|
40 |
mode="constant", constant_values=tokenizer.pad_id)
|
41 |
input_tensor = torch.from_numpy(prompt).to(dtype=torch.long, device=model.device)
|
42 |
-
input_tensor = input_tensor.unsqueeze(0)
|
43 |
cur_len = input_tensor.shape[1]
|
44 |
bar = tqdm.tqdm(desc="generating", total=max_len - cur_len)
|
45 |
with bar:
|
46 |
while cur_len < max_len:
|
47 |
-
end = False
|
48 |
-
hidden = model.forward(input_tensor)[
|
49 |
next_token_seq = None
|
50 |
-
|
51 |
for i in range(max_token_seq):
|
52 |
-
mask = torch.zeros(tokenizer.vocab_size, dtype=torch.int64, device=model.device)
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
if
|
58 |
-
mask_ids
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
logits = model.forward_token(hidden, next_token_seq)[:, -1:]
|
67 |
scores = torch.softmax(logits / temp, dim=-1) * mask
|
68 |
-
|
69 |
if i == 0:
|
70 |
-
next_token_seq =
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
|
|
|
|
|
|
76 |
else:
|
77 |
-
next_token_seq = torch.cat([next_token_seq,
|
78 |
-
if len(tokenizer.events[
|
79 |
break
|
80 |
if next_token_seq.shape[1] < max_token_seq:
|
81 |
next_token_seq = F.pad(next_token_seq, (0, max_token_seq - next_token_seq.shape[1]),
|
@@ -84,8 +105,8 @@ def generate(model: MIDIModel, prompt=None, max_len=512, temp=1.0, top_p=0.98, t
|
|
84 |
input_tensor = torch.cat([input_tensor, next_token_seq], dim=1)
|
85 |
cur_len += 1
|
86 |
bar.update(1)
|
87 |
-
yield next_token_seq.
|
88 |
-
if end:
|
89 |
break
|
90 |
|
91 |
|
@@ -96,8 +117,9 @@ def create_msg(name, data):
|
|
96 |
def send_msgs(msgs):
|
97 |
return json.dumps(msgs)
|
98 |
|
99 |
-
|
100 |
-
|
|
|
101 |
remove_empty_channels, seed, seed_rand, gen_events, temp, top_p, top_k, allow_cc):
|
102 |
if "large" in model_name:
|
103 |
return gen_events // 10 + 15
|
@@ -106,9 +128,9 @@ def get_duration(model_name, tab, mid_seq, continuation_state, instruments, drum
|
|
106 |
|
107 |
|
108 |
@spaces.GPU(duration=get_duration)
|
109 |
-
def run(model_name, tab, mid_seq, continuation_state, instruments, drum_kit, bpm, time_sig,
|
110 |
-
reduce_cc_st, remap_track_channel, add_default_instr, remove_empty_channels,
|
111 |
-
gen_events, temp, top_p, top_k, allow_cc):
|
112 |
model = models[model_name]
|
113 |
model.to(device=opt.device)
|
114 |
tokenizer = model.tokenizer
|
@@ -156,8 +178,8 @@ def run(model_name, tab, mid_seq, continuation_state, instruments, drum_kit, bpm
|
|
156 |
patches[9] = drum_kits2number[drum_kit]
|
157 |
for i, (c, p) in enumerate(patches.items()):
|
158 |
mid.append(tokenizer.event2tokens(["patch_change", 0, 0, i + 1, c, p]))
|
159 |
-
|
160 |
-
|
161 |
if len(instruments) > 0:
|
162 |
disable_patch_change = True
|
163 |
disable_channels = [i for i in range(16) if i not in patches]
|
@@ -167,84 +189,91 @@ def run(model_name, tab, mid_seq, continuation_state, instruments, drum_kit, bpm
|
|
167 |
remap_track_channel=remap_track_channel,
|
168 |
add_default_instr=add_default_instr,
|
169 |
remove_empty_channels=remove_empty_channels)
|
170 |
-
mid = np.asarray(mid, dtype=np.int64)
|
171 |
mid = mid[:int(midi_events)]
|
172 |
-
|
173 |
-
|
174 |
-
mid_seq.append(token_seq.tolist())
|
175 |
elif tab == 2 and mid_seq is not None:
|
176 |
-
continuation_state.append(len(mid_seq))
|
177 |
mid = np.asarray(mid_seq, dtype=np.int64)
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
else:
|
179 |
continuation_state = [0]
|
180 |
-
|
181 |
-
mid =
|
|
|
182 |
|
183 |
if mid is not None:
|
184 |
-
max_len +=
|
185 |
|
186 |
-
events = [tokenizer.tokens2event(tokens) for tokens in mid_seq]
|
187 |
init_msgs = [create_msg("progress", [0, gen_events])]
|
188 |
-
if tab
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
208 |
|
209 |
-
events = [tokenizer.tokens2event(tokens) for tokens in mid_seq]
|
210 |
-
