diff --git a/.gitattributes b/.gitattributes
new file mode 100644
index 0000000000000000000000000000000000000000..c7d9f3332a950355d5a77d85000f05e6f45435ea
--- /dev/null
+++ b/.gitattributes
@@ -0,0 +1,34 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
diff --git a/README.md b/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..4941cee152fe1f4c77e73c527f3becd6a3543913
--- /dev/null
+++ b/README.md
@@ -0,0 +1,4 @@
+---
+license: mit
+duplicated_from: lj1995/VoiceConversionWebUI
+---
diff --git a/RVC-beta-v20230331.7z b/RVC-beta-v20230331.7z
new file mode 100644
index 0000000000000000000000000000000000000000..325769e0386b14ada6af50e395aa1cda434e68e4
--- /dev/null
+++ b/RVC-beta-v20230331.7z
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65724924d51c5baeed3ee00476fe9dead8f3d0e01ab3923d6eab7439a3993010
+size 3142549647
diff --git a/RVC-beta.7z b/RVC-beta.7z
new file mode 100644
index 0000000000000000000000000000000000000000..3ea8d0663c937a429ae7c483ce4d1aaacda56c2b
--- /dev/null
+++ b/RVC-beta.7z
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4783aaa047d3a84b6b072f0c331b184fa997e2f06e14cd4c7a4df7a8c5f21e8a
+size 3175821212
diff --git a/RVC-beta20230416.7z b/RVC-beta20230416.7z
new file mode 100644
index 0000000000000000000000000000000000000000..b40399b216c21604f0fe2f40bf22b74544a5c4dc
--- /dev/null
+++ b/RVC-beta20230416.7z
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f5aaa39f1ef195e575e16c46d1d39671687645e2f7684b332e334dd525d9238
+size 3147462195
diff --git a/RVC-beta20230428.7z b/RVC-beta20230428.7z
new file mode 100644
index 0000000000000000000000000000000000000000..45a975351a809d429c6f41aa4ee7c7b4a4f5d902
--- /dev/null
+++ b/RVC-beta20230428.7z
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87fe7619af6a6ddf58c9445b17beb567a620ed7efad5eb8bec1dededc05295c0
+size 3175819910
diff --git a/config.py b/config.py
new file mode 100644
index 0000000000000000000000000000000000000000..67ab01846a7a23988c58105fe03c031344387dd3
--- /dev/null
+++ b/config.py
@@ -0,0 +1,38 @@
+############离线VC参数
+inp_root=r"白鹭霜华长条"#对输入目录下所有音频进行转换，别放非音频文件
+opt_root=r"opt"#输出目录
+f0_up_key=0#升降调，整数，男转女12，女转男-12
+person=r"weights\洛天依v3.pt"#目前只有洛天依v3
+############硬件参数
+device = "cuda:0"#填写cuda:x或cpu，x指代第几张卡，只支持N卡加速
+is_half=True#9-10-20-30-40系显卡无脑True，不影响质量，>=20显卡开启有加速
+n_cpu=0#默认0用上所有线程，写数字限制CPU资源使用
+############下头别动
+import torch
+if(torch.cuda.is_available()==False):
+ print("没有发现支持的N卡，使用CPU进行推理")
+ device="cpu"
+ is_half=False
+if(device!="cpu"):
+ gpu_name=torch.cuda.get_device_name(int(device.split(":")[-1]))
+ if("16"in gpu_name):
+ print("16系显卡强制单精度")
+ is_half=False
+from multiprocessing import cpu_count
+if(n_cpu==0):n_cpu=cpu_count()
+if(is_half==True):
+ #6G显存配置
+ x_pad = 3
+ x_query = 10
+ x_center = 60
+ x_max = 65
+else:
+ #5G显存配置
+ x_pad = 1
+ # x_query = 6
+ # x_center = 30
+ # x_max = 32
+ #6G显存配置
+ x_query = 6
+ x_center = 38
+ x_max = 41
\ No newline at end of file
diff --git a/go-web.bat b/go-web.bat
new file mode 100644
index 0000000000000000000000000000000000000000..e150a778ac9a47315b34da8198ee5b388362e11e
--- /dev/null
+++ b/go-web.bat
@@ -0,0 +1 @@
+runtime\python.exe infer-web.py
\ No newline at end of file
diff --git a/go.bat b/go.bat
new file mode 100644
index 0000000000000000000000000000000000000000..f7ce87ada2ef0439bf36e258de80f5f3344d6498
--- /dev/null
+++ b/go.bat
@@ -0,0 +1 @@
+runtime\python.exe infer.py
\ No newline at end of file
diff --git a/hubert_base.pt b/hubert_base.pt
new file mode 100644
index 0000000000000000000000000000000000000000..72f47ab58564f01d5cc8b05c63bdf96d944551ff
--- /dev/null
+++ b/hubert_base.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f54b40fd2802423a5643779c4861af1e9ee9c1564dc9d32f54f20b5ffba7db96
+size 189507909
diff --git a/infer-web.py b/infer-web.py
new file mode 100644
index 0000000000000000000000000000000000000000..0852fdfcab55ba452fcd8a5986f3bd1870b338bf
--- /dev/null
+++ b/infer-web.py
@@ -0,0 +1,193 @@
+import torch, pdb, os,traceback,sys,warnings,shutil
+now_dir=os.getcwd()
+sys.path.append(now_dir)
+tmp=os.path.join(now_dir,"TEMP")
+shutil.rmtree(tmp,ignore_errors=True)
+os.makedirs(tmp,exist_ok=True)
+os.environ["TEMP"]=tmp
+warnings.filterwarnings("ignore")
+torch.manual_seed(114514)
+from infer_pack.models import SynthesizerTrnMs256NSF as SynthesizerTrn256
+from scipy.io import wavfile
+from fairseq import checkpoint_utils
+import gradio as gr
+import librosa
+import logging
+from vc_infer_pipeline import VC
+import soundfile as sf
+from config import is_half,device,is_half
+from infer_uvr5 import _audio_pre_
+logging.getLogger('numba').setLevel(logging.WARNING)
+
+models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task(["hubert_base.pt"],suffix="",)
+hubert_model = models[0]
+hubert_model = hubert_model.to(device)
+if(is_half):hubert_model = hubert_model.half()
+else:hubert_model = hubert_model.float()
+hubert_model.eval()
+
+
+weight_root="weights"
+weight_uvr5_root="uvr5_weights"
+names=[]
+for name in os.listdir(weight_root):names.append(name.replace(".pt",""))
+uvr5_names=[]
+for name in os.listdir(weight_uvr5_root):uvr5_names.append(name.replace(".pth",""))
+
+def get_vc(sid):
+ person = "%s/%s.pt" % (weight_root, sid)
+ cpt = torch.load(person, map_location="cpu")
+ dv = cpt["dv"]
+ tgt_sr = cpt["config"][-1]
+ net_g = SynthesizerTrn256(*cpt["config"], is_half=is_half)
+ net_g.load_state_dict(cpt["weight"], strict=True)
+ net_g.eval().to(device)
+ if (is_half):net_g = net_g.half()
+ else:net_g = net_g.float()
+ vc = VC(tgt_sr, device, is_half)
+ return dv,tgt_sr,net_g,vc
+
+def vc_single(sid,input_audio,f0_up_key,f0_file):
+ if input_audio is None:return "You need to upload an audio", None
+ f0_up_key = int(f0_up_key)
+ try:
+ if(type(input_audio)==str):
+ print("processing %s" % input_audio)
+ audio, sampling_rate = sf.read(input_audio)
+ else:
+ sampling_rate, audio = input_audio
+ audio = audio.astype("float32") / 32768
+ if(type(sid)==str):dv, tgt_sr, net_g, vc=get_vc(sid)
+ else:dv,tgt_sr,net_g,vc=sid
+ if len(audio.shape) > 1:
+ audio = librosa.to_mono(audio.transpose(1, 0))
+ if sampling_rate != 16000:
+ audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)
+ times = [0, 0, 0]
+ audio_opt=vc.pipeline(hubert_model,net_g,dv,audio,times,f0_up_key,f0_file=f0_file)
+ print(times)
+ return "Success", (tgt_sr, audio_opt)
+ except:
+ info=traceback.format_exc()
+ print(info)
+ return info,(None,None)
+ finally:
+ print("clean_empty_cache")
+ del net_g,dv,vc
+ torch.cuda.empty_cache()
+
+def vc_multi(sid,dir_path,opt_root,paths,f0_up_key):
+ try:
+ dir_path=dir_path.strip(" ")#防止小白拷路径头尾带了空格
+ opt_root=opt_root.strip(" ")
+ os.makedirs(opt_root, exist_ok=True)
+ dv, tgt_sr, net_g, vc = get_vc(sid)
+ try:
+ if(dir_path!=""):paths=[os.path.join(dir_path,name)for name in os.listdir(dir_path)]
+ else:paths=[path.name for path in paths]
+ except:
+ traceback.print_exc()
+ paths = [path.name for path in paths]
+ infos=[]
+ for path in paths:
+ info,opt=vc_single([dv,tgt_sr,net_g,vc],path,f0_up_key,f0_file=None)
+ if(info=="Success"):
+ try:
+ tgt_sr,audio_opt=opt
+ wavfile.write("%s/%s" % (opt_root, os.path.basename(path)), tgt_sr, audio_opt)
+ except:
+ info=traceback.format_exc()
+ infos.append("%s->%s"%(os.path.basename(path),info))
+ return "\n".join(infos)
+ except:
+ return traceback.format_exc()
+ finally:
+ print("clean_empty_cache")
+ del net_g,dv,vc
+ torch.cuda.empty_cache()
+
+def uvr(model_name,inp_root,save_root_vocal,save_root_ins):
+ infos = []
+ try:
+ inp_root = inp_root.strip(" ")# 防止小白拷路径头尾带了空格
+ save_root_vocal = save_root_vocal.strip(" ")
+ save_root_ins = save_root_ins.strip(" ")
+ pre_fun = _audio_pre_(model_path=os.path.join(weight_uvr5_root,model_name+".pth"), device=device, is_half=is_half)
+ for name in os.listdir(inp_root):
+ inp_path=os.path.join(inp_root,name)
+ try:
+ pre_fun._path_audio_(inp_path , save_root_ins,save_root_vocal)
+ infos.append("%s->Success"%(os.path.basename(inp_path)))
+ except:
+ infos.append("%s->%s" % (os.path.basename(inp_path),traceback.format_exc()))
+ except:
+ infos.append(traceback.format_exc())
+ finally:
+ try:
+ del pre_fun.model
+ del pre_fun
+ except:
+ traceback.print_exc()
+ print("clean_empty_cache")
+ torch.cuda.empty_cache()
+ return "\n".join(infos)
+
+with gr.Blocks() as app:
+ with gr.Tabs():
+ with gr.TabItem("推理"):
+ with gr.Group():
+ gr.Markdown(value="""
+ 使用软件者、传播软件导出的声音者自负全责。如不认可该条款，则不能使用/引用软件包内所有代码和文件。<br>
+ 目前仅开放白菜音色，后续将扩展为本地训练推理工具，用户可训练自己的音色进行社区共享。<br>
+ 男转女推荐+12key，女转男推荐-12key，如果音域爆炸导致音色失真也可以自己调整到合适音域
+ """)
+ with gr.Row():
+ with gr.Column():
+ sid0 = gr.Dropdown(label="音色", choices=names)
+ vc_transform0 = gr.Number(label="变调（整数，半音数量，升八度12降八度-12）", value=12)
+ f0_file = gr.File(label="F0曲线文件，可选，一行一个音高，代替默认F0及升降调")
+ input_audio0 = gr.Audio(label="上传音频")
+ but0=gr.Button("转换", variant="primary")
+ with gr.Column():
+ vc_output1 = gr.Textbox(label="输出信息")
+ vc_output2 = gr.Audio(label="输出音频")
+ but0.click(vc_single, [sid0, input_audio0, vc_transform0,f0_file], [vc_output1, vc_output2])
+ with gr.Group():
+ gr.Markdown(value="""
+ 批量转换，上传多个音频文件，在指定文件夹（默认opt）下输出转换的音频。<br>
+ 合格的文件夹路径格式举例：E:\codes\py39\\vits_vc_gpu\白鹭霜华测试样例（去文件管理器地址栏拷就行了）
+ """)
+ with gr.Row():
+ with gr.Column():
+ sid1 = gr.Dropdown(label="音色", choices=names)
+ vc_transform1 = gr.Number(label="变调（整数，半音数量，升八度12降八度-12）", value=12)
+ opt_input = gr.Textbox(label="指定输出文件夹",value="opt")
+ with gr.Column():
+ dir_input = gr.Textbox(label="输入待处理音频文件夹路径")
+ inputs = gr.File(file_count="multiple", label="也可批量输入音频文件，二选一，优先读文件夹")
+ but1=gr.Button("转换", variant="primary")
+ vc_output3 = gr.Textbox(label="输出信息")
+ but1.click(vc_multi, [sid1, dir_input,opt_input,inputs, vc_transform1], [vc_output3])
+
+ with gr.TabItem("数据处理"):
+ with gr.Group():
+ gr.Markdown(value="""
+ 人声伴奏分离批量处理，使用UVR5模型。<br>
+ 不带和声用HP2，带和声且提取的人声不需要和声用HP5<br>
+ 合格的文件夹路径格式举例：E:\codes\py39\\vits_vc_gpu\白鹭霜华测试样例（去文件管理器地址栏拷就行了）
+ """)
+ with gr.Row():
+ with gr.Column():
