Spaces:
Runtime error
Runtime error
Upload app.py
Browse files
app.py
CHANGED
|
@@ -5,8 +5,8 @@ import pytz
|
|
| 5 |
from pathlib import Path
|
| 6 |
|
| 7 |
def current_time():
|
| 8 |
-
|
| 9 |
-
|
| 10 |
|
| 11 |
print(f"[{current_time()}] 开始部署空间...")
|
| 12 |
|
|
@@ -83,58 +83,58 @@ SAMPLE_RATE = 16000
|
|
| 83 |
SF2_PATH = 'SGM-v2.01-Sal-Guit-Bass-V1.3.sf2'
|
| 84 |
|
| 85 |
def upload_audio(audio, sample_rate):
|
| 86 |
-
|
| 87 |
-
|
| 88 |
|
| 89 |
|
| 90 |
print(f"[{current_time()}] 日志:开始包装模型...")
|
| 91 |
class InferenceModel(object):
|
| 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 |
return {
|
| 139 |
'encoder_input_tokens': (self.batch_size, self.inputs_length),
|
| 140 |
'decoder_input_tokens': (self.batch_size, self.outputs_length)
|
|
@@ -144,10 +144,10 @@ class InferenceModel(object):
|
|
| 144 |
"""解析用于训练模型的 gin 文件。"""
|
| 145 |
print(f"[{current_time()}] 日志:解析 gin 文件")
|
| 146 |
gin_bindings = [
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
]
|
| 152 |
with gin.unlock_config():
|
| 153 |
gin.parse_config_files_and_bindings(gin_files, gin_bindings, finalize_config=False)
|
|
@@ -158,11 +158,11 @@ class InferenceModel(object):
|
|
| 158 |
model_config = gin.get_configurable(network.T5Config)()
|
| 159 |
module = network.Transformer(config=model_config)
|
| 160 |
return models.ContinuousInputsEncoderDecoderModel(
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
|
| 167 |
|
| 168 |
def restore_from_checkpoint(self, checkpoint_path):
|
|
@@ -175,12 +175,12 @@ class InferenceModel(object):
|
|
| 175 |
partitioner=self.partitioner)
|
| 176 |
|
| 177 |
restore_checkpoint_cfg = t5x.utils.RestoreCheckpointConfig(
|
| 178 |
-
|
| 179 |
|
| 180 |
train_state_axes = train_state_initializer.train_state_axes
|
| 181 |
self._predict_fn = self._get_predict_fn(train_state_axes)
|
| 182 |
self._train_state = train_state_initializer.from_checkpoint_or_scratch(
|
| 183 |
-
|
| 184 |
|
| 185 |
@functools.lru_cache()
|
| 186 |
def _get_predict_fn(self, train_state_axes):
|
|
@@ -189,11 +189,11 @@ class InferenceModel(object):
|
|
| 189 |
def partial_predict_fn(params, batch, decode_rng):
|
| 190 |
return self.model.predict_batch_with_aux(params, batch, decoder_params={'decode_rng': None})
|
| 191 |
return self.partitioner.partition(
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
)
|
| 198 |
|
| 199 |
def predict_tokens(self, batch, seed=0):
|
|
@@ -252,16 +252,16 @@ class InferenceModel(object):
|
|
| 252 |
def preprocess(self, ds):
|
| 253 |
pp_chain = [
|
| 254 |
functools.partial(
|
| 255 |
-
|
| 256 |
-
|
| 257 |
-
|
| 258 |
-
|
| 259 |
-
|
| 260 |
# 在训练期间进行缓存。
|
| 261 |
preprocessors.add_dummy_targets,
|
| 262 |
functools.partial(
|
| 263 |
-
|
| 264 |
-
|
| 265 |
]
|
| 266 |
for pp in pp_chain:
|
| 267 |
ds = pp(ds)
|
|
@@ -273,10 +273,10 @@ class InferenceModel(object):
|
|
| 273 |
# 向下取整到最接近的符号化时间步。
|
| 274 |
start_time -= start_time % (1 / self.codec.steps_per_second)
|
| 275 |
return {
|
| 276 |
-
|
| 277 |
-
|
| 278 |
-
|
| 279 |
-
|
| 280 |
}
|
| 281 |
|
| 282 |
@staticmethod
|
|
@@ -308,11 +308,11 @@ article = "<p style='text-align: center'>出错了?试试把文件转换为MP3
|
|
| 308 |
examples=[['canon.flac'], ['download.wav']]
|
| 309 |
|
| 310 |
gr.Interface(
|
| 311 |
-
|
| 312 |
-
|
| 313 |
-
|
| 314 |
-
|
| 315 |
-
|
| 316 |
-
|
| 317 |
-
|
| 318 |
).launch(server_port=7861)
|
|
|
|
| 5 |
from pathlib import Path
|
| 6 |
|
| 7 |
def current_time():
|
| 8 |
+
current = datetime.datetime.now(pytz.timezone('Asia/Shanghai')).strftime("%Y年-%m月-%d日 %H时:%M分:%S秒")
|
| 9 |
+
return current
|
| 10 |
|
| 11 |
print(f"[{current_time()}] 开始部署空间...")
