File size: 11,552 Bytes
9a76abc d640048 9a76abc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 | import gradio as gr
from app.original import *
def train_tabs():
with gr.TabItem(i18n("训练")):
gr.Markdown(
value=i18n(
"step1: 填写实验配置. 实验数据放在logs下, 每个实验一个文件夹, 需手工输入实验名路径, 内含实验配置, 日志, 训练得到的模型文件. "
)
)
with gr.Row():
exp_dir1 = gr.Textbox(label=i18n("输入实验名"), value="mi-test")
sr2 = gr.Radio(
label=i18n("目标采样率"),
choices=["40k", "48k"],
value="40k",
interactive=True,
)
if_f0_3 = gr.Radio(
label=i18n("模型是否带音高指导(唱歌一定要, 语音可以不要)"),
choices=[i18n("是"), i18n("否")],
value=i18n("是"),
interactive=True,
)
version19 = gr.Radio(
label=i18n("版本"),
choices=["v1", "v2"],
value="v2",
interactive=True,
visible=True,
)
np7 = gr.Slider(
minimum=0,
maximum=config.n_cpu,
step=1,
label=i18n("提取音高和处理数据使用的CPU进程数"),
value=int(np.ceil(config.n_cpu / 1.5)),
interactive=True,
)
with gr.Group(): # 暂时单人的, 后面支持最多4人的#数据处理
gr.Markdown(
value=i18n(
"step2a: 自动遍历训练文件夹下所有可解码成音频的文件并进行切片归一化, 在实验目录下生成2个wav文件夹; 暂时只支持单人训练. "
)
)
with gr.Row():
trainset_dir4 = gr.Textbox(
label=i18n("输入训练文件夹路径"),
value=i18n("E:\\语音音频+标注\\米津玄师\\src"),
)
spk_id5 = gr.Slider(
minimum=0,
maximum=4,
step=1,
label=i18n("请指定说话人id"),
value=0,
interactive=True,
)
but1 = gr.Button(i18n("处理数据"), variant="primary")
info1 = gr.Textbox(label=i18n("输出信息"), value="")
but1.click(
preprocess_dataset,
[trainset_dir4, exp_dir1, sr2, np7],
[info1],
api_name="train_preprocess",
)
with gr.Group():
gr.Markdown(
value=i18n(
"step2b: 使用CPU提取音高(如果模型带音高), 使用GPU提取特征(选择卡号)"
)
)
with gr.Row():
with gr.Column():
gpus6 = gr.Textbox(
label=i18n(
"以-分隔输入使用的卡号, 例如 0-1-2 使用卡0和卡1和卡2"
),
value=gpus,
interactive=True,
visible=F0GPUVisible,
)
gpu_info9 = gr.Textbox(
label=i18n("显卡信息"), value=gpu_info, visible=F0GPUVisible
)
with gr.Column():
f0method8 = gr.Radio(
label=i18n(
"选择音高提取算法:输入歌声可用pm提速,高质量语音但CPU差可用dio提速,harvest质量更好但慢,rmvpe效果最好且微吃CPU/GPU"
),
choices=["pm", "harvest", "dio", "rmvpe", "rmvpe_gpu"],
value="rmvpe_gpu",
interactive=True,
)
gpus_rmvpe = gr.Textbox(
label=i18n(
"rmvpe卡号配置：以-分隔输入使用的不同进程卡号,例如0-0-1使用在卡0上跑2个进程并在卡1上跑1个进程"
),
value="%s-%s" % (gpus, gpus),
interactive=True,
visible=F0GPUVisible,
)
but2 = gr.Button(i18n("特征提取"), variant="primary")
info2 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=8)
f0method8.change(
fn=change_f0_method,
inputs=[f0method8],
outputs=[gpus_rmvpe],
)
but2.click(
extract_f0_feature,
[
gpus6,
np7,
f0method8,
if_f0_3,
exp_dir1,
version19,
gpus_rmvpe,
],
[info2],
api_name="train_extract_f0_feature",
)
with gr.Group():
gr.Markdown(value=i18n("step3: 填写训练设置, 开始训练模型和索引"))
with gr.Row():
save_epoch10 = gr.Slider(
minimum=1,
maximum=50,
step=1,
label=i18n("保存频率save_every_epoch"),
value=5,
interactive=True,
)
total_epoch11 = gr.Slider(
minimum=2,
maximum=1000,
step=1,
label=i18n("总训练轮数total_epoch"),
value=20,
interactive=True,
)
batch_size12 = gr.Slider(
minimum=1,
maximum=40,
step=1,
label=i18n("每张显卡的batch_size"),
value=default_batch_size,
interactive=True,
)
if_save_latest13 = gr.Radio(
label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"),
choices=[i18n("是"), i18n("否")],
value=i18n("否"),
interactive=True,
)
if_cache_gpu17 = gr.Radio(
label=i18n(
"是否缓存所有训练集至显存. 10min以下小数据可缓存以加速训练, 大数据缓存会炸显存也加不了多少速"
),
choices=[i18n("是"), i18n("否")],
value=i18n("否"),
interactive=True,
)
if_save_every_weights18 = gr.Radio(
label=i18n(
"是否在每次保存时间点将最终小模型保存至weights文件夹"
),
choices=[i18n("是"), i18n("否")],
value=i18n("否"),
interactive=True,
)
with gr.Row():
pretrained_G14 = gr.Textbox(
label=i18n("加载预训练底模G路径"),
value="assets/pretrained_v2/f0G40k.pth",
interactive=True,
)
pretrained_D15 = gr.Textbox(
label=i18n("加载预训练底模D路径"),
value="assets/pretrained_v2/f0D40k.pth",
interactive=True,
)
sr2.change(
change_sr2,
[sr2, if_f0_3, version19],
[pretrained_G14, pretrained_D15],
)
version19.change(
change_version19,
[sr2, if_f0_3, version19],
[pretrained_G14, pretrained_D15, sr2],
)
if_f0_3.change(
change_f0,
[if_f0_3, sr2, version19],
[f0method8, gpus_rmvpe, pretrained_G14, pretrained_D15],
)
gpus16 = gr.Textbox(
label=i18n(
"以-分隔输入使用的卡号, 例如 0-1-2 使用卡0和卡1和卡2"
),
value=gpus,
interactive=True,
)
but3 = gr.Button(i18n("训练模型"), variant="primary")
but4 = gr.Button(i18n("训练特征索引"), variant="primary")
but5 = gr.Button(i18n("一键训练"), variant="primary")
info3 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=10)
but3.click(
click_train,
[
exp_dir1,
sr2,
if_f0_3,
spk_id5,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
],
info3,
api_name="train_start",
)
but4.click(train_index, [exp_dir1, version19], info3)
but5.click(
train1key,
[
exp_dir1,
sr2,
if_f0_3,
trainset_dir4,
spk_id5,
np7,
f0method8,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
gpus_rmvpe,
],
info3,
api_name="train_start_all",
)
|