+ dir_wav_input = gr.Textbox(label="输入待处理音频文件夹路径")
+ wav_inputs = gr.File(file_count="multiple", label="也可批量输入音频文件，二选一，优先读文件夹")
+ with gr.Column():
+ model_choose = gr.Dropdown(label="模型", choices=uvr5_names)
+ opt_vocal_root = gr.Textbox(label="指定输出人声文件夹",value="opt")
+ opt_ins_root = gr.Textbox(label="指定输出乐器文件夹",value="opt")
+ but2=gr.Button("转换", variant="primary")
+ vc_output4 = gr.Textbox(label="输出信息")
+ but2.click(uvr, [model_choose, dir_wav_input,opt_vocal_root,opt_ins_root], [vc_output4])
+ with gr.TabItem("训练-待开放"):pass
+
+ # app.launch(server_name="0.0.0.0",server_port=7860)
+ app.launch(server_name="127.0.0.1",server_port=7860)
\ No newline at end of file
diff --git a/infer.py b/infer.py
new file mode 100644
index 0000000000000000000000000000000000000000..4bd61ad0227aa2fabb798dd36ea5a21957e5af3e
--- /dev/null
+++ b/infer.py
@@ -0,0 +1,48 @@
+import torch, pdb, os,sys,librosa,warnings,traceback
+warnings.filterwarnings("ignore")
+torch.manual_seed(114514)
+sys.path.append(os.getcwd())
+from config import inp_root,opt_root,f0_up_key,person,is_half,device
+os.makedirs(opt_root,exist_ok=True)
+import soundfile as sf
+from infer_pack.models import SynthesizerTrnMs256NSF as SynthesizerTrn256
+from scipy.io import wavfile
+from fairseq import checkpoint_utils
+import scipy.signal as signal
+from vc_infer_pipeline import VC
+
+models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task(["hubert_base.pt"],suffix="",)
+model = models[0]
+model = model.to(device)
+if(is_half):model = model.half()
+else:model = model.float()
+model.eval()
+
+cpt=torch.load(person,map_location="cpu")
+dv=cpt["dv"]
+tgt_sr=cpt["config"][-1]
+net_g = SynthesizerTrn256(*cpt["config"],is_half=is_half)
+net_g.load_state_dict(cpt["weight"],strict=True)
+net_g.eval().to(device)
+if(is_half):net_g = net_g.half()
+else:net_g = net_g.float()
+
+vc=VC(tgt_sr,device,is_half)
+
+for name in os.listdir(inp_root):
+ try:
+ wav_path="%s\%s"%(inp_root,name)
+ print("processing %s"%wav_path)
+ audio, sampling_rate = sf.read(wav_path)
+ if len(audio.shape) > 1:
+ audio = librosa.to_mono(audio.transpose(1, 0))
+ if sampling_rate != vc.sr:
+ audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=vc.sr)
+
+ times = [0, 0, 0]
+ audio_opt=vc.pipeline(model,net_g,dv,audio,times,f0_up_key,f0_file=None)
+ wavfile.write("%s/%s"%(opt_root,name), tgt_sr, audio_opt)
+ except:
+ traceback.print_exc()
+
+print(times)
diff --git a/infer_pack/__pycache__/attentions.cpython-39.pyc b/infer_pack/__pycache__/attentions.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6ab2db43cbc0901bb05e0b90f35f75e90f232a54
Binary files /dev/null and b/infer_pack/__pycache__/attentions.cpython-39.pyc differ
diff --git a/infer_pack/__pycache__/commons.cpython-39.pyc b/infer_pack/__pycache__/commons.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8f504828a4fc7914bd7cbfaa7e4eea991dfbe564
Binary files /dev/null and b/infer_pack/__pycache__/commons.cpython-39.pyc differ
diff --git a/infer_pack/__pycache__/models.cpython-39.pyc b/infer_pack/__pycache__/models.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..930c18d0b8ea25f4de6d0fef6b2ecb4179b1d0d8
Binary files /dev/null and b/infer_pack/__pycache__/models.cpython-39.pyc differ
diff --git a/infer_pack/__pycache__/modules.cpython-39.pyc b/infer_pack/__pycache__/modules.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..acc721e102cf83f070434812ee90f7c378c0cb43
Binary files /dev/null and b/infer_pack/__pycache__/modules.cpython-39.pyc differ
diff --git a/infer_pack/__pycache__/transforms.cpython-39.pyc b/infer_pack/__pycache__/transforms.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c3ba6c2b8ab455fe048fea17582dada45b96f5bc
Binary files /dev/null and b/infer_pack/__pycache__/transforms.cpython-39.pyc differ
diff --git a/infer_pack/attentions.py b/infer_pack/attentions.py
new file mode 100644
index 0000000000000000000000000000000000000000..77cb63ffccf3e33badf22d50862a64ba517b487f
--- /dev/null
+++ b/infer_pack/attentions.py
@@ -0,0 +1,417 @@
+import copy
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+from infer_pack import commons
+from infer_pack import modules
+from infer_pack.modules import LayerNorm
+
+
+class Encoder(nn.Module):
+ def __init__(
+ self,
+ hidden_channels,
+ filter_channels,
+ n_heads,
+ n_layers,
+ kernel_size=1,
+ p_dropout=0.0,
+ window_size=10,
+ **kwargs
+ ):
+ super().__init__()
+ self.hidden_channels = hidden_channels
+ self.filter_channels = filter_channels
+ self.n_heads = n_heads
+ self.n_layers = n_layers
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ self.window_size = window_size
+
+ self.drop = nn.Dropout(p_dropout)
+ self.attn_layers = nn.ModuleList()
+ self.norm_layers_1 = nn.ModuleList()
+ self.ffn_layers = nn.ModuleList()
+ self.norm_layers_2 = nn.ModuleList()
+ for i in range(self.n_layers):
+ self.attn_layers.append(
+ MultiHeadAttention(
+ hidden_channels,
+ hidden_channels,
+ n_heads,
+ p_dropout=p_dropout,
+ window_size=window_size,
+ )
+ )
+ self.norm_layers_1.append(LayerNorm(hidden_channels))
+ self.ffn_layers.append(
+ FFN(
+ hidden_channels,
+ hidden_channels,
+ filter_channels,
+ kernel_size,
+ p_dropout=p_dropout,
+ )
+ )
+ self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+ def forward(self, x, x_mask):
+ attn_mask = x_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+ x = x * x_mask
+ for i in range(self.n_layers):
+ y = self.attn_layers[i](x, x, attn_mask)
+ y = self.drop(y)
+ x = self.norm_layers_1[i](x + y)
+
+ y = self.ffn_layers[i](x, x_mask)
+ y = self.drop(y)
+ x = self.norm_layers_2[i](x + y)
+ x = x * x_mask
+ return x
+
+
+class Decoder(nn.Module):
+ def __init__(
+ self,
+ hidden_channels,
+ filter_channels,
+ n_heads,
+ n_layers,
+ kernel_size=1,
+ p_dropout=0.0,
+ proximal_bias=False,
+ proximal_init=True,
+ **kwargs
+ ):
+ super().__init__()
+ self.hidden_channels = hidden_channels
+ self.filter_channels = filter_channels
+ self.n_heads = n_heads
+ self.n_layers = n_layers
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ self.proximal_bias = proximal_bias
+ self.proximal_init = proximal_init
+
+ self.drop = nn.Dropout(p_dropout)
+ self.self_attn_layers = nn.ModuleList()
+ self.norm_layers_0 = nn.ModuleList()
+ self.encdec_attn_layers = nn.ModuleList()
+ self.norm_layers_1 = nn.ModuleList()
+ self.ffn_layers = nn.ModuleList()
+ self.norm_layers_2 = nn.ModuleList()
+ for i in range(self.n_layers):
+ self.self_attn_layers.append(
+ MultiHeadAttention(
+ hidden_channels,
+ hidden_channels,
+ n_heads,
+ p_dropout=p_dropout,
+ proximal_bias=proximal_bias,
+ proximal_init=proximal_init,
+ )
+ )
+ self.norm_layers_0.append(LayerNorm(hidden_channels))
+ self.encdec_attn_layers.append(
+ MultiHeadAttention(
+ hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout
+ )
+ )
+ self.norm_layers_1.append(LayerNorm(hidden_channels))
+ self.ffn_layers.append(
+ FFN(
+ hidden_channels,
+ hidden_channels,
+ filter_channels,
+ kernel_size,
+ p_dropout=p_dropout,
+ causal=True,
+ )
+ )
+ self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+ def forward(self, x, x_mask, h, h_mask):
+ """
+ x: decoder input
+ h: encoder output
+ """
+ self_attn_mask = commons.subsequent_mask(x_mask.size(2)).to(
+ device=x.device, dtype=x.dtype
+ )
+ encdec_attn_mask = h_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+ x = x * x_mask
+ for i in range(self.n_layers):
+ y = self.self_attn_layers[i](x, x, self_attn_mask)
+ y = self.drop(y)
+ x = self.norm_layers_0[i](x + y)
+
+ y = self.encdec_attn_layers[i](x, h, encdec_attn_mask)
+ y = self.drop(y)
+ x = self.norm_layers_1[i](x + y)
+
+ y = self.ffn_layers[i](x, x_mask)
+ y = self.drop(y)
+ x = self.norm_layers_2[i](x + y)
+ x = x * x_mask
+ return x
+
+
+class MultiHeadAttention(nn.Module):
+ def __init__(
+ self,
+ channels,
+ out_channels,
+ n_heads,
+ p_dropout=0.0,
+ window_size=None,
+ heads_share=True,
+ block_length=None,
+ proximal_bias=False,
+ proximal_init=False,
+ ):
+ super().__init__()
+ assert channels % n_heads == 0
+
+ self.channels = channels
+ self.out_channels = out_channels
+ self.n_heads = n_heads
+ self.p_dropout = p_dropout
+ self.window_size = window_size
+ self.heads_share = heads_share
+ self.block_length = block_length
+ self.proximal_bias = proximal_bias
+ self.proximal_init = proximal_init
+ self.attn = None
+
+ self.k_channels = channels // n_heads
+ self.conv_q = nn.Conv1d(channels, channels, 1)
+ self.conv_k = nn.Conv1d(channels, channels, 1)
+ self.conv_v = nn.Conv1d(channels, channels, 1)
+ self.conv_o = nn.Conv1d(channels, out_channels, 1)
+ self.drop = nn.Dropout(p_dropout)
+
+ if window_size is not None:
+ n_heads_rel = 1 if heads_share else n_heads
+ rel_stddev = self.k_channels**-0.5
+ self.emb_rel_k = nn.Parameter(
+ torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels)
+ * rel_stddev
+ )
+ self.emb_rel_v = nn.Parameter(
+ torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels)
+ * rel_stddev
+ )
+
+ nn.init.xavier_uniform_(self.conv_q.weight)
+ nn.init.xavier_uniform_(self.conv_k.weight)
+ nn.init.xavier_uniform_(self.conv_v.weight)
+ if proximal_init:
+ with torch.no_grad():
+ self.conv_k.weight.copy_(self.conv_q.weight)
+ self.conv_k.bias.copy_(self.conv_q.bias)
+
+ def forward(self, x, c, attn_mask=None):
+ q = self.conv_q(x)
+ k = self.conv_k(c)
+ v = self.conv_v(c)
+
+ x, self.attn = self.attention(q, k, v, mask=attn_mask)
+
+ x = self.conv_o(x)
+ return x
+
+ def attention(self, query, key, value, mask=None):
+ # reshape [b, d, t] -> [b, n_h, t, d_k]
+ b, d, t_s, t_t = (*key.size(), query.size(2))
+ query = query.view(b, self.n_heads, self.k_channels, t_t).transpose(2, 3)
+ key = key.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+ value = value.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+
+ scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1))
+ if self.window_size is not None:
+ assert (
+ t_s == t_t
+ ), "Relative attention is only available for self-attention."