mid = tokenizer.detokenize(mid_seq)
|
211 |
-
audio = synthesizer.synthesis(MIDI.score2opus(mid))
|
212 |
-
with open(f"output.mid", 'wb') as f:
|
213 |
-
f.write(MIDI.score2midi(mid))
|
214 |
-
end_msgs = [create_msg("visualizer_clear", tokenizer.version),
|
215 |
-
create_msg("visualizer_append", events),
|
216 |
-
create_msg("visualizer_end", None),
|
217 |
-
create_msg("progress", [0, 0])]
|
218 |
-
yield mid_seq, continuation_state, "output.mid", (44100, audio), seed, send_msgs(end_msgs)
|
219 |
|
220 |
-
|
221 |
-
def cancel_run(model_name, mid_seq):
|
222 |
if mid_seq is None:
|
223 |
return None, None, []
|
224 |
tokenizer = models[model_name].tokenizer
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
-
|
231 |
-
|
232 |
-
|
233 |
-
|
234 |
-
|
|
|
|
|
|
|
|
|
|
|
235 |
|
236 |
|
237 |
def undo_continuation(model_name, mid_seq, continuation_state):
|
238 |
if mid_seq is None or len(continuation_state) < 2:
|
239 |
return mid_seq, continuation_state, send_msgs([])
|
240 |
-
mid_seq = mid_seq[:continuation_state[-1]]
|
241 |
-
continuation_state = continuation_state[:-1]
|
242 |
tokenizer = models[model_name].tokenizer
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
248 |
return mid_seq, continuation_state, send_msgs(end_msgs)
|
249 |
|
250 |
|
@@ -296,13 +325,14 @@ if __name__ == "__main__":
|
|
296 |
opt = parser.parse_args()
|
297 |
soundfont_path = hf_hub_download_retry(repo_id="skytnt/midi-model", filename="soundfont.sf2")
|
298 |
synthesizer = MidiSynthesizer(soundfont_path)
|
299 |
-
models_info = {
|
300 |
-
|
301 |
-
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
|
|
306 |
models = {}
|
307 |
if opt.device == "cuda":
|
308 |
torch.backends.cudnn.deterministic = True
|
@@ -391,7 +421,12 @@ if __name__ == "__main__":
|
|
391 |
example2 = gr.Examples([[file, 128] for file in glob.glob("example/*.mid")],
|
392 |
[input_midi, input_midi_events])
|
393 |
with gr.TabItem("last output prompt") as tab3:
|
394 |
-
gr.Markdown("Continue generating on the last output.
|
|
|
|
|
|
|
|
|
|
|
395 |
undo_btn = gr.Button("undo the last continuation")
|
396 |
|
397 |
tab1.select(lambda: 0, None, tab_select, queue=False)
|
@@ -413,21 +448,29 @@ if __name__ == "__main__":
|
|
413 |
stop_btn = gr.Button("stop and output")
|
414 |
output_midi_seq = gr.State()
|
415 |
output_continuation_state = gr.State([0])
|
416 |
-
|
417 |
-
|
418 |
-
|
|
|
|
|
|
|
|
|
|
|
419 |
run_event = run_btn.click(run, [input_model, tab_select, output_midi_seq, output_continuation_state,
|
420 |
-
input_instruments, input_drum_kit, input_bpm,
|
421 |
-
|
|
|
422 |
input_add_default_instr, input_remove_empty_channels,
|
423 |
input_seed, input_seed_rand, input_gen_events, input_temp, input_top_p,
|
424 |
input_top_k, input_allow_cc],
|
425 |
-
[output_midi_seq, output_continuation_state,
|
426 |
-
|
427 |
-
|
428 |
-
|
429 |
-
|
430 |
-
|
|
|
|
|
431 |
undo_btn.click(undo_continuation, [input_model, output_midi_seq, output_continuation_state],
|
432 |
[output_midi_seq, output_continuation_state, js_msg], queue=False)
|
433 |
app.queue().launch(server_port=opt.port, share=opt.share, inbrowser=True)
|
|
|
18 |
from midi_synthesizer import MidiSynthesizer
|
19 |
|
20 |
MAX_SEED = np.iinfo(np.int32).max
|
21 |
+
OUTPUT_BATCH_SIZE = 4
|
22 |
in_space = os.getenv("SYSTEM") == "spaces"
|
23 |
|
24 |
|
25 |
@torch.inference_mode()
|
26 |
+
def generate(model: MIDIModel, prompt=None, batch_size=1, max_len=512, temp=1.0, top_p=0.98, top_k=20,
|
27 |
disable_patch_change=False, disable_control_change=False, disable_channels=None, generator=None):
|
28 |
tokenizer = model.tokenizer
|
29 |
if disable_channels is not None:
|
|
|
34 |
if prompt is None:
|
35 |
input_tensor = torch.full((1, max_token_seq), tokenizer.pad_id, dtype=torch.long, device=model.device)
|
36 |