|
| 12 |
|
|
|
|
| 83 |
SF2_PATH = 'SGM-v2.01-Sal-Guit-Bass-V1.3.sf2'
|
| 84 |
|
| 85 |
def upload_audio(audio, sample_rate):
|
| 86 |
+
return note_seq.audio_io.wav_data_to_samples_librosa(
|
| 87 |
+
audio, sample_rate=sample_rate)
|
| 88 |
|
| 89 |
|
| 90 |
print(f"[{current_time()}] 日志:开始包装模型...")
|
| 91 |
class InferenceModel(object):
|
| 92 |
+
"""音乐转录的 T5X 模型包装器。"""
|
| 93 |
+
|
| 94 |
+
def __init__(self, checkpoint_path, model_type='mt3'):
|
| 95 |
+
if model_type == 'ismir2021':
|
| 96 |
+
num_velocity_bins = 127
|
| 97 |
+
self.encoding_spec = note_sequences.NoteEncodingSpec
|
| 98 |
+
self.inputs_length = 512
|
| 99 |
+
elif model_type == 'mt3':
|
| 100 |
+
num_velocity_bins = 1
|
| 101 |
+
self.encoding_spec = note_sequences.NoteEncodingWithTiesSpec
|
| 102 |
+
self.inputs_length = 256
|
| 103 |
+
else:
|
| 104 |
+
raise ValueError('unknown model_type: %s' % model_type)
|
| 105 |
+
|
| 106 |
+
gin_files = ['/home/user/app/mt3/gin/model.gin',
|
| 107 |
+
'/home/user/app/mt3/gin/mt3.gin']
|
| 108 |
+
|
| 109 |
+
self.batch_size = 8
|
| 110 |
+
self.outputs_length = 1024
|
| 111 |
+
self.sequence_length = {'inputs': self.inputs_length,
|
| 112 |
+
'targets': self.outputs_length}
|
| 113 |
+
|
| 114 |
+
self.partitioner = t5x.partitioning.PjitPartitioner(
|
| 115 |
+
model_parallel_submesh=None, num_partitions=1)
|
| 116 |
+
|
| 117 |
+
print(f"[{current_time()}] 日志:构建编解码器")
|
| 118 |
+
self.spectrogram_config = spectrograms.SpectrogramConfig()
|
| 119 |
+
self.codec = vocabularies.build_codec(
|
| 120 |
+
vocab_config=vocabularies.VocabularyConfig(
|
| 121 |
+
num_velocity_bins=num_velocity_bins)
|
| 122 |
+
)
|
| 123 |
+
self.vocabulary = vocabularies.vocabulary_from_codec(self.codec)
|
| 124 |
+
self.output_features = {
|
| 125 |
+
'inputs': seqio.ContinuousFeature(dtype=tf.float32, rank=2),
|
| 126 |
+
'targets': seqio.Feature(vocabulary=self.vocabulary),
|
| 127 |
+
}
|
| 128 |
+
|
| 129 |
+
print(f"[{current_time()}] 日志:创建 T5X 模型")
|
| 130 |
+
self._parse_gin(gin_files)
|
| 131 |
+
self.model = self._load_model()
|
| 132 |
+
|
| 133 |
+
print(f"[{current_time()}] 日志:恢复模型检查点")
|
| 134 |
+
self.restore_from_checkpoint(checkpoint_path)
|
| 135 |
+
|
| 136 |
+
@property
|
| 137 |
+
def input_shapes(self):
|
| 138 |
return {
|
| 139 |
'encoder_input_tokens': (self.batch_size, self.inputs_length),
|
| 140 |
'decoder_input_tokens': (self.batch_size, self.outputs_length)
|
|
|
|
| 144 |
"""解析用于训练模型的 gin 文件。"""
|
| 145 |
print(f"[{current_time()}] 日志:解析 gin 文件")
|
| 146 |
gin_bindings = [
|
| 147 |
+
'from __gin__ import dynamic_registration',
|
| 148 |
+
'from mt3 import vocabularies',
|
| 149 |
+
'[email protected]()',
|
| 150 |
+
'vocabularies.VocabularyConfig.num_velocity_bins=%NUM_VELOCITY_BINS'
|
| 151 |
]
|
| 152 |
with gin.unlock_config():
|
| 153 |
gin.parse_config_files_and_bindings(gin_files, gin_bindings, finalize_config=False)
|
|
|
|
| 158 |
model_config = gin.get_configurable(network.T5Config)()