+ key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s)
+ rel_logits = self._matmul_with_relative_keys(
+ query / math.sqrt(self.k_channels), key_relative_embeddings
+ )
+ scores_local = self._relative_position_to_absolute_position(rel_logits)
+ scores = scores + scores_local
+ if self.proximal_bias:
+ assert t_s == t_t, "Proximal bias is only available for self-attention."
+ scores = scores + self._attention_bias_proximal(t_s).to(
+ device=scores.device, dtype=scores.dtype
+ )
+ if mask is not None:
+ scores = scores.masked_fill(mask == 0, -1e4)
+ if self.block_length is not None:
+ assert (
+ t_s == t_t
+ ), "Local attention is only available for self-attention."
+ block_mask = (
+ torch.ones_like(scores)
+ .triu(-self.block_length)
+ .tril(self.block_length)
+ )
+ scores = scores.masked_fill(block_mask == 0, -1e4)
+ p_attn = F.softmax(scores, dim=-1) # [b, n_h, t_t, t_s]
+ p_attn = self.drop(p_attn)
+ output = torch.matmul(p_attn, value)
+ if self.window_size is not None:
+ relative_weights = self._absolute_position_to_relative_position(p_attn)
+ value_relative_embeddings = self._get_relative_embeddings(
+ self.emb_rel_v, t_s
+ )
+ output = output + self._matmul_with_relative_values(
+ relative_weights, value_relative_embeddings
+ )
+ output = (
+ output.transpose(2, 3).contiguous().view(b, d, t_t)
+ ) # [b, n_h, t_t, d_k] -> [b, d, t_t]
+ return output, p_attn
+
+ def _matmul_with_relative_values(self, x, y):
+ """
+ x: [b, h, l, m]
+ y: [h or 1, m, d]
+ ret: [b, h, l, d]
+ """
+ ret = torch.matmul(x, y.unsqueeze(0))
+ return ret
+
+ def _matmul_with_relative_keys(self, x, y):
+ """
+ x: [b, h, l, d]
+ y: [h or 1, m, d]
+ ret: [b, h, l, m]
+ """
+ ret = torch.matmul(x, y.unsqueeze(0).transpose(-2, -1))
+ return ret
+
+ def _get_relative_embeddings(self, relative_embeddings, length):
+ max_relative_position = 2 * self.window_size + 1
+ # Pad first before slice to avoid using cond ops.
+ pad_length = max(length - (self.window_size + 1), 0)
+ slice_start_position = max((self.window_size + 1) - length, 0)
+ slice_end_position = slice_start_position + 2 * length - 1
+ if pad_length > 0:
+ padded_relative_embeddings = F.pad(
+ relative_embeddings,
+ commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]),
+ )
+ else:
+ padded_relative_embeddings = relative_embeddings
+ used_relative_embeddings = padded_relative_embeddings[
+ :, slice_start_position:slice_end_position
+ ]
+ return used_relative_embeddings
+
+ def _relative_position_to_absolute_position(self, x):
+ """
+ x: [b, h, l, 2*l-1]
+ ret: [b, h, l, l]
+ """
+ batch, heads, length, _ = x.size()
+ # Concat columns of pad to shift from relative to absolute indexing.
+ x = F.pad(x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, 1]]))
+
+ # Concat extra elements so to add up to shape (len+1, 2*len-1).
+ x_flat = x.view([batch, heads, length * 2 * length])
+ x_flat = F.pad(
+ x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [0, length - 1]])
+ )
+
+ # Reshape and slice out the padded elements.
+ x_final = x_flat.view([batch, heads, length + 1, 2 * length - 1])[
+ :, :, :length, length - 1 :
+ ]
+ return x_final
+
+ def _absolute_position_to_relative_position(self, x):
+ """
+ x: [b, h, l, l]
+ ret: [b, h, l, 2*l-1]
+ """
+ batch, heads, length, _ = x.size()
+ # padd along column
+ x = F.pad(
+ x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, length - 1]])
+ )
+ x_flat = x.view([batch, heads, length**2 + length * (length - 1)])
+ # add 0's in the beginning that will skew the elements after reshape
+ x_flat = F.pad(x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [length, 0]]))
+ x_final = x_flat.view([batch, heads, length, 2 * length])[:, :, :, 1:]
+ return x_final
+
+ def _attention_bias_proximal(self, length):
+ """Bias for self-attention to encourage attention to close positions.
+ Args:
+ length: an integer scalar.
+ Returns:
+ a Tensor with shape [1, 1, length, length]
+ """
+ r = torch.arange(length, dtype=torch.float32)
+ diff = torch.unsqueeze(r, 0) - torch.unsqueeze(r, 1)
+ return torch.unsqueeze(torch.unsqueeze(-torch.log1p(torch.abs(diff)), 0), 0)
+
+
+class FFN(nn.Module):
+ def __init__(
+ self,
+ in_channels,
+ out_channels,
+ filter_channels,
+ kernel_size,
+ p_dropout=0.0,
+ activation=None,
+ causal=False,
+ ):
+ super().__init__()
+ self.in_channels = in_channels
+ self.out_channels = out_channels
+ self.filter_channels = filter_channels
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ self.activation = activation
+ self.causal = causal
+
+ if causal:
+ self.padding = self._causal_padding
+ else:
+ self.padding = self._same_padding
+
+ self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size)
+ self.conv_2 = nn.Conv1d(filter_channels, out_channels, kernel_size)
+ self.drop = nn.Dropout(p_dropout)
+
+ def forward(self, x, x_mask):
+ x = self.conv_1(self.padding(x * x_mask))
+ if self.activation == "gelu":
+ x = x * torch.sigmoid(1.702 * x)
+ else:
+ x = torch.relu(x)
+ x = self.drop(x)
+ x = self.conv_2(self.padding(x * x_mask))
+ return x * x_mask
+
+ def _causal_padding(self, x):
+ if self.kernel_size == 1:
+ return x
+ pad_l = self.kernel_size - 1
+ pad_r = 0
+ padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+ x = F.pad(x, commons.convert_pad_shape(padding))
+ return x
+
+ def _same_padding(self, x):
+ if self.kernel_size == 1:
+ return x
+ pad_l = (self.kernel_size - 1) // 2
+ pad_r = self.kernel_size // 2
+ padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+ x = F.pad(x, commons.convert_pad_shape(padding))
+ return x
diff --git a/infer_pack/commons.py b/infer_pack/commons.py
new file mode 100644
index 0000000000000000000000000000000000000000..ba2dad2c884a34d3ffcf6e0795d04d764d6a5eec
--- /dev/null
+++ b/infer_pack/commons.py
@@ -0,0 +1,164 @@
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+
+def init_weights(m, mean=0.0, std=0.01):
+ classname = m.__class__.__name__
+ if classname.find("Conv") != -1:
+ m.weight.data.normal_(mean, std)
+
+
+def get_padding(kernel_size, dilation=1):
+ return int((kernel_size * dilation - dilation) / 2)
+
+
+def convert_pad_shape(pad_shape):
+ l = pad_shape[::-1]
+ pad_shape = [item for sublist in l for item in sublist]
+ return pad_shape
+
+
+def kl_divergence(m_p, logs_p, m_q, logs_q):
+ """KL(P||Q)"""
+ kl = (logs_q - logs_p) - 0.5
+ kl += (
+ 0.5 * (torch.exp(2.0 * logs_p) + ((m_p - m_q) ** 2)) * torch.exp(-2.0 * logs_q)
+ )
+ return kl
+
+
+def rand_gumbel(shape):
+ """Sample from the Gumbel distribution, protect from overflows."""
+ uniform_samples = torch.rand(shape) * 0.99998 + 0.00001
+ return -torch.log(-torch.log(uniform_samples))
+
+
+def rand_gumbel_like(x):
+ g = rand_gumbel(x.size()).to(dtype=x.dtype, device=x.device)
+ return g
+
+
+def slice_segments(x, ids_str, segment_size=4):
+ ret = torch.zeros_like(x[:, :, :segment_size])
+ for i in range(x.size(0)):
+ idx_str = ids_str[i]
+ idx_end = idx_str + segment_size
+ ret[i] = x[i, :, idx_str:idx_end]
+ return ret
+def slice_segments2(x, ids_str, segment_size=4):
+ ret = torch.zeros_like(x[:, :segment_size])
+ for i in range(x.size(0)):
+ idx_str = ids_str[i]
+ idx_end = idx_str + segment_size
+ ret[i] = x[i, idx_str:idx_end]
+ return ret
+
+
+def rand_slice_segments(x, x_lengths=None, segment_size=4):
+ b, d, t = x.size()
+ if x_lengths is None:
+ x_lengths = t
+ ids_str_max = x_lengths - segment_size + 1
+ ids_str = (torch.rand([b]).to(device=x.device) * ids_str_max).to(dtype=torch.long)
+ ret = slice_segments(x, ids_str, segment_size)
+ return ret, ids_str
+
+
+def get_timing_signal_1d(length, channels, min_timescale=1.0, max_timescale=1.0e4):
+ position = torch.arange(length, dtype=torch.float)
+ num_timescales = channels // 2
+ log_timescale_increment = math.log(float(max_timescale) / float(min_timescale)) / (
+ num_timescales - 1
+ )
+ inv_timescales = min_timescale * torch.exp(
+ torch.arange(num_timescales, dtype=torch.float) * -log_timescale_increment
+ )
+ scaled_time = position.unsqueeze(0) * inv_timescales.unsqueeze(1)
+ signal = torch.cat([torch.sin(scaled_time), torch.cos(scaled_time)], 0)
+ signal = F.pad(signal, [0, 0, 0, channels % 2])
+ signal = signal.view(1, channels, length)
+ return signal
+
+
+def add_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4):
+ b, channels, length = x.size()
+ signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+ return x + signal.to(dtype=x.dtype, device=x.device)
+
+
+def cat_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4, axis=1):
+ b, channels, length = x.size()
+ signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+ return torch.cat([x, signal.to(dtype=x.dtype, device=x.device)], axis)
+
+
+def subsequent_mask(length):
+ mask = torch.tril(torch.ones(length, length)).unsqueeze(0).unsqueeze(0)
+ return mask
+
+
+@torch.jit.script
+def fused_add_tanh_sigmoid_multiply(input_a, input_b, n_channels):
+ n_channels_int = n_channels[0]
+ in_act = input_a + input_b
+ t_act = torch.tanh(in_act[:, :n_channels_int, :])
+ s_act = torch.sigmoid(in_act[:, n_channels_int:, :])
+ acts = t_act * s_act
+ return acts
+
+
+def convert_pad_shape(pad_shape):
+ l = pad_shape[::-1]
+ pad_shape = [item for sublist in l for item in sublist]
+ return pad_shape
+
+
+def shift_1d(x):
+ x = F.pad(x, convert_pad_shape([[0, 0], [0, 0], [1, 0]]))[:, :, :-1]
+ return x
+
+
+def sequence_mask(length, max_length=None):
+ if max_length is None:
+ max_length = length.max()
+ x = torch.arange(max_length, dtype=length.dtype, device=length.device)
+ return x.unsqueeze(0) < length.unsqueeze(1)
+
+
+def generate_path(duration, mask):
+ """
+ duration: [b, 1, t_x]
+ mask: [b, 1, t_y, t_x]
+ """
+ device = duration.device
+
+ b, _, t_y, t_x = mask.shape
+ cum_duration = torch.cumsum(duration, -1)
+
+ cum_duration_flat = cum_duration.view(b * t_x)
+ path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype)
+ path = path.view(b, t_x, t_y)
+ path = path - F.pad(path, convert_pad_shape([[0, 0], [1, 0], [0, 0]]))[:, :-1]
+ path = path.unsqueeze(1).transpose(2, 3) * mask
+ return path
+
+
+def clip_grad_value_(parameters, clip_value, norm_type=2):
+ if isinstance(parameters, torch.Tensor):
+ parameters = [parameters]
+ parameters = list(filter(lambda p: p.grad is not None, parameters))
+ norm_type = float(norm_type)
+ if clip_value is not None:
+ clip_value = float(clip_value)
+
+ total_norm = 0
+ for p in parameters:
+ param_norm = p.grad.data.norm(norm_type)
+ total_norm += param_norm.item() ** norm_type
+ if clip_value is not None:
+ p.grad.data.clamp_(min=-clip_value, max=clip_value)