input_tensor[0, 0] = tokenizer.bos_id # bos
|
37 |
+
input_tensor = input_tensor.unsqueeze(0)
|
38 |
+
input_tensor = torch.cat([input_tensor] * batch_size, dim=0)
|
39 |
else:
|
40 |
+
if len(prompt.shape) == 2:
|
41 |
+
prompt = prompt[None, :]
|
42 |
+
prompt = np.repeat(prompt, repeats=batch_size, axis=0)
|
43 |
+
elif prompt.shape[0] == 1:
|
44 |
+
prompt = np.repeat(prompt, repeats=batch_size, axis=0)
|
45 |
+
elif len(prompt.shape) != 3 or prompt.shape[0] != batch_size:
|
46 |
+
raise ValueError(f"invalid shape for prompt, {prompt.shape}")
|
47 |
+
prompt = prompt[..., :max_token_seq]
|
48 |
if prompt.shape[-1] < max_token_seq:
|
49 |
+
prompt = np.pad(prompt, ((0, 0), (0, 0), (0, max_token_seq - prompt.shape[-1])),
|
50 |
mode="constant", constant_values=tokenizer.pad_id)
|
51 |
input_tensor = torch.from_numpy(prompt).to(dtype=torch.long, device=model.device)
|
|
|
52 |
cur_len = input_tensor.shape[1]
|
53 |
bar = tqdm.tqdm(desc="generating", total=max_len - cur_len)
|
54 |
with bar:
|
55 |
while cur_len < max_len:
|
56 |
+
end = [False] * batch_size
|
57 |
+
hidden = model.forward(input_tensor)[:, -1]
|
58 |
next_token_seq = None
|
59 |
+
event_names = [""] * batch_size
|
60 |
for i in range(max_token_seq):
|
61 |
+
mask = torch.zeros((batch_size, tokenizer.vocab_size), dtype=torch.int64, device=model.device)
|
62 |
+
for b in range(batch_size):
|
63 |
+
if end[b]:
|
64 |
+
mask[b, tokenizer.pad_id] = 1
|
65 |
+
continue
|
66 |
+
if i == 0:
|
67 |
+
mask_ids = list(tokenizer.event_ids.values()) + [tokenizer.eos_id]
|
68 |
+
if disable_patch_change:
|
69 |
+
mask_ids.remove(tokenizer.event_ids["patch_change"])
|
70 |
+
if disable_control_change:
|
71 |
+
mask_ids.remove(tokenizer.event_ids["control_change"])
|
72 |
+
mask[b, mask_ids] = 1
|
73 |
+
else:
|
74 |
+
param_names = tokenizer.events[event_names[b]]
|
75 |
+
if i > len(param_names):
|
76 |
+
mask[b, tokenizer.pad_id] = 1
|
77 |
+
continue
|
78 |
+
param_name = param_names[i - 1]
|
79 |
+
mask_ids = tokenizer.parameter_ids[param_name]
|
80 |
+
if param_name == "channel":
|
81 |
+
mask_ids = [i for i in mask_ids if i not in disable_channels]
|
82 |
+
mask[b, mask_ids] = 1
|
83 |
+
mask = mask.unsqueeze(1)
|
84 |
logits = model.forward_token(hidden, next_token_seq)[:, -1:]
|
85 |
scores = torch.softmax(logits / temp, dim=-1) * mask
|
86 |
+
samples = model.sample_top_p_k(scores, top_p, top_k, generator=generator)
|
87 |
if i == 0:
|
88 |
+
next_token_seq = samples
|
89 |
+
for b in range(batch_size):
|
90 |
+
if end[b]:
|
91 |
+
continue
|
92 |
+
eid = samples[b].item()
|
93 |
+
if eid == tokenizer.eos_id:
|
94 |
+
end[b] = True
|
95 |
+
else:
|
96 |
+
event_names[b] = tokenizer.id_events[eid]
|
97 |
else:
|
98 |
+
next_token_seq = torch.cat([next_token_seq, samples], dim=1)
|
99 |
+
if all([len(tokenizer.events[event_names[b]]) == i for b in range(batch_size) if not end[b]]):
|
100 |
break
|
101 |
if next_token_seq.shape[1] < max_token_seq:
|
102 |
next_token_seq = F.pad(next_token_seq, (0, max_token_seq - next_token_seq.shape[1]),
|
|
|
105 |
input_tensor = torch.cat([input_tensor, next_token_seq], dim=1)
|
106 |
cur_len += 1
|
107 |
bar.update(1)
|
108 |
+
yield next_token_seq[:, 0].cpu().numpy()
|
109 |
+
if all(end):
|
110 |
break
|
111 |
|
112 |
|
|
|
117 |
def send_msgs(msgs):
|
118 |
return json.dumps(msgs)
|
119 |
|
120 |
+
|
121 |
+
def get_duration(model_name, tab, mid_seq, continuation_state, continuation_select, instruments, drum_kit, bpm,
|
122 |
+
time_sig, key_sig, mid, midi_events, reduce_cc_st, remap_track_channel, add_default_instr,
|
123 |
remove_empty_channels, seed, seed_rand, gen_events, temp, top_p, top_k, allow_cc):
|
124 |
if "large" in model_name:
|
125 |
return gen_events // 10 + 15
|
|
|
128 |
|
129 |
|
130 |
@spaces.GPU(duration=get_duration)