|
| 159 |
module = network.Transformer(config=model_config)
|
| 160 |
return models.ContinuousInputsEncoderDecoderModel(
|
| 161 |
+
module=module,
|
| 162 |
+
input_vocabulary=self.output_features['inputs'].vocabulary,
|
| 163 |
+
output_vocabulary=self.output_features['targets'].vocabulary,
|
| 164 |
+
optimizer_def=t5x.adafactor.Adafactor(decay_rate=0.8, step_offset=0),
|
| 165 |
+
input_depth=spectrograms.input_depth(self.spectrogram_config))
|
| 166 |
|
| 167 |
|
| 168 |
def restore_from_checkpoint(self, checkpoint_path):
|
|
|
|
| 175 |
partitioner=self.partitioner)
|
| 176 |
|
| 177 |
restore_checkpoint_cfg = t5x.utils.RestoreCheckpointConfig(
|
| 178 |
+
path=checkpoint_path, mode='specific', dtype='float32')
|
| 179 |
|
| 180 |
train_state_axes = train_state_initializer.train_state_axes
|
| 181 |
self._predict_fn = self._get_predict_fn(train_state_axes)
|
| 182 |
self._train_state = train_state_initializer.from_checkpoint_or_scratch(
|
| 183 |
+
[restore_checkpoint_cfg], init_rng=jax.random.PRNGKey(0))
|
| 184 |
|
| 185 |
@functools.lru_cache()
|
| 186 |
def _get_predict_fn(self, train_state_axes):
|
|
|
|
| 189 |
def partial_predict_fn(params, batch, decode_rng):
|
| 190 |
return self.model.predict_batch_with_aux(params, batch, decoder_params={'decode_rng': None})
|
| 191 |
return self.partitioner.partition(
|
| 192 |
+
partial_predict_fn,
|
| 193 |
+
in_axis_resources=(
|
| 194 |
+
train_state_axes.params,
|
| 195 |
+
t5x.partitioning.PartitionSpec('data',), None),
|
| 196 |
+
out_axis_resources=t5x.partitioning.PartitionSpec('data',)
|
| 197 |
)
|
| 198 |
|
| 199 |
def predict_tokens(self, batch, seed=0):
|
|
|
|
| 252 |
def preprocess(self, ds):
|
| 253 |
pp_chain = [
|
| 254 |
functools.partial(
|
| 255 |
+
t5.data.preprocessors.split_tokens_to_inputs_length,
|
| 256 |
+
sequence_length=self.sequence_length,
|
| 257 |
+
output_features=self.output_features,
|
| 258 |
+
feature_key='inputs',
|
| 259 |
+
additional_feature_keys=['input_times']),
|
| 260 |
# 在训练期间进行缓存。
|
| 261 |
preprocessors.add_dummy_targets,
|
| 262 |
functools.partial(
|
| 263 |
+
preprocessors.compute_spectrograms,
|
| 264 |
+
spectrogram_config=self.spectrogram_config)
|
| 265 |
]
|
| 266 |
for pp in pp_chain:
|
| 267 |
ds = pp(ds)
|
|
|
|
| 273 |
# 向下取整到最接近的符号化时间步。
|
| 274 |
start_time -= start_time % (1 / self.codec.steps_per_second)
|
| 275 |
return {
|
| 276 |
+
'est_tokens': tokens,
|
| 277 |
+
'start_time': start_time,
|
| 278 |
+
# 内部 MT3 代码期望原始输入,这里不使用。
|
| 279 |
+
'raw_inputs': []
|
| 280 |
}
|
| 281 |
|
| 282 |
@staticmethod
|
|
|
|
| 308 |
examples=[['canon.flac'], ['download.wav']]
|
| 309 |
|
| 310 |
gr.Interface(
|
| 311 |
+
inference,
|
| 312 |
+
gr.Audio(type="filepath", label="输入"),
|
| 313 |
+
outputs=gr.File(label="输出"),
|
| 314 |
+
title=title,
|
| 315 |
+
description=description,
|
| 316 |
+
article=article,
|
| 317 |
+
examples=examples
|
| 318 |
).launch(server_port=7861)
|