+ total_norm = total_norm ** (1.0 / norm_type)
+ return total_norm
diff --git a/infer_pack/models.py b/infer_pack/models.py
new file mode 100644
index 0000000000000000000000000000000000000000..7ffa50643b51dc50289e34f58e30c41e19746102
--- /dev/null
+++ b/infer_pack/models.py
@@ -0,0 +1,664 @@
+import math,pdb,os
+from time import time as ttime
+import torch
+from torch import nn
+from torch.nn import functional as F
+from infer_pack import modules
+from infer_pack import attentions
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from infer_pack.commons import init_weights
+import numpy as np
+from infer_pack import commons
+class TextEncoder256(nn.Module):
+ def __init__(
+ self, out_channels, hidden_channels, filter_channels, n_heads, n_layers, kernel_size, p_dropout, f0=True ):
+ super().__init__()
+ self.out_channels = out_channels
+ self.hidden_channels = hidden_channels
+ self.filter_channels = filter_channels
+ self.n_heads = n_heads
+ self.n_layers = n_layers
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ self.emb_phone = nn.Linear(256, hidden_channels)
+ self.lrelu=nn.LeakyReLU(0.1,inplace=True)
+ if(f0==True):
+ self.emb_pitch = nn.Embedding(256, hidden_channels) # pitch 256
+ self.encoder = attentions.Encoder(
+ hidden_channels, filter_channels, n_heads, n_layers, kernel_size, p_dropout
+ )
+ self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+ def forward(self, phone, pitch, lengths):
+ if(pitch==None):
+ x = self.emb_phone(phone)
+ else:
+ x = self.emb_phone(phone) + self.emb_pitch(pitch)
+ x = x * math.sqrt(self.hidden_channels) # [b, t, h]
+ x=self.lrelu(x)
+ x = torch.transpose(x, 1, -1) # [b, h, t]
+ x_mask = torch.unsqueeze(commons.sequence_mask(lengths, x.size(2)), 1).to(
+ x.dtype
+ )
+ x = self.encoder(x * x_mask, x_mask)
+ stats = self.proj(x) * x_mask
+
+ m, logs = torch.split(stats, self.out_channels, dim=1)
+ return m, logs, x_mask
+class TextEncoder256km(nn.Module):
+ def __init__(
+ self, out_channels, hidden_channels, filter_channels, n_heads, n_layers, kernel_size, p_dropout, f0=True ):
+ super().__init__()
+ self.out_channels = out_channels
+ self.hidden_channels = hidden_channels
+ self.filter_channels = filter_channels
+ self.n_heads = n_heads
+ self.n_layers = n_layers
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ # self.emb_phone = nn.Linear(256, hidden_channels)
+ self.emb_phone = nn.Embedding(500, hidden_channels)
+ self.lrelu=nn.LeakyReLU(0.1,inplace=True)
+ if(f0==True):
+ self.emb_pitch = nn.Embedding(256, hidden_channels) # pitch 256
+ self.encoder = attentions.Encoder(
+ hidden_channels, filter_channels, n_heads, n_layers, kernel_size, p_dropout
+ )
+ self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+ def forward(self, phone, pitch, lengths):
+ if(pitch==None):
+ x = self.emb_phone(phone)
+ else:
+ x = self.emb_phone(phone) + self.emb_pitch(pitch)
+ x = x * math.sqrt(self.hidden_channels) # [b, t, h]
+ x=self.lrelu(x)
+ x = torch.transpose(x, 1, -1) # [b, h, t]
+ x_mask = torch.unsqueeze(commons.sequence_mask(lengths, x.size(2)), 1).to(
+ x.dtype
+ )
+ x = self.encoder(x * x_mask, x_mask)
+ stats = self.proj(x) * x_mask
+
+ m, logs = torch.split(stats, self.out_channels, dim=1)
+ return m, logs, x_mask
+class ResidualCouplingBlock(nn.Module):
+ def __init__(
+ self,
+ channels,
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ n_flows=4,
+ gin_channels=0,
+ ):
+ super().__init__()
+ self.channels = channels
+ self.hidden_channels = hidden_channels
+ self.kernel_size = kernel_size
+ self.dilation_rate = dilation_rate
+ self.n_layers = n_layers
+ self.n_flows = n_flows
+ self.gin_channels = gin_channels
+
+ self.flows = nn.ModuleList()
+ for i in range(n_flows):
+ self.flows.append(
+ modules.ResidualCouplingLayer(
+ channels,
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ gin_channels=gin_channels,
+ mean_only=True,
+ )
+ )
+ self.flows.append(modules.Flip())
+
+ def forward(self, x, x_mask, g=None, reverse=False):
+ if not reverse:
+ for flow in self.flows:
+ x, _ = flow(x, x_mask, g=g, reverse=reverse)
+ else:
+ for flow in reversed(self.flows):
+ x = flow(x, x_mask, g=g, reverse=reverse)
+ return x
+
+ def remove_weight_norm(self):
+ for i in range(self.n_flows):
+ self.flows[i * 2].remove_weight_norm()
+class PosteriorEncoder(nn.Module):
+ def __init__(
+ self,
+ in_channels,
+ out_channels,
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ gin_channels=0,
+ ):
+ super().__init__()
+ self.in_channels = in_channels
+ self.out_channels = out_channels
+ self.hidden_channels = hidden_channels
+ self.kernel_size = kernel_size
+ self.dilation_rate = dilation_rate
+ self.n_layers = n_layers
+ self.gin_channels = gin_channels
+
+ self.pre = nn.Conv1d(in_channels, hidden_channels, 1)
+ self.enc = modules.WN(
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ gin_channels=gin_channels,
+ )
+ self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+ def forward(self, x, x_lengths, g=None):
+ x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(
+ x.dtype
+ )
+ x = self.pre(x) * x_mask
+ x = self.enc(x, x_mask, g=g)
+ stats = self.proj(x) * x_mask
+ m, logs = torch.split(stats, self.out_channels, dim=1)
+ z = (m + torch.randn_like(m) * torch.exp(logs)) * x_mask
+ return z, m, logs, x_mask
+
+ def remove_weight_norm(self):
+ self.enc.remove_weight_norm()
+class Generator(torch.nn.Module):
+ def __init__(
+ self,
+ initial_channel,
+ resblock,
+ resblock_kernel_sizes,
+ resblock_dilation_sizes,
+ upsample_rates,
+ upsample_initial_channel,
+ upsample_kernel_sizes,
+ gin_channels=0,
+ ):
+ super(Generator, self).__init__()
+ self.num_kernels = len(resblock_kernel_sizes)
+ self.num_upsamples = len(upsample_rates)
+ self.conv_pre = Conv1d(
+ initial_channel, upsample_initial_channel, 7, 1, padding=3
+ )
+ resblock = modules.ResBlock1 if resblock == "1" else modules.ResBlock2
+
+ self.ups = nn.ModuleList()
+ for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
+ self.ups.append(
+ weight_norm(
+ ConvTranspose1d(
+ upsample_initial_channel // (2**i),
+ upsample_initial_channel // (2 ** (i + 1)),
+ k,
+ u,
+ padding=(k - u) // 2,
+ )
+ )
+ )
+
+ self.resblocks = nn.ModuleList()
+ for i in range(len(self.ups)):
+ ch = upsample_initial_channel // (2 ** (i + 1))
+ for j, (k, d) in enumerate(
+ zip(resblock_kernel_sizes, resblock_dilation_sizes)
+ ):
+ self.resblocks.append(resblock(ch, k, d))
+
+ self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
+ self.ups.apply(init_weights)
+
+ if gin_channels != 0:
+ self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1)
+
+ def forward(self, x, g=None):
+ x = self.conv_pre(x)
+ if g is not None:
+ x = x + self.cond(g)
+
+ for i in range(self.num_upsamples):
+ x = F.leaky_relu(x, modules.LRELU_SLOPE)
+ x = self.ups[i](x)
+ xs = None
+ for j in range(self.num_kernels):
+ if xs is None:
+ xs = self.resblocks[i * self.num_kernels + j](x)
+ else:
+ xs += self.resblocks[i * self.num_kernels + j](x)
+ x = xs / self.num_kernels
+ x = F.leaky_relu(x)
+ x = self.conv_post(x)
+ x = torch.tanh(x)
+
+ return x
+
+ def remove_weight_norm(self):
+ for l in self.ups:
+ remove_weight_norm(l)
+ for l in self.resblocks:
+ l.remove_weight_norm()
+class SineGen(torch.nn.Module):
+ """ Definition of sine generator
+ SineGen(samp_rate, harmonic_num = 0,
+ sine_amp = 0.1, noise_std = 0.003,
+ voiced_threshold = 0,
+ flag_for_pulse=False)
+ samp_rate: sampling rate in Hz
+ harmonic_num: number of harmonic overtones (default 0)
+ sine_amp: amplitude of sine-wavefrom (default 0.1)
+ noise_std: std of Gaussian noise (default 0.003)
+ voiced_thoreshold: F0 threshold for U/V classification (default 0)
+ flag_for_pulse: this SinGen is used inside PulseGen (default False)
+ Note: when flag_for_pulse is True, the first time step of a voiced
+ segment is always sin(np.pi) or cos(0)
+ """
+
+ def __init__(self, samp_rate, harmonic_num=0,
+ sine_amp=0.1, noise_std=0.003,
+ voiced_threshold=0,
+ flag_for_pulse=False):
+ super(SineGen, self).__init__()
+ self.sine_amp = sine_amp
+ self.noise_std = noise_std
+ self.harmonic_num = harmonic_num
+ self.dim = self.harmonic_num + 1
+ self.sampling_rate = samp_rate
+ self.voiced_threshold = voiced_threshold
+
+ def _f02uv(self, f0):
+ # generate uv signal
+ uv = torch.ones_like(f0)
+ uv = uv * (f0 > self.voiced_threshold)
+ return uv
+
+ def forward(self, f0,upp):
+ """ sine_tensor, uv = forward(f0)
+ input F0: tensor(batchsize=1, length, dim=1)
+ f0 for unvoiced steps should be 0
+ output sine_tensor: tensor(batchsize=1, length, dim)
+ output uv: tensor(batchsize=1, length, 1)
+ """
+ with torch.no_grad():
+ f0 = f0[:, None].transpose(1, 2)
+ f0_buf = torch.zeros(f0.shape[0], f0.shape[1], self.dim,device=f0.device)
+ # fundamental component
+ f0_buf[:, :, 0] = f0[:, :, 0]
+ for idx in np.arange(self.harmonic_num):f0_buf[:, :, idx + 1] = f0_buf[:, :, 0] * (idx + 2)# idx + 2: the (idx+1)-th overtone, (idx+2)-th harmonic
+ rad_values = (f0_buf / self.sampling_rate) % 1###%1意味着n_har的乘积无法后处理优化
+ rand_ini = torch.rand(f0_buf.shape[0], f0_buf.shape[2], device=f0_buf.device)
+ rand_ini[:, 0] = 0
+ rad_values[:, 0, :] = rad_values[:, 0, :] + rand_ini
+ tmp_over_one = torch.cumsum(rad_values, 1)# % 1 #####%1意味着后面的cumsum无法再优化
+ tmp_over_one*=upp
+ tmp_over_one=F.interpolate(tmp_over_one.transpose(2, 1), scale_factor=upp, mode='linear', align_corners=True).transpose(2, 1)
+ rad_values=F.interpolate(rad_values.transpose(2, 1), scale_factor=upp, mode='nearest').transpose(2, 1)#######
+ tmp_over_one%=1
+ tmp_over_one_idx = (tmp_over_one[:, 1:, :] - tmp_over_one[:, :-1, :]) < 0
+ cumsum_shift = torch.zeros_like(rad_values)
+ cumsum_shift[:, 1:, :] = tmp_over_one_idx * -1.0
+ sine_waves = torch.sin(torch.cumsum(rad_values + cumsum_shift, dim=1) * 2 * np.pi)
+ sine_waves = sine_waves * self.sine_amp
+ uv = self._f02uv(f0)
+ uv = F.interpolate(uv.transpose(2, 1), scale_factor=upp, mode='nearest').transpose(2, 1)
+ noise_amp = uv * self.noise_std + (1 - uv) * self.sine_amp / 3
+ noise = noise_amp * torch.randn_like(sine_waves)
+ sine_waves = sine_waves * uv + noise
+ return sine_waves, uv, noise
+class SourceModuleHnNSF(torch.nn.Module):
+ """ SourceModule for hn-nsf
+ SourceModule(sampling_rate, harmonic_num=0, sine_amp=0.1,
+ add_noise_std=0.003, voiced_threshod=0)
+ sampling_rate: sampling_rate in Hz
+ harmonic_num: number of harmonic above F0 (default: 0)
+ sine_amp: amplitude of sine source signal (default: 0.1)
+ add_noise_std: std of additive Gaussian noise (default: 0.003)