|
131 |
+
def run(model_name, tab, mid_seq, continuation_state, continuation_select, instruments, drum_kit, bpm, time_sig,
|
132 |
+
key_sig, mid, midi_events, reduce_cc_st, remap_track_channel, add_default_instr, remove_empty_channels,
|
133 |
+
seed, seed_rand, gen_events, temp, top_p, top_k, allow_cc):
|
134 |
model = models[model_name]
|
135 |
model.to(device=opt.device)
|
136 |
tokenizer = model.tokenizer
|
|
|
178 |
patches[9] = drum_kits2number[drum_kit]
|
179 |
for i, (c, p) in enumerate(patches.items()):
|
180 |
mid.append(tokenizer.event2tokens(["patch_change", 0, 0, i + 1, c, p]))
|
181 |
+
mid = np.asarray([mid] * OUTPUT_BATCH_SIZE, dtype=np.int64)
|
182 |
+
mid_seq = mid.tolist()
|
183 |
if len(instruments) > 0:
|
184 |
disable_patch_change = True
|
185 |
disable_channels = [i for i in range(16) if i not in patches]
|
|
|
189 |
remap_track_channel=remap_track_channel,
|
190 |
add_default_instr=add_default_instr,
|
191 |
remove_empty_channels=remove_empty_channels)
|
|
|
192 |
mid = mid[:int(midi_events)]
|
193 |
+
mid = np.asarray([mid] * OUTPUT_BATCH_SIZE, dtype=np.int64)
|
194 |
+
mid_seq = mid.tolist()
|
|
|
195 |
elif tab == 2 and mid_seq is not None:
|
|
|
196 |
mid = np.asarray(mid_seq, dtype=np.int64)
|
197 |
+
if continuation_select > 0:
|
198 |
+
continuation_state.append(mid_seq)
|
199 |
+
mid = np.repeat(mid[continuation_select - 1:continuation_select], repeats=OUTPUT_BATCH_SIZE, axis=0)
|
200 |
+
mid_seq = mid.tolist()
|
201 |
+
else:
|
202 |
+
continuation_state.append(mid.shape[1])
|
203 |
else:
|
204 |
continuation_state = [0]
|
205 |
+
mid = [[tokenizer.bos_id] + [tokenizer.pad_id] * (tokenizer.max_token_seq - 1)]
|
206 |
+
mid = np.asarray([mid] * OUTPUT_BATCH_SIZE, dtype=np.int64)
|
207 |
+
mid_seq = mid.tolist()
|
208 |
|
209 |
if mid is not None:
|
210 |
+
max_len += mid.shape[1]
|
211 |
|
|
|
212 |
init_msgs = [create_msg("progress", [0, gen_events])]
|
213 |
+
if not (tab == 2 and continuation_select == 0):
|
214 |
+
for i in range(OUTPUT_BATCH_SIZE):
|
215 |
+
events = [tokenizer.tokens2event(tokens) for tokens in mid_seq[i]]
|
216 |
+
init_msgs += [create_msg("visualizer_clear", [i, tokenizer.version]),
|
217 |
+
create_msg("visualizer_append", [i, events])]
|
218 |
+
yield mid_seq, continuation_state, seed, send_msgs(init_msgs)
|
219 |
+
midi_generator = generate(model, mid, batch_size=OUTPUT_BATCH_SIZE, max_len=max_len, temp=temp,
|
220 |
+
top_p=top_p, top_k=top_k, disable_patch_change=disable_patch_change,
|
221 |
+
disable_control_change=not allow_cc, disable_channels=disable_channels,
|
222 |
+
generator=generator)
|
223 |
+
events = [list() for i in range(OUTPUT_BATCH_SIZE)]
|
224 |
+
t = time.time()
|
225 |
+
for i, token_seqs in enumerate(midi_generator):
|
226 |
+
token_seqs = token_seqs.tolist()
|
227 |
+
for j in range(OUTPUT_BATCH_SIZE):
|
228 |
+
token_seq = token_seqs[j]
|
229 |
+
mid_seq[j].append(token_seq)
|
230 |
+
events[j].append(tokenizer.tokens2event(token_seq))
|
231 |
+
if time.time() - t > 0.2:
|
232 |
+
msgs = [create_msg("progress", [i + 1, gen_events])]
|
233 |
+
for j in range(OUTPUT_BATCH_SIZE):
|
234 |
+
msgs += [create_msg("visualizer_append", [j, events[j]])]
|
235 |
+
events[j] = list()
|
236 |
+
yield mid_seq, continuation_state, seed, send_msgs(msgs)
|
237 |
+
t = time.time()
|
238 |
+
yield mid_seq, continuation_state, seed, send_msgs([])
|
239 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
240 |
|
241 |
+
def finish_run(model_name, mid_seq):
|
|
|
242 |
if mid_seq is None:
|
243 |
return None, None, []
|
244 |
tokenizer = models[model_name].tokenizer
|
245 |
+
outputs = []
|
246 |
+
end_msgs = [create_msg("progress", [0, 0])]
|
247 |
+
if not os.path.exists("outputs"):
|
248 |
+
os.mkdir("outputs")
|