+ note that amplitude of noise in unvoiced is decided
+ by sine_amp
+ voiced_threshold: threhold to set U/V given F0 (default: 0)
+ Sine_source, noise_source = SourceModuleHnNSF(F0_sampled)
+ F0_sampled (batchsize, length, 1)
+ Sine_source (batchsize, length, 1)
+ noise_source (batchsize, length 1)
+ uv (batchsize, length, 1)
+ """
+
+ def __init__(self, sampling_rate, harmonic_num=0, sine_amp=0.1,
+ add_noise_std=0.003, voiced_threshod=0,is_half=True):
+ super(SourceModuleHnNSF, self).__init__()
+
+ self.sine_amp = sine_amp
+ self.noise_std = add_noise_std
+ self.is_half=is_half
+ # to produce sine waveforms
+ self.l_sin_gen = SineGen(sampling_rate, harmonic_num,
+ sine_amp, add_noise_std, voiced_threshod)
+
+ # to merge source harmonics into a single excitation
+ self.l_linear = torch.nn.Linear(harmonic_num + 1, 1)
+ self.l_tanh = torch.nn.Tanh()
+
+ def forward(self, x,upp=None):
+ sine_wavs, uv, _ = self.l_sin_gen(x,upp)
+ if(self.is_half==True):sine_wavs=sine_wavs.half()
+ sine_merge = self.l_tanh(self.l_linear(sine_wavs))
+ return sine_merge,None,None# noise, uv
+class GeneratorNSF(torch.nn.Module):
+ def __init__(
+ self,
+ initial_channel,
+ resblock,
+ resblock_kernel_sizes,
+ resblock_dilation_sizes,
+ upsample_rates,
+ upsample_initial_channel,
+ upsample_kernel_sizes,
+ gin_channels=0,
+ sr=40000,
+ is_half=False
+ ):
+ super(GeneratorNSF, self).__init__()
+ self.num_kernels = len(resblock_kernel_sizes)
+ self.num_upsamples = len(upsample_rates)
+
+ self.f0_upsamp = torch.nn.Upsample(scale_factor=np.prod(upsample_rates))
+ self.m_source = SourceModuleHnNSF(
+ sampling_rate=sr,
+ harmonic_num=0,
+ is_half=is_half
+ )
+ self.noise_convs = nn.ModuleList()
+ self.conv_pre = Conv1d(
+ initial_channel, upsample_initial_channel, 7, 1, padding=3
+ )
+ resblock = modules.ResBlock1 if resblock == "1" else modules.ResBlock2
+
+ self.ups = nn.ModuleList()
+ for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
+ c_cur = upsample_initial_channel // (2 ** (i + 1))
+ self.ups.append(
+ weight_norm(
+ ConvTranspose1d(
+ upsample_initial_channel // (2**i),
+ upsample_initial_channel // (2 ** (i + 1)),
+ k,
+ u,
+ padding=(k - u) // 2,
+ )
+ )
+ )
+ if i + 1 < len(upsample_rates):
+ stride_f0 = np.prod(upsample_rates[i + 1:])
+ self.noise_convs.append(Conv1d(
+ 1, c_cur, kernel_size=stride_f0 * 2, stride=stride_f0, padding=stride_f0 // 2))
+ else:
+ self.noise_convs.append(Conv1d(1, c_cur, kernel_size=1))
+
+ self.resblocks = nn.ModuleList()
+ for i in range(len(self.ups)):
+ ch = upsample_initial_channel // (2 ** (i + 1))
+ for j, (k, d) in enumerate(
+ zip(resblock_kernel_sizes, resblock_dilation_sizes)
+ ):
+ self.resblocks.append(resblock(ch, k, d))
+
+ self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
+ self.ups.apply(init_weights)
+
+ if gin_channels != 0:
+ self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1)
+
+ self.upp=np.prod(upsample_rates)
+
+ def forward(self, x, f0,g=None):
+ har_source, noi_source, uv = self.m_source(f0,self.upp)
+ har_source = har_source.transpose(1, 2)
+ x = self.conv_pre(x)
+ if g is not None:
+ x = x + self.cond(g)
+
+ for i in range(self.num_upsamples):
+ x = F.leaky_relu(x, modules.LRELU_SLOPE)
+ x = self.ups[i](x)
+ x_source = self.noise_convs[i](har_source)
+ x = x + x_source
+ xs = None
+ for j in range(self.num_kernels):
+ if xs is None:
+ xs = self.resblocks[i * self.num_kernels + j](x)
+ else:
+ xs += self.resblocks[i * self.num_kernels + j](x)
+ x = xs / self.num_kernels
+ x = F.leaky_relu(x)
+ x = self.conv_post(x)
+ x = torch.tanh(x)
+ return x
+
+ def remove_weight_norm(self):
+ for l in self.ups:
+ remove_weight_norm(l)
+ for l in self.resblocks:
+ l.remove_weight_norm()
+class SynthesizerTrnMs256NSF(nn.Module):
+ """
+ Synthesizer for Training
+ """
+
+ def __init__(
+ self,
+ spec_channels,
+ segment_size,
+ inter_channels,
+ hidden_channels,
+ filter_channels,
+ n_heads,
+ n_layers,
+ kernel_size,
+ p_dropout,
+ resblock,
+ resblock_kernel_sizes,
+ resblock_dilation_sizes,
+ upsample_rates,
+ upsample_initial_channel,
+ upsample_kernel_sizes,
+ spk_embed_dim,
+ gin_channels=0,
+ sr=40000,
+ **kwargs
+ ):
+
+ super().__init__()
+ self.spec_channels = spec_channels
+ self.inter_channels = inter_channels
+ self.hidden_channels = hidden_channels
+ self.filter_channels = filter_channels
+ self.n_heads = n_heads
+ self.n_layers = n_layers
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ self.resblock = resblock
+ self.resblock_kernel_sizes = resblock_kernel_sizes
+ self.resblock_dilation_sizes = resblock_dilation_sizes
+ self.upsample_rates = upsample_rates
+ self.upsample_initial_channel = upsample_initial_channel
+ self.upsample_kernel_sizes = upsample_kernel_sizes
+ self.segment_size = segment_size
+ self.gin_channels = gin_channels
+ self.spk_embed_dim=spk_embed_dim
+ self.enc_p = TextEncoder256(
+ inter_channels,
+ hidden_channels,
+ filter_channels,
+ n_heads,
+ n_layers,
+ kernel_size,
+ p_dropout,
+ )
+ self.dec = GeneratorNSF(
+ inter_channels,
+ resblock,
+ resblock_kernel_sizes,
+ resblock_dilation_sizes,
+ upsample_rates,
+ upsample_initial_channel,
+ upsample_kernel_sizes,
+ gin_channels=0,
+ sr=sr,
+ is_half=kwargs["is_half"]
+ )
+ self.enc_q = PosteriorEncoder(
+ spec_channels,
+ inter_channels,
+ hidden_channels,
+ 5,
+ 1,
+ 16,
+ gin_channels=gin_channels,
+ )
+ self.flow = ResidualCouplingBlock(
+ inter_channels, hidden_channels, 5, 1, 3, gin_channels=gin_channels
+ )
+ self.emb_g = nn.Linear(self.spk_embed_dim, gin_channels)
+
+ def remove_weight_norm(self):
+ self.dec.remove_weight_norm()
+ self.flow.remove_weight_norm()
+ self.enc_q.remove_weight_norm()
+
+ def infer(self, phone, phone_lengths, pitch,pitchf, ds,max_len=None):
+ m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+ if("float16"in str(m_p.dtype)):ds=ds.half()
+ ds=ds.to(m_p.device)
+ g = self.emb_g(ds).unsqueeze(-1) # [b, h, 1]#
+ z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66) * x_mask
+
+ z = self.flow(z_p, x_mask, g=g, reverse=True)
+ o = self.dec((z * x_mask)[:, :, :max_len],pitchf, g=None)
+ return o, x_mask, (z, z_p, m_p, logs_p)
+class SynthesizerTrn256NSFkm(nn.Module):
+ """
+ Synthesizer for Training
+ """
+
+ def __init__(
+ self,
+ spec_channels,
+ segment_size,
+ inter_channels,
+ hidden_channels,
+ filter_channels,
+ n_heads,
+ n_layers,
+ kernel_size,
+ p_dropout,
+ resblock,
+ resblock_kernel_sizes,
+ resblock_dilation_sizes,
+ upsample_rates,
+ upsample_initial_channel,
+ upsample_kernel_sizes,
+ spk_embed_dim,
+ gin_channels=0,
+ sr=40000,
+ **kwargs
+ ):
+
+ super().__init__()
+ self.spec_channels = spec_channels
+ self.inter_channels = inter_channels
+ self.hidden_channels = hidden_channels
+ self.filter_channels = filter_channels
+ self.n_heads = n_heads
+ self.n_layers = n_layers
+ self.kernel_size = kernel_size
+ self.p_dropout = p_dropout
+ self.resblock = resblock
+ self.resblock_kernel_sizes = resblock_kernel_sizes
+ self.resblock_dilation_sizes = resblock_dilation_sizes
+ self.upsample_rates = upsample_rates
+ self.upsample_initial_channel = upsample_initial_channel
+ self.upsample_kernel_sizes = upsample_kernel_sizes
+ self.segment_size = segment_size
+ self.gin_channels = gin_channels
+
+ self.enc_p = TextEncoder256km(
+ inter_channels,
+ hidden_channels,
+ filter_channels,
+ n_heads,
+ n_layers,
+ kernel_size,
+ p_dropout,
+ )
+ self.dec = GeneratorNSF(
+ inter_channels,
+ resblock,
+ resblock_kernel_sizes,
+ resblock_dilation_sizes,
+ upsample_rates,
+ upsample_initial_channel,
+ upsample_kernel_sizes,
+ gin_channels=0,
+ sr=sr,
+ is_half=kwargs["is_half"]
+ )
+ self.enc_q = PosteriorEncoder(
+ spec_channels,
+ inter_channels,
+ hidden_channels,
+ 5,
+ 1,
+ 16,
+ gin_channels=gin_channels,
+ )
+ self.flow = ResidualCouplingBlock(
+ inter_channels, hidden_channels, 5, 1, 3, gin_channels=gin_channels
+ )
+
+ def remove_weight_norm(self):
+ self.dec.remove_weight_norm()
+ self.flow.remove_weight_norm()
+ self.enc_q.remove_weight_norm()
+
+ def forward(self, phone, phone_lengths, pitch, pitchf, y, y_lengths):
+ m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+
+ z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=None)
+ z_p = self.flow(z, y_mask, g=None)
+
+ z_slice, ids_slice = commons.rand_slice_segments(
+ z, y_lengths, self.segment_size
+ )
+
+ pitchf = commons.slice_segments2(
+ pitchf, ids_slice, self.segment_size
+ )
+ o = self.dec(z_slice, pitchf,g=None)
+ return o, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q)
+
+ def infer(self, phone, phone_lengths, pitch, nsff0,max_len=None):
+ # torch.cuda.synchronize()
+ # t0=ttime()
+ m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+ # torch.cuda.synchronize()
+ # t1=ttime()
+ z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66) * x_mask
+ # torch.cuda.synchronize()
+ # t2=ttime()
+ z = self.flow(z_p, x_mask, g=None, reverse=True)
+ # torch.cuda.synchronize()
+ # t3=ttime()
+ o = self.dec((z * x_mask)[:, :, :max_len], nsff0,g=None)
+ # torch.cuda.synchronize()
+ # t4=ttime()
+ # print(1233333333333333333333333,t1-t0,t2-t1,t3-t2,t4-t3)
+ return o, x_mask, (z, z_p, m_p, logs_p)
\ No newline at end of file
diff --git a/infer_pack/modules.py b/infer_pack/modules.py
new file mode 100644
index 0000000000000000000000000000000000000000..960481cedad9a6106f2bf0b9e86e82b120f7b33f
--- /dev/null
+++ b/infer_pack/modules.py
@@ -0,0 +1,522 @@
+import copy
+import math
+import numpy as np
+import scipy
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm
+
+from infer_pack import commons
+from infer_pack.commons import init_weights, get_padding
+from infer_pack.transforms import piecewise_rational_quadratic_transform
+
+
+LRELU_SLOPE = 0.1
+
+
+class LayerNorm(nn.Module):
+ def __init__(self, channels, eps=1e-5):
+ super().__init__()
+ self.channels = channels
+ self.eps = eps
+
+ self.gamma = nn.Parameter(torch.ones(channels))
+ self.beta = nn.Parameter(torch.zeros(channels))
+
+ def forward(self, x):
+ x = x.transpose(1, -1)
+ x = F.layer_norm(x, (self.channels,), self.gamma, self.beta, self.eps)
+ return x.transpose(1, -1)
+
+
+class ConvReluNorm(nn.Module):
+ def __init__(
+ self,
+ in_channels,
+ hidden_channels,
+ out_channels,
+ kernel_size,
+ n_layers,
+ p_dropout,
+ ):
+ super().__init__()
+ self.in_channels = in_channels
+ self.hidden_channels = hidden_channels
+ self.out_channels = out_channels
+ self.kernel_size = kernel_size
+ self.n_layers = n_layers
+ self.p_dropout = p_dropout
+ assert n_layers > 1, "Number of layers should be larger than 0."
+
+ self.conv_layers = nn.ModuleList()
+ self.norm_layers = nn.ModuleList()
+ self.conv_layers.append(
+ nn.Conv1d(
+ in_channels, hidden_channels, kernel_size, padding=kernel_size // 2
+ )
+ )
+ self.norm_layers.append(LayerNorm(hidden_channels))
+ self.relu_drop = nn.Sequential(nn.ReLU(), nn.Dropout(p_dropout))
+ for _ in range(n_layers - 1):
+ self.conv_layers.append(
+ nn.Conv1d(
+ hidden_channels,
+ hidden_channels,
+ kernel_size,
+ padding=kernel_size // 2,
+ )
+ )
+ self.norm_layers.append(LayerNorm(hidden_channels))
+ self.proj = nn.Conv1d(hidden_channels, out_channels, 1)
+ self.proj.weight.data.zero_()
+ self.proj.bias.data.zero_()
+
+ def forward(self, x, x_mask):
+ x_org = x
+ for i in range(self.n_layers):
+ x = self.conv_layers[i](x * x_mask)
+ x = self.norm_layers[i](x)
+ x = self.relu_drop(x)
+ x = x_org + self.proj(x)
+ return x * x_mask
+
+
+class DDSConv(nn.Module):
+ """
+ Dialted and Depth-Separable Convolution
+ """
+
+ def __init__(self, channels, kernel_size, n_layers, p_dropout=0.0):
+ super().__init__()
+ self.channels = channels
+ self.kernel_size = kernel_size
+ self.n_layers = n_layers
+ self.p_dropout = p_dropout
+
+ self.drop = nn.Dropout(p_dropout)
+ self.convs_sep = nn.ModuleList()
+ self.convs_1x1 = nn.ModuleList()
+ self.norms_1 = nn.ModuleList()
+ self.norms_2 = nn.ModuleList()
+ for i in range(n_layers):
+ dilation = kernel_size**i
+ padding = (kernel_size * dilation - dilation) // 2
+ self.convs_sep.append(
+ nn.Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ groups=channels,
+ dilation=dilation,
+ padding=padding,
+ )
+ )
+ self.convs_1x1.append(nn.Conv1d(channels, channels, 1))
+ self.norms_1.append(LayerNorm(channels))
+ self.norms_2.append(LayerNorm(channels))
+
+ def forward(self, x, x_mask, g=None):
+ if g is not None:
+ x = x + g
+ for i in range(self.n_layers):
+ y = self.convs_sep[i](x * x_mask)
+ y = self.norms_1[i](y)
+ y = F.gelu(y)
+ y = self.convs_1x1[i](y)
+ y = self.norms_2[i](y)
+ y = F.gelu(y)
+ y = self.drop(y)
+ x = x + y
+ return x * x_mask
+
+
+class WN(torch.nn.Module):
+ def __init__(
+ self,
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ gin_channels=0,
+ p_dropout=0,
+ ):
+ super(WN, self).__init__()
+ assert kernel_size % 2 == 1
+ self.hidden_channels = hidden_channels
+ self.kernel_size = (kernel_size,)
+ self.dilation_rate = dilation_rate
+ self.n_layers = n_layers
+ self.gin_channels = gin_channels
+ self.p_dropout = p_dropout
+
+ self.in_layers = torch.nn.ModuleList()
+ self.res_skip_layers = torch.nn.ModuleList()
+ self.drop = nn.Dropout(p_dropout)
+
+ if gin_channels != 0:
+ cond_layer = torch.nn.Conv1d(
+ gin_channels, 2 * hidden_channels * n_layers, 1
+ )
+ self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name="weight")
+
+ for i in range(n_layers):
+ dilation = dilation_rate**i
+ padding = int((kernel_size * dilation - dilation) / 2)
+ in_layer = torch.nn.Conv1d(
+ hidden_channels,
+ 2 * hidden_channels,
+ kernel_size,
+ dilation=dilation,
+ padding=padding,
+ )
+ in_layer = torch.nn.utils.weight_norm(in_layer, name="weight")
+ self.in_layers.append(in_layer)
+
+ # last one is not necessary
+ if i < n_layers - 1:
+ res_skip_channels = 2 * hidden_channels
+ else:
+ res_skip_channels = hidden_channels
+
+ res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1)
+ res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name="weight")
+ self.res_skip_layers.append(res_skip_layer)
+
+ def forward(self, x, x_mask, g=None, **kwargs):
+ output = torch.zeros_like(x)
+ n_channels_tensor = torch.IntTensor([self.hidden_channels])
+
+ if g is not None:
+ g = self.cond_layer(g)
+
+ for i in range(self.n_layers):
+ x_in = self.in_layers[i](x)
+ if g is not None:
+ cond_offset = i * 2 * self.hidden_channels
+ g_l = g[:, cond_offset : cond_offset + 2 * self.hidden_channels, :]
+ else:
+ g_l = torch.zeros_like(x_in)
+
+ acts = commons.fused_add_tanh_sigmoid_multiply(x_in, g_l, n_channels_tensor)
+ acts = self.drop(acts)
+
+ res_skip_acts = self.res_skip_layers[i](acts)
+ if i < self.n_layers - 1:
+ res_acts = res_skip_acts[:, : self.hidden_channels, :]
+ x = (x + res_acts) * x_mask
+ output = output + res_skip_acts[:, self.hidden_channels :, :]
+ else:
+ output = output + res_skip_acts
+ return output * x_mask
+
+ def remove_weight_norm(self):
+ if self.gin_channels != 0:
+ torch.nn.utils.remove_weight_norm(self.cond_layer)
+ for l in self.in_layers:
+ torch.nn.utils.remove_weight_norm(l)
+ for l in self.res_skip_layers:
+ torch.nn.utils.remove_weight_norm(l)
+
+
+class ResBlock1(torch.nn.Module):
+ def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5)):
+ super(ResBlock1, self).__init__()
+ self.convs1 = nn.ModuleList(
+ [
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=dilation[0],
+ padding=get_padding(kernel_size, dilation[0]),
+ )
+ ),
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=dilation[1],
+ padding=get_padding(kernel_size, dilation[1]),
+ )
+ ),
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=dilation[2],
+ padding=get_padding(kernel_size, dilation[2]),
+ )
+ ),
+ ]
+ )
+ self.convs1.apply(init_weights)
+
+ self.convs2 = nn.ModuleList(
+ [
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=1,
+ padding=get_padding(kernel_size, 1),
+ )
+ ),
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=1,
+ padding=get_padding(kernel_size, 1),
+ )
+ ),
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=1,
+ padding=get_padding(kernel_size, 1),
+ )
+ ),
+ ]
+ )
+ self.convs2.apply(init_weights)
+
+ def forward(self, x, x_mask=None):
+ for c1, c2 in zip(self.convs1, self.convs2):
+ xt = F.leaky_relu(x, LRELU_SLOPE)
+ if x_mask is not None:
+ xt = xt * x_mask
+ xt = c1(xt)
+ xt = F.leaky_relu(xt, LRELU_SLOPE)
+ if x_mask is not None:
+ xt = xt * x_mask
+ xt = c2(xt)
+ x = xt + x
+ if x_mask is not None:
+ x = x * x_mask
+ return x
+
+ def remove_weight_norm(self):
+ for l in self.convs1:
+ remove_weight_norm(l)
+ for l in self.convs2:
+ remove_weight_norm(l)
+
+
+class ResBlock2(torch.nn.Module):
+ def __init__(self, channels, kernel_size=3, dilation=(1, 3)):
+ super(ResBlock2, self).__init__()
+ self.convs = nn.ModuleList(
+ [
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=dilation[0],
+ padding=get_padding(kernel_size, dilation[0]),
+ )
+ ),
+ weight_norm(
+ Conv1d(
+ channels,
+ channels,
+ kernel_size,
+ 1,
+ dilation=dilation[1],
+ padding=get_padding(kernel_size, dilation[1]),
+ )
+ ),
+ ]
+ )
+ self.convs.apply(init_weights)
+
+ def forward(self, x, x_mask=None):
+ for c in self.convs:
+ xt = F.leaky_relu(x, LRELU_SLOPE)
+ if x_mask is not None:
+ xt = xt * x_mask
+ xt = c(xt)
+ x = xt + x
+ if x_mask is not None:
+ x = x * x_mask
+ return x
+
+ def remove_weight_norm(self):
+ for l in self.convs:
+ remove_weight_norm(l)
+
+
+class Log(nn.Module):
+ def forward(self, x, x_mask, reverse=False, **kwargs):
+ if not reverse:
+ y = torch.log(torch.clamp_min(x, 1e-5)) * x_mask
+ logdet = torch.sum(-y, [1, 2])
+ return y, logdet
+ else:
+ x = torch.exp(x) * x_mask
+ return x
+
+
+class Flip(nn.Module):
+ def forward(self, x, *args, reverse=False, **kwargs):
+ x = torch.flip(x, [1])
+ if not reverse:
+ logdet = torch.zeros(x.size(0)).to(dtype=x.dtype, device=x.device)
+ return x, logdet
+ else:
+ return x
+
+
+class ElementwiseAffine(nn.Module):
+ def __init__(self, channels):
+ super().__init__()
+ self.channels = channels
+ self.m = nn.Parameter(torch.zeros(channels, 1))
+ self.logs = nn.Parameter(torch.zeros(channels, 1))
+
+ def forward(self, x, x_mask, reverse=False, **kwargs):
+ if not reverse:
+ y = self.m + torch.exp(self.logs) * x
+ y = y * x_mask
+ logdet = torch.sum(self.logs * x_mask, [1, 2])
+ return y, logdet
+ else:
+ x = (x - self.m) * torch.exp(-self.logs) * x_mask
+ return x
+
+
+class ResidualCouplingLayer(nn.Module):
+ def __init__(
+ self,
+ channels,
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ p_dropout=0,
+ gin_channels=0,
+ mean_only=False,
+ ):
+ assert channels % 2 == 0, "channels should be divisible by 2"
+ super().__init__()
+ self.channels = channels
+ self.hidden_channels = hidden_channels
+ self.kernel_size = kernel_size
+ self.dilation_rate = dilation_rate
+ self.n_layers = n_layers
+ self.half_channels = channels // 2
+ self.mean_only = mean_only
+
+ self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
+ self.enc = WN(
+ hidden_channels,
+ kernel_size,
+ dilation_rate,
+ n_layers,
+ p_dropout=p_dropout,
+ gin_channels=gin_channels,
+ )
+ self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1)
+ self.post.weight.data.zero_()
+ self.post.bias.data.zero_()
+
+ def forward(self, x, x_mask, g=None, reverse=False):
+ x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+ h = self.pre(x0) * x_mask
+ h = self.enc(h, x_mask, g=g)
+ stats = self.post(h) * x_mask
+ if not self.mean_only:
+ m, logs = torch.split(stats, [self.half_channels] * 2, 1)
+ else:
+ m = stats
+ logs = torch.zeros_like(m)
+
+ if not reverse:
+ x1 = m + x1 * torch.exp(logs) * x_mask
+ x = torch.cat([x0, x1], 1)
+ logdet = torch.sum(logs, [1, 2])
+ return x, logdet
+ else:
+ x1 = (x1 - m) * torch.exp(-logs) * x_mask
+ x = torch.cat([x0, x1], 1)
+ return x
+
+ def remove_weight_norm(self):
+ self.enc.remove_weight_norm()
+
+
+class ConvFlow(nn.Module):
+ def __init__(
+ self,
+ in_channels,
+ filter_channels,
+ kernel_size,
+ n_layers,
+ num_bins=10,
+ tail_bound=5.0,
+ ):
+ super().__init__()
+ self.in_channels = in_channels
+ self.filter_channels = filter_channels
+ self.kernel_size = kernel_size
+ self.n_layers = n_layers
+ self.num_bins = num_bins
+ self.tail_bound = tail_bound
+ self.half_channels = in_channels // 2
+
+ self.pre = nn.Conv1d(self.half_channels, filter_channels, 1)
+ self.convs = DDSConv(filter_channels, kernel_size, n_layers, p_dropout=0.0)
+ self.proj = nn.Conv1d(
+ filter_channels, self.half_channels * (num_bins * 3 - 1), 1
+ )
+ self.proj.weight.data.zero_()
+ self.proj.bias.data.zero_()
+
+ def forward(self, x, x_mask, g=None, reverse=False):
+ x0, x1 = torch.split(x, [self.half_channels] * 2, 1)
+ h = self.pre(x0)
+ h = self.convs(h, x_mask, g=g)
+ h = self.proj(h) * x_mask
+
+ b, c, t = x0.shape
+ h = h.reshape(b, c, -1, t).permute(0, 1, 3, 2) # [b, cx?, t] -> [b, c, t, ?]