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 += [(44100, audio), f"outputs/output{i + 1}.mid"]
|
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([])
|
|
|
|
|
265 |
tokenizer = models[model_name].tokenizer
|
266 |
+
if isinstance(continuation_state[-1], list):
|
267 |
+
mid_seq = continuation_state[-1]
|
268 |
+
else:
|
269 |
+
mid_seq = [ms[:continuation_state[-1]] for ms in mid_seq]
|
270 |
+
continuation_state = continuation_state[:-1]
|
271 |
+
end_msgs = [create_msg("progress", [0, 0])]
|
272 |
+
for i in range(OUTPUT_BATCH_SIZE):
|
273 |
+
events = [tokenizer.tokens2event(tokens) for tokens in mid_seq[i]]
|
274 |
+
end_msgs += [create_msg("visualizer_clear", [i, tokenizer.version]),
|
275 |
+
create_msg("visualizer_append", [i, events]),
|
276 |
+
create_msg("visualizer_end", i)]
|
277 |
return mid_seq, continuation_state, send_msgs(end_msgs)
|
278 |
|
279 |
|
|
|
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"],
|
330 |
+
"generic pretrain model (tv2o-large) by asigalov61": ["asigalov61/Music-Llama", "", "tv2o-large"],
|
331 |
+
"generic pretrain model (tv2o-medium) by asigalov61": ["asigalov61/Music-Llama-Medium", "", "tv2o-medium"],
|
332 |
+
"generic pretrain model (tv1-medium) by skytnt": ["skytnt/midi-model", "", "tv1-medium"],
|
333 |
+
"j-pop finetune model (tv2o-medium) by skytnt": ["skytnt/midi-model-ft", "jpop-tv2o-medium/", "tv2o-medium"],
|
334 |
+
"touhou finetune model (tv2o-medium) by skytnt": ["skytnt/midi-model-ft", "touhou-tv2o-medium/", "tv2o-medium"],
|
335 |
+
}
|
336 |
models = {}
|
337 |
if opt.device == "cuda":
|
338 |
torch.backends.cudnn.deterministic = True
|
|
|
421 |
example2 = gr.Examples([[file, 128] for file in glob.glob("example/*.mid")],
|
422 |
[input_midi, input_midi_events])
|
423 |
with gr.TabItem("last output prompt") as tab3:
|
424 |
+
gr.Markdown("Continue generating on the last output.")
|
425 |
+
input_continuation_select = gr.Radio(label="select output to continue generating", value="all",
|
426 |
+
choices=["all"] + [f"output{i + 1}" for i in
|
427 |
+
range(OUTPUT_BATCH_SIZE)],
|
428 |
+
type="index"
|
429 |
+
)
|
430 |
undo_btn = gr.Button("undo the last continuation")
|
431 |
|
432 |
tab1.select(lambda: 0, None, tab_select, queue=False)
|
|
|
448 |
stop_btn = gr.Button("stop and output")
|
449 |
output_midi_seq = gr.State()
|
450 |
output_continuation_state = gr.State([0])
|
451 |
+
batch_outputs = []
|
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 |
+
batch_outputs += [output_audio, output_midi]
|
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,
|
462 |
+
input_reduce_cc_st, input_remap_track_channel,
|
463 |
input_add_default_instr, input_remove_empty_channels,
|
464 |
input_seed, input_seed_rand, input_gen_events, input_temp, input_top_p,
|
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 |
+
inputs=[input_model, output_midi_seq],
|
470 |
+
outputs=batch_outputs + [js_msg],
|
471 |
+
queue=False)
|
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)
|
javascript/app.js
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
/**
|
2 |
* 自动绕过 shadowRoot 的 querySelector
|
3 |
* @param {string} selector - 要查询的 CSS 选择器
|
@@ -594,33 +596,49 @@ class MidiVisualizer extends HTMLElement{
|
|
594 |
customElements.define('midi-visualizer', MidiVisualizer);
|
595 |
|
596 |
(()=>{
|
597 |
-
|
598 |
-
|
599 |
-
|
600 |
-
|
601 |
-
|
602 |
-
|
603 |
-
|
604 |
-
|
605 |
-
|
606 |
-
|
607 |
-
}
|
608 |
-
let midi_audio = app.querySelector("#midi_audio");
|
609 |
-
if (!!midi_audio){
|
610 |
-
let midi_audio_cursor = midi_audio.deepQuerySelector(".cursor");
|
611 |
-
if(!!midi_audio_cursor && midi_audio_cursor_inited!==midi_audio_cursor){
|
612 |
-
midi_visualizer.bindWaveformCursor(midi_audio_cursor)
|
613 |
-
midi_audio_cursor_inited = midi_audio_cursor
|
614 |
}
|
615 |
-
let
|
616 |
-
if(!!