+
+ unnormalized_widths = h[..., : self.num_bins] / math.sqrt(self.filter_channels)
+ unnormalized_heights = h[..., self.num_bins : 2 * self.num_bins] / math.sqrt(
+ self.filter_channels
+ )
+ unnormalized_derivatives = h[..., 2 * self.num_bins :]
+
+ x1, logabsdet = piecewise_rational_quadratic_transform(
+ x1,
+ unnormalized_widths,
+ unnormalized_heights,
+ unnormalized_derivatives,
+ inverse=reverse,
+ tails="linear",
+ tail_bound=self.tail_bound,
+ )
+
+ x = torch.cat([x0, x1], 1) * x_mask
+ logdet = torch.sum(logabsdet * x_mask, [1, 2])
+ if not reverse:
+ return x, logdet
+ else:
+ return x
diff --git a/infer_pack/transforms.py b/infer_pack/transforms.py
new file mode 100644
index 0000000000000000000000000000000000000000..4793d67ca5a5630e0ffe0f9fb29445c949e64dae
--- /dev/null
+++ b/infer_pack/transforms.py
@@ -0,0 +1,193 @@
+import torch
+from torch.nn import functional as F
+
+import numpy as np
+
+
+DEFAULT_MIN_BIN_WIDTH = 1e-3
+DEFAULT_MIN_BIN_HEIGHT = 1e-3
+DEFAULT_MIN_DERIVATIVE = 1e-3
+
+
+def piecewise_rational_quadratic_transform(inputs,
+ unnormalized_widths,
+ unnormalized_heights,
+ unnormalized_derivatives,
+ inverse=False,
+ tails=None,
+ tail_bound=1.,
+ min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+ min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+ min_derivative=DEFAULT_MIN_DERIVATIVE):
+
+ if tails is None:
+ spline_fn = rational_quadratic_spline
+ spline_kwargs = {}
+ else:
+ spline_fn = unconstrained_rational_quadratic_spline
+ spline_kwargs = {
+ 'tails': tails,
+ 'tail_bound': tail_bound
+ }
+
+ outputs, logabsdet = spline_fn(
+ inputs=inputs,
+ unnormalized_widths=unnormalized_widths,
+ unnormalized_heights=unnormalized_heights,
+ unnormalized_derivatives=unnormalized_derivatives,
+ inverse=inverse,
+ min_bin_width=min_bin_width,
+ min_bin_height=min_bin_height,
+ min_derivative=min_derivative,
+ **spline_kwargs
+ )
+ return outputs, logabsdet
+
+
+def searchsorted(bin_locations, inputs, eps=1e-6):
+ bin_locations[..., -1] += eps
+ return torch.sum(
+ inputs[..., None] >= bin_locations,
+ dim=-1
+ ) - 1
+
+
+def unconstrained_rational_quadratic_spline(inputs,
+ unnormalized_widths,
+ unnormalized_heights,
+ unnormalized_derivatives,
+ inverse=False,
+ tails='linear',
+ tail_bound=1.,
+ min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+ min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+ min_derivative=DEFAULT_MIN_DERIVATIVE):
+ inside_interval_mask = (inputs >= -tail_bound) & (inputs <= tail_bound)
+ outside_interval_mask = ~inside_interval_mask
+
+ outputs = torch.zeros_like(inputs)
+ logabsdet = torch.zeros_like(inputs)
+
+ if tails == 'linear':
+ unnormalized_derivatives = F.pad(unnormalized_derivatives, pad=(1, 1))
+ constant = np.log(np.exp(1 - min_derivative) - 1)
+ unnormalized_derivatives[..., 0] = constant
+ unnormalized_derivatives[..., -1] = constant
+
+ outputs[outside_interval_mask] = inputs[outside_interval_mask]
+ logabsdet[outside_interval_mask] = 0
+ else:
+ raise RuntimeError('{} tails are not implemented.'.format(tails))
+
+ outputs[inside_interval_mask], logabsdet[inside_interval_mask] = rational_quadratic_spline(
+ inputs=inputs[inside_interval_mask],
+ unnormalized_widths=unnormalized_widths[inside_interval_mask, :],
+ unnormalized_heights=unnormalized_heights[inside_interval_mask, :],
+ unnormalized_derivatives=unnormalized_derivatives[inside_interval_mask, :],
+ inverse=inverse,
+ left=-tail_bound, right=tail_bound, bottom=-tail_bound, top=tail_bound,
+ min_bin_width=min_bin_width,
+ min_bin_height=min_bin_height,
+ min_derivative=min_derivative
+ )
+
+ return outputs, logabsdet
+
+def rational_quadratic_spline(inputs,
+ unnormalized_widths,
+ unnormalized_heights,
+ unnormalized_derivatives,
+ inverse=False,
+ left=0., right=1., bottom=0., top=1.,
+ min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+ min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+ min_derivative=DEFAULT_MIN_DERIVATIVE):
+ if torch.min(inputs) < left or torch.max(inputs) > right:
+ raise ValueError('Input to a transform is not within its domain')
+
+ num_bins = unnormalized_widths.shape[-1]
+
+ if min_bin_width * num_bins > 1.0:
+ raise ValueError('Minimal bin width too large for the number of bins')
+ if min_bin_height * num_bins > 1.0:
+ raise ValueError('Minimal bin height too large for the number of bins')
+
+ widths = F.softmax(unnormalized_widths, dim=-1)
+ widths = min_bin_width + (1 - min_bin_width * num_bins) * widths
+ cumwidths = torch.cumsum(widths, dim=-1)
+ cumwidths = F.pad(cumwidths, pad=(1, 0), mode='constant', value=0.0)
+ cumwidths = (right - left) * cumwidths + left
+ cumwidths[..., 0] = left
+ cumwidths[..., -1] = right
+ widths = cumwidths[..., 1:] - cumwidths[..., :-1]
+
+ derivatives = min_derivative + F.softplus(unnormalized_derivatives)
+
+ heights = F.softmax(unnormalized_heights, dim=-1)
+ heights = min_bin_height + (1 - min_bin_height * num_bins) * heights
+ cumheights = torch.cumsum(heights, dim=-1)
+ cumheights = F.pad(cumheights, pad=(1, 0), mode='constant', value=0.0)
+ cumheights = (top - bottom) * cumheights + bottom
+ cumheights[..., 0] = bottom
+ cumheights[..., -1] = top
+ heights = cumheights[..., 1:] - cumheights[..., :-1]
+
+ if inverse:
+ bin_idx = searchsorted(cumheights, inputs)[..., None]
+ else:
+ bin_idx = searchsorted(cumwidths, inputs)[..., None]
+
+ input_cumwidths = cumwidths.gather(-1, bin_idx)[..., 0]
+ input_bin_widths = widths.gather(-1, bin_idx)[..., 0]
+
+ input_cumheights = cumheights.gather(-1, bin_idx)[..., 0]
+ delta = heights / widths
+ input_delta = delta.gather(-1, bin_idx)[..., 0]
+
+ input_derivatives = derivatives.gather(-1, bin_idx)[..., 0]
+ input_derivatives_plus_one = derivatives[..., 1:].gather(-1, bin_idx)[..., 0]
+
+ input_heights = heights.gather(-1, bin_idx)[..., 0]
+
+ if inverse:
+ a = (((inputs - input_cumheights) * (input_derivatives
+ + input_derivatives_plus_one
+ - 2 * input_delta)
+ + input_heights * (input_delta - input_derivatives)))
+ b = (input_heights * input_derivatives
+ - (inputs - input_cumheights) * (input_derivatives
+ + input_derivatives_plus_one
+ - 2 * input_delta))
+ c = - input_delta * (inputs - input_cumheights)
+
+ discriminant = b.pow(2) - 4 * a * c
+ assert (discriminant >= 0).all()
+
+ root = (2 * c) / (-b - torch.sqrt(discriminant))
+ outputs = root * input_bin_widths + input_cumwidths
+
+ theta_one_minus_theta = root * (1 - root)
+ denominator = input_delta + ((input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+ * theta_one_minus_theta)
+ derivative_numerator = input_delta.pow(2) * (input_derivatives_plus_one * root.pow(2)
+ + 2 * input_delta * theta_one_minus_theta
+ + input_derivatives * (1 - root).pow(2))
+ logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+ return outputs, -logabsdet
+ else:
+ theta = (inputs - input_cumwidths) / input_bin_widths
+ theta_one_minus_theta = theta * (1 - theta)
+
+ numerator = input_heights * (input_delta * theta.pow(2)
+ + input_derivatives * theta_one_minus_theta)
+ denominator = input_delta + ((input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+ * theta_one_minus_theta)
+ outputs = input_cumheights + numerator / denominator
+
+ derivative_numerator = input_delta.pow(2) * (input_derivatives_plus_one * theta.pow(2)
+ + 2 * input_delta * theta_one_minus_theta
+ + input_derivatives * (1 - theta).pow(2))
+ logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+ return outputs, logabsdet
diff --git a/infer_uvr5.py b/infer_uvr5.py
new file mode 100644
index 0000000000000000000000000000000000000000..cd52f2d018d76e85039172b7ab35c50fb21f4fd2
--- /dev/null
+++ b/infer_uvr5.py
@@ -0,0 +1,108 @@
+import os,sys,torch,warnings,pdb
+warnings.filterwarnings("ignore")
+import librosa
+import importlib
+import numpy as np
+import hashlib , math
+from tqdm import tqdm
+from uvr5_pack.lib_v5 import spec_utils
+from uvr5_pack.utils import _get_name_params,inference
+from uvr5_pack.lib_v5.model_param_init import ModelParameters
+from scipy.io import wavfile
+
+class _audio_pre_():
+ def __init__(self, model_path,device,is_half):
+ self.model_path = model_path
+ self.device = device
+ self.data = {
+ # Processing Options
+ 'postprocess': False,
+ 'tta': False,
+ # Constants
+ 'window_size': 512,
+ 'agg': 10,
+ 'high_end_process': 'mirroring',
+ }
+ nn_arch_sizes = [
+ 31191, # default
+ 33966,61968, 123821, 123812, 537238 # custom
+ ]
+ self.nn_architecture = list('{}KB'.format(s) for s in nn_arch_sizes)
+ model_size = math.ceil(os.stat(model_path ).st_size / 1024)
+ nn_architecture = '{}KB'.format(min(nn_arch_sizes, key=lambda x:abs(x-model_size)))
+ nets = importlib.import_module('uvr5_pack.lib_v5.nets' + f'_{nn_architecture}'.replace('_{}KB'.format(nn_arch_sizes[0]), ''), package=None)
+ model_hash = hashlib.md5(open(model_path,'rb').read()).hexdigest()
+ param_name ,model_params_d = _get_name_params(model_path , model_hash)
+
+ mp = ModelParameters(model_params_d)
+ model = nets.CascadedASPPNet(mp.param['bins'] * 2)
+ cpk = torch.load( model_path , map_location='cpu')
+ model.load_state_dict(cpk)
+ model.eval()
+ if(is_half==True):model = model.half().to(device)
+ else:model = model.to(device)
+
+ self.mp = mp
+ self.model = model
+
+ def _path_audio_(self, music_file ,ins_root=None,vocal_root=None):
+ if(ins_root is None and vocal_root is None):return "No save root."