|
617 |
-
|
618 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
619 |
}
|
620 |
-
}
|
621 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
622 |
|
623 |
let hasProgressBar = false;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
624 |
|
625 |
function createProgressBar(progressbarContainer){
|
626 |
let parentProgressbar = progressbarContainer.parentNode;
|
@@ -653,15 +671,15 @@ customElements.define('midi-visualizer', MidiVisualizer);
|
|
653 |
hasProgressBar = false;
|
654 |
}
|
655 |
|
656 |
-
function setProgressBar(
|
657 |
if (!hasProgressBar)
|
658 |
-
createProgressBar(
|
659 |
if (hasProgressBar && total === 0){
|
660 |
-
removeProgressBar(
|
661 |
return
|
662 |
}
|
663 |
-
let parentProgressbar =
|
664 |
-
let divProgress = parentProgressbar.querySelector(".progressDiv");
|
665 |
let divInner = parentProgressbar.querySelector(".progress");
|
666 |
if(total===0)
|
667 |
total = 1;
|
@@ -679,24 +697,30 @@ customElements.define('midi-visualizer', MidiVisualizer);
|
|
679 |
}
|
680 |
})
|
681 |
function handleMsg(msg){
|
|
|
682 |
switch (msg.name) {
|
683 |
case "visualizer_clear":
|
684 |
-
|
685 |
-
|
|
|
|
|
686 |
break;
|
687 |
case "visualizer_append":
|
688 |
-
msg.data
|
689 |
-
|
|
|
|
|
690 |
})
|
691 |
break;
|
692 |
case "visualizer_end":
|
693 |
-
|
694 |
-
|
|
|
695 |
break;
|
696 |
case "progress":
|
697 |
let progress = msg.data[0]
|
698 |
let total = msg.data[1]
|
699 |
-
setProgressBar(
|
700 |
break;
|
701 |
default:
|
702 |
}
|
|
|
1 |
+
const MIDI_OUTPUT_BATCH_SIZE=4;
|
2 |
+
|
3 |
/**
|
4 |
* 自动绕过 shadowRoot 的 querySelector
|
5 |
* @param {string} selector - 要查询的 CSS 选择器
|
|
|
596 |
customElements.define('midi-visualizer', MidiVisualizer);
|
597 |
|
598 |
(()=>{
|
599 |
+
function midi_visualizer_setup(idx, midi_visualizer){
|
600 |
+
let midi_visualizer_container_inited = null
|
601 |
+
let midi_audio_audio_inited = null;
|
602 |
+
let midi_audio_cursor_inited = null;
|
603 |
+
onUiUpdate((m)=>{
|
604 |
+
let app = gradioApp()
|
605 |
+
let midi_visualizer_container = app.querySelector(`#midi_visualizer_container_${idx}`);
|
606 |
+
if(!!midi_visualizer_container && midi_visualizer_container_inited!== midi_visualizer_container){
|
607 |
+
midi_visualizer_container.appendChild(midi_visualizer)
|
608 |
+
midi_visualizer_container_inited = midi_visualizer_container;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
609 |
}
|
610 |
+
let midi_audio = app.querySelector(`#midi_audio_${idx}`);
|
611 |
+
if (!!midi_audio){
|
612 |
+
let midi_audio_cursor = midi_audio.deepQuerySelector(".cursor");
|
613 |
+
if(!!midi_audio_cursor && midi_audio_cursor_inited!==midi_audio_cursor){
|
614 |
+
midi_visualizer.bindWaveformCursor(midi_audio_cursor)
|
615 |
+
midi_audio_cursor_inited = midi_audio_cursor
|
616 |
+
}
|
617 |
+
let midi_audio_audio = midi_audio.deepQuerySelector("audio");
|
618 |
+
if(!!midi_audio_audio && midi_audio_audio_inited!==midi_audio_audio){
|
619 |
+
midi_visualizer.bindAudioPlayer(midi_audio_audio)
|
620 |
+
midi_audio_audio_inited = midi_audio_audio
|
621 |
+
}
|
622 |
}
|
623 |
+
});
|
624 |
+
}
|
625 |
+
|
626 |
+
let midi_visualizers = []
|
627 |
+
for (let i = 0; i < MIDI_OUTPUT_BATCH_SIZE ; i++){
|
628 |
+
let midi_visualizer = document.createElement('midi-visualizer');
|
629 |
+
midi_visualizers.push(midi_visualizer);
|
630 |
+
midi_visualizer_setup(i, midi_visualizer)
|
631 |
+
}
|
632 |
|
633 |
let hasProgressBar = false;
|
634 |
+
let output_tabs_inited = null;
|
635 |
+
onUiUpdate((m)=>{
|
636 |
+
let app = gradioApp()
|
637 |
+
let output_tabs = app.querySelector("#output_tabs");
|
638 |
+
if(!!output_tabs && output_tabs_inited!== output_tabs){
|
639 |
+
output_tabs_inited = output_tabs;
|
640 |
+
}
|
641 |
+
});
|
642 |
|
643 |
function createProgressBar(progressbarContainer){
|
644 |
let parentProgressbar = progressbarContainer.parentNode;
|
|
|
671 |
hasProgressBar = false;
|
672 |
}
|
673 |
|
674 |
+
function setProgressBar(progress, total){
|
675 |
if (!hasProgressBar)
|
676 |
+
createProgressBar(output_tabs_inited)
|
677 |
if (hasProgressBar && total === 0){
|
678 |
+
removeProgressBar(output_tabs_inited)
|
679 |
return
|
680 |
}
|
681 |
+
let parentProgressbar = output_tabs_inited.parentNode;
|
682 |
+
// let divProgress = parentProgressbar.querySelector(".progressDiv");
|
683 |
let divInner = parentProgressbar.querySelector(".progress");
|
684 |
if(total===0)
|
685 |
total = 1;
|
|
|
697 |
}
|
698 |
})
|
699 |
function handleMsg(msg){
|
700 |
+
let idx;
|
701 |
switch (msg.name) {
|
702 |
case "visualizer_clear":
|
703 |
+
idx = msg.data[0];
|
704 |
+
let ver = msg.data[1];
|
705 |
+
midi_visualizers[idx].clearMidiEvents(false);
|
706 |
+
midi_visualizers[idx].version = ver;
|
707 |
break;
|
708 |
case "visualizer_append":
|
709 |
+
idx = msg.data[0];
|
710 |
+
let events = msg.data[1];
|
711 |
+
events.forEach( value => {
|
712 |
+
midi_visualizers[idx].appendMidiEvent(value);
|
713 |
})
|
714 |
break;
|
715 |
case "visualizer_end":
|
716 |
+
idx = msg.data;
|
717 |
+
midi_visualizers[idx].finishAppendMidiEvent()
|
718 |
+
midi_visualizers[idx].setPlayTime(0);
|
719 |
break;
|
720 |
case "progress":
|
721 |
let progress = msg.data[0]
|
722 |
let total = msg.data[1]
|
723 |
+
setProgressBar(progress, total)
|
724 |
break;
|
725 |
default:
|
726 |
}
|
midi_model.py
CHANGED
@@ -111,49 +111,69 @@ class MIDIModel(nn.Module):
|
|
111 |
return next_token
|
112 |
|
113 |
@torch.inference_mode()
|
114 |
-
def generate(self, prompt=None, max_len=512, temp=1.0, top_p=0.98, top_k=20, generator=None):
|
115 |
tokenizer = self.tokenizer
|
116 |
max_token_seq = tokenizer.max_token_seq
|
117 |
if prompt is None:
|
118 |
input_tensor = torch.full((1, max_token_seq), tokenizer.pad_id, dtype=torch.long, device=self.device)
|
119 |
input_tensor[0, 0] = tokenizer.bos_id # bos
|
|
|
|
|
120 |
else:
|
121 |
-
prompt
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
122 |
if prompt.shape[-1] < max_token_seq:
|
123 |
-
prompt = np.pad(prompt, ((0, 0), (0, max_token_seq - prompt.shape[-1])),
|
124 |
mode="constant", constant_values=tokenizer.pad_id)
|
125 |
input_tensor = torch.from_numpy(prompt).to(dtype=torch.long, device=self.device)
|
126 |
-
|
127 |
cur_len = input_tensor.shape[1]
|
128 |
bar = tqdm.tqdm(desc="generating", total=max_len - cur_len)
|
129 |
with bar:
|
130 |
while cur_len < max_len:
|
131 |
-
end = False
|
132 |
-
hidden = self.forward(input_tensor)[
|
133 |
next_token_seq = None
|
134 |
-
|
135 |
for i in range(max_token_seq):
|
136 |
-
mask = torch.zeros(tokenizer.vocab_size, dtype=torch.int64, device=self.device)
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
143 |
logits = self.forward_token(hidden, next_token_seq)[:, -1:]
|
144 |
scores = torch.softmax(logits / temp, dim=-1) * mask
|
145 |
-
|
146 |
if i == 0:
|
147 |
-
next_token_seq =
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
|
|
|
|
|
|
153 |
else:
|
154 |
-
next_token_seq = torch.cat([next_token_seq,
|
155 |
-
if len(tokenizer.events[
|
156 |
break
|
|
|
157 |
if next_token_seq.shape[1] < max_token_seq:
|
158 |
next_token_seq = F.pad(next_token_seq, (0, max_token_seq - next_token_seq.shape[1]),
|
159 |
"constant", value=tokenizer.pad_id)
|
@@ -161,6 +181,7 @@ class MIDIModel(nn.Module):
|
|
161 |
input_tensor = torch.cat([input_tensor, next_token_seq], dim=1)
|
162 |
cur_len += 1
|
163 |
bar.update(1)
|
164 |
-
|
|
|
165 |
break
|
166 |
-
return input_tensor
|
|
|
111 |
return next_token
|
112 |
|
113 |
@torch.inference_mode()
|
114 |
+
def generate(self, prompt=None, batch_size=1, max_len=512, temp=1.0, top_p=0.98, top_k=20, generator=None):
|
115 |
tokenizer = self.tokenizer
|
116 |
max_token_seq = tokenizer.max_token_seq
|
117 |
if prompt is None:
|
118 |
input_tensor = torch.full((1, max_token_seq), tokenizer.pad_id, dtype=torch.long, device=self.device)
|
119 |
input_tensor[0, 0] = tokenizer.bos_id # bos
|
120 |
+
input_tensor = input_tensor.unsqueeze(0)
|
121 |
+
input_tensor = torch.cat([input_tensor] * batch_size, dim=0)
|
122 |
else:
|
123 |
+
if len(prompt.shape) == 2:
|
124 |
+
prompt = prompt[None, :]
|
125 |
+
prompt = np.repeat(prompt, repeats=batch_size, axis=0)
|
126 |
+
elif prompt.shape[0] == 1:
|
127 |
+
prompt = np.repeat(prompt, repeats=batch_size, axis=0)
|
128 |
+
else:
|
129 |
+
raise ValueError(f"invalid shape for prompt, {prompt.shape}")
|
130 |
+
prompt = prompt[..., :max_token_seq]
|
131 |
if prompt.shape[-1] < max_token_seq:
|
132 |
+
prompt = np.pad(prompt, ((0, 0), (0, 0), (0, max_token_seq - prompt.shape[-1])),
|
133 |
mode="constant", constant_values=tokenizer.pad_id)
|
134 |
input_tensor = torch.from_numpy(prompt).to(dtype=torch.long, device=self.device)
|
135 |
+
|
136 |
cur_len = input_tensor.shape[1]
|
137 |
bar = tqdm.tqdm(desc="generating", total=max_len - cur_len)
|
138 |
with bar:
|
139 |
while cur_len < max_len:
|
140 |
+
end = [False] * batch_size
|
141 |
+
hidden = self.forward(input_tensor)[:, -1]
|
142 |
next_token_seq = None
|
143 |
+
event_names = [""] * batch_size
|
144 |
for i in range(max_token_seq):
|
145 |
+
mask = torch.zeros((batch_size, tokenizer.vocab_size), dtype=torch.int64, device=self.device)
|
146 |
+
for b in range(batch_size):
|
147 |
+
if end[b]:
|
148 |
+
mask[b, tokenizer.pad_id] = 1
|
149 |
+
continue
|
150 |
+
if i == 0:
|
151 |
+
mask[b, list(tokenizer.event_ids.values()) + [tokenizer.eos_id]] = 1
|
152 |
+
else:
|
153 |
+
param_names = tokenizer.events[event_names[b]]
|
154 |
+
if i > len(param_names):
|
155 |
+
mask[b, tokenizer.pad_id] = 1
|
156 |
+
continue
|
157 |
+
mask[b, tokenizer.parameter_ids[param_names[i - 1]]] = 1
|
158 |
+
mask = mask.unsqueeze(1)
|
159 |
logits = self.forward_token(hidden, next_token_seq)[:, -1:]
|
160 |
scores = torch.softmax(logits / temp, dim=-1) * mask
|
161 |
+
samples = self.sample_top_p_k(scores, top_p, top_k, generator=generator)
|
162 |
if i == 0:
|
163 |
+
next_token_seq = samples
|
164 |
+
for b in range(batch_size):
|
165 |
+
if end[b]:
|
166 |
+
continue
|
167 |
+
eid = samples[b].item()
|
168 |
+
if eid == tokenizer.eos_id:
|
169 |
+
end[b] = True
|
170 |
+
else:
|
171 |
+
event_names[b] = tokenizer.id_events[eid]
|
172 |
else:
|
173 |
+
next_token_seq = torch.cat([next_token_seq, samples], dim=1)
|
174 |
+
if all([len(tokenizer.events[event_names[b]]) == i for b in range(batch_size) if not end[b]]):
|
175 |
break
|
176 |
+
|
177 |
if next_token_seq.shape[1] < max_token_seq:
|
178 |
next_token_seq = F.pad(next_token_seq, (0, max_token_seq - next_token_seq.shape[1]),
|
179 |
"constant", value=tokenizer.pad_id)
|
|
|
181 |
input_tensor = torch.cat([input_tensor, next_token_seq], dim=1)
|
182 |
cur_len += 1
|
183 |
bar.update(1)
|
184 |
+
|
185 |
+
if all(end):
|
186 |
break
|
187 |
+
return input_tensor.cpu().numpy()
|