+ name=os.path.basename(music_file)
+ if(ins_root is not None):os.makedirs(ins_root, exist_ok=True)
+ if(vocal_root is not None):os.makedirs(vocal_root , exist_ok=True)
+ X_wave, y_wave, X_spec_s, y_spec_s = {}, {}, {}, {}
+ bands_n = len(self.mp.param['band'])
+ # print(bands_n)
+ for d in range(bands_n, 0, -1):
+ bp = self.mp.param['band'][d]
+ if d == bands_n: # high-end band
+ X_wave[d], _ = librosa.core.load(
+ music_file, bp['sr'], False, dtype=np.float32, res_type=bp['res_type'])
+ if X_wave[d].ndim == 1:
+ X_wave[d] = np.asfortranarray([X_wave[d], X_wave[d]])
+ else: # lower bands
+ X_wave[d] = librosa.core.resample(X_wave[d+1], self.mp.param['band'][d+1]['sr'], bp['sr'], res_type=bp['res_type'])
+ # Stft of wave source
+ X_spec_s[d] = spec_utils.wave_to_spectrogram_mt(X_wave[d], bp['hl'], bp['n_fft'], self.mp.param['mid_side'], self.mp.param['mid_side_b2'], self.mp.param['reverse'])
+ # pdb.set_trace()
+ if d == bands_n and self.data['high_end_process'] != 'none':
+ input_high_end_h = (bp['n_fft']//2 - bp['crop_stop']) + ( self.mp.param['pre_filter_stop'] - self.mp.param['pre_filter_start'])
+ input_high_end = X_spec_s[d][:, bp['n_fft']//2-input_high_end_h:bp['n_fft']//2, :]
+
+ X_spec_m = spec_utils.combine_spectrograms(X_spec_s, self.mp)
+ aggresive_set = float(self.data['agg']/100)
+ aggressiveness = {'value': aggresive_set, 'split_bin': self.mp.param['band'][1]['crop_stop']}
+ with torch.no_grad():
+ pred, X_mag, X_phase = inference(X_spec_m,self.device,self.model, aggressiveness,self.data)
+ # Postprocess
+ if self.data['postprocess']:
+ pred_inv = np.clip(X_mag - pred, 0, np.inf)
+ pred = spec_utils.mask_silence(pred, pred_inv)
+ y_spec_m = pred * X_phase
+ v_spec_m = X_spec_m - y_spec_m
+
+ if (ins_root is not None):
+ if self.data['high_end_process'].startswith('mirroring'):
+ input_high_end_ = spec_utils.mirroring(self.data['high_end_process'], y_spec_m, input_high_end, self.mp)
+ wav_instrument = spec_utils.cmb_spectrogram_to_wave(y_spec_m, self.mp,input_high_end_h, input_high_end_)
+ else:
+ wav_instrument = spec_utils.cmb_spectrogram_to_wave(y_spec_m, self.mp)
+ print ('%s instruments done'%name)
+ wavfile.write(os.path.join(ins_root, 'instrument_{}.wav'.format(name) ), self.mp.param['sr'], (np.array(wav_instrument)*32768).astype("int16")) #
+ if (vocal_root is not None):
+ if self.data['high_end_process'].startswith('mirroring'):
+ input_high_end_ = spec_utils.mirroring(self.data['high_end_process'], v_spec_m, input_high_end, self.mp)
+ wav_vocals = spec_utils.cmb_spectrogram_to_wave(v_spec_m, self.mp, input_high_end_h, input_high_end_)
+ else:
+ wav_vocals = spec_utils.cmb_spectrogram_to_wave(v_spec_m, self.mp)
+ print ('%s vocals done'%name)
+ wavfile.write(os.path.join(vocal_root , 'vocal_{}.wav'.format(name) ), self.mp.param['sr'], (np.array(wav_vocals)*32768).astype("int16"))
+
+if __name__ == '__main__':
+ device = 'cuda'
+ is_half=True
+ model_path='uvr5_weights/2_HP-UVR.pth'
+ pre_fun = _audio_pre_(model_path=model_path,device=device,is_half=True)
+ audio_path = '神女劈观.aac'
+ save_path = 'opt'
+ pre_fun._path_audio_(audio_path , save_path,save_path)
diff --git a/myinfer.py b/myinfer.py
new file mode 100644
index 0000000000000000000000000000000000000000..2c3f12df4f651af8030e9b97a928c683bf332ffc
--- /dev/null
+++ b/myinfer.py
@@ -0,0 +1,156 @@
+'''
+runtime\python.exe myinfer.py 0 "E:\codes\py39\RVC-beta\todo-songs\1111.wav" "E:\codes\py39\logs\mi-test\added_IVF677_Flat_nprobe_7.index" harvest "test.wav" "weights/mi-test.pth" 0.6 cuda:0 True
+'''
+import os,sys,pdb,torch
+now_dir = os.getcwd()
+sys.path.append(now_dir)
+import argparse
+import glob
+import sys
+import torch
+from multiprocessing import cpu_count
+class Config:
+ def __init__(self,device,is_half):
+ self.device = device
+ self.is_half = is_half
+ self.n_cpu = 0
+ self.gpu_name = None
+ self.gpu_mem = None
+ self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config()
+
+ def device_config(self) -> tuple:
+ if torch.cuda.is_available():
+ i_device = int(self.device.split(":")[-1])
+ self.gpu_name = torch.cuda.get_device_name(i_device)
+ if (
+ ("16" in self.gpu_name and "V100" not in self.gpu_name.upper())
+ or "P40" in self.gpu_name.upper()
+ or "1060" in self.gpu_name
+ or "1070" in self.gpu_name
+ or "1080" in self.gpu_name
+ ):
+ print("16系/10系显卡和P40强制单精度")
+ self.is_half = False
+ for config_file in ["32k.json", "40k.json", "48k.json"]:
+ with open(f"configs/{config_file}", "r") as f:
+ strr = f.read().replace("true", "false")
+ with open(f"configs/{config_file}", "w") as f:
+ f.write(strr)
+ with open("trainset_preprocess_pipeline_print.py", "r") as f:
+ strr = f.read().replace("3.7", "3.0")
+ with open("trainset_preprocess_pipeline_print.py", "w") as f:
+ f.write(strr)
+ else:
+ self.gpu_name = None
+ self.gpu_mem = int(
+ torch.cuda.get_device_properties(i_device).total_memory
+ / 1024
+ / 1024
+ / 1024
+ + 0.4
+ )
+ if self.gpu_mem <= 4:
+ with open("trainset_preprocess_pipeline_print.py", "r") as f:
+ strr = f.read().replace("3.7", "3.0")
+ with open("trainset_preprocess_pipeline_print.py", "w") as f:
+ f.write(strr)
+ elif torch.backends.mps.is_available():
+ print("没有发现支持的N卡, 使用MPS进行推理")
+ self.device = "mps"
+ else:
+ print("没有发现支持的N卡, 使用CPU进行推理")
+ self.device = "cpu"
+ self.is_half = True
+
+ if self.n_cpu == 0:
+ self.n_cpu = cpu_count()
+
+ if self.is_half:
+ # 6G显存配置
+ x_pad = 3
+ x_query = 10
+ x_center = 60
+ x_max = 65
+ else:
+ # 5G显存配置
+ x_pad = 1
+ x_query = 6
+ x_center = 38
+ x_max = 41
+
+ if self.gpu_mem != None and self.gpu_mem <= 4:
+ x_pad = 1
+ x_query = 5
+ x_center = 30
+ x_max = 32
+
+ return x_pad, x_query, x_center, x_max
+
+f0up_key=sys.argv[1]
+input_path=sys.argv[2]
+index_path=sys.argv[3]
+f0method=sys.argv[4]#harvest or pm
+opt_path=sys.argv[5]
+model_path=sys.argv[6]
+index_rate=float(sys.argv[7])
+device=sys.argv[8]
+is_half=bool(sys.argv[9])
+print(sys.argv)
+config=Config(device,is_half)
+now_dir=os.getcwd()
+sys.path.append(now_dir)
+from vc_infer_pipeline import VC
+from infer_pack.models import SynthesizerTrnMs256NSFsid, SynthesizerTrnMs256NSFsid_nono
+from my_utils import load_audio
+from fairseq import checkpoint_utils
+from scipy.io import wavfile
+
+hubert_model=None
+def load_hubert():
+ global hubert_model
+ models, saved_cfg, task = checkpoint_utils.load_model_ensemble_and_task(["hubert_base.pt"],suffix="",)
+ hubert_model = models[0]
+ hubert_model = hubert_model.to(device)
+ if(is_half):hubert_model = hubert_model.half()
+ else:hubert_model = hubert_model.float()
+ hubert_model.eval()
+
+def vc_single(sid,input_audio,f0_up_key,f0_file,f0_method,file_index,index_rate):
+ global tgt_sr,net_g,vc,hubert_model
+ if input_audio is None:return "You need to upload an audio", None
+ f0_up_key = int(f0_up_key)
+ audio=load_audio(input_audio,16000)
+ times = [0, 0, 0]
+ if(hubert_model==None):load_hubert()
+ if_f0 = cpt.get("f0", 1)
+ # audio_opt=vc.pipeline(hubert_model,net_g,sid,audio,times,f0_up_key,f0_method,file_index,file_big_npy,index_rate,if_f0,f0_file=f0_file)
+ audio_opt=vc.pipeline(hubert_model,net_g,sid,audio,times,f0_up_key,f0_method,file_index,index_rate,if_f0,f0_file=f0_file)
+ print(times)
+ return audio_opt
+
+
+def get_vc(model_path):
+ global n_spk,tgt_sr,net_g,vc,cpt,device,is_half
+ print("loading pth %s"%model_path)
+ cpt = torch.load(model_path, map_location="cpu")
+ tgt_sr = cpt["config"][-1]
+ cpt["config"][-3]=cpt["weight"]["emb_g.weight"].shape[0]#n_spk
+ if_f0=cpt.get("f0",1)
+ if(if_f0==1):
+ net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half)
+ else:
+ net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
+ del net_g.enc_q
+ print(net_g.load_state_dict(cpt["weight"], strict=False)) # 不加这一行清不干净，真奇葩
+ net_g.eval().to(device)
+ if (is_half):net_g = net_g.half()
+ else:net_g = net_g.float()
+ vc = VC(tgt_sr, config)
+ n_spk=cpt["config"][-3]
+ # return {"visible": True,"maximum": n_spk, "__type__": "update"}
+
+
+get_vc(model_path)
+wav_opt=vc_single(0,input_path,f0up_key,None,f0method,index_path,index_rate)
+wavfile.write(opt_path, tgt_sr, wav_opt)
+
diff --git a/pretrained/D32k.pth b/pretrained/D32k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..346293c7023f67178a4c85da94cc3e85bdb320ef
--- /dev/null
+++ b/pretrained/D32k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ab20645829460fdad0d3c44254f1ab53c32cae50c22a66c926ae5aa30abda6f
+size 109978943
diff --git a/pretrained/D40k.pth b/pretrained/D40k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..11fd6ad63daeb5d786e78d843a6f0800c3d2394c
--- /dev/null
+++ b/pretrained/D40k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:547f66dbbcd9023b9051ed244d12ab043ba8a4e854b154cc28761ac7c002909b
+size 109978943
diff --git a/pretrained/D48k.pth b/pretrained/D48k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..48beafd328bedd8cf493bbf5f4a9d37d2260024a
--- /dev/null
+++ b/pretrained/D48k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cc013fa60ed9c3f902f5bd99f48c7e3b9352d763d4d3cd6bc241c37b0bfd9ad
+size 109978943
diff --git a/pretrained/G32k.pth b/pretrained/G32k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..b32d2b44a9f6668964a41721bf41381daa5eccaf
--- /dev/null
+++ b/pretrained/G32k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90e0458e39efd50beef48ab398c9f88cec2b405d69565e2ad958a25882936aa1
+size 72653893
diff --git a/pretrained/G40k.pth b/pretrained/G40k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..4b07c594c703b27428cd796fdc205562c8974dba
--- /dev/null
+++ b/pretrained/G40k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1e7e413d5b1009777c39cd232ac4d91a81382161d19350fe15bf3f137b01425
+size 72763063
diff --git a/pretrained/G48k.pth b/pretrained/G48k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..75a3b815961447e9df877314ad8f3bb458d72758
--- /dev/null
+++ b/pretrained/G48k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d4a19433aa20617d0712dce2d379620b2f9b80e21b58d6081ba442f203e26d0
+size 72850501
diff --git a/pretrained/f0D32k.pth b/pretrained/f0D32k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..9f36ffb9c4afdf77440f4619688d985daa25b107
--- /dev/null
+++ b/pretrained/f0D32k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:294db3087236e2c75260d6179056791c9231245daf5d0485545d9e54c4057c77
+size 109978943
diff --git a/pretrained/f0D40k.pth b/pretrained/f0D40k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..da7dad91a825dac3a2ed0fb418c99be4a804811e
--- /dev/null
+++ b/pretrained/f0D40k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d4f5a441594b470d67579958b2fd4c6b992852ded28ff9e72eda67abcebe423
+size 109978943
diff --git a/pretrained/f0D48k.pth b/pretrained/f0D48k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..789032e0a223731da818f55c6f65df22a354d6e3
--- /dev/null
+++ b/pretrained/f0D48k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b84c8bf347ad1e539c842e8f2a4c36ecd9e7fb23c16041189e4877e9b07925c
+size 109978943
diff --git a/pretrained/f0G32k.pth b/pretrained/f0G32k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..d5e2abac08b9cd774c7647a7605582c0fe97d7b9
--- /dev/null
+++ b/pretrained/f0G32k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68c08e16e47d2796427b518c9b702edeae5c9d11f8d62608215f10c776d6170b
+size 72795627
diff --git a/pretrained/f0G40k.pth b/pretrained/f0G40k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..dca2d0aff1d5093e98153b562f813778bf5ea04a
--- /dev/null
+++ b/pretrained/f0G40k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a562c00b615367ed2bd97857346a26ce05b5c231629264f5ca2e19ce5f8d59ea
+size 72909665
diff --git a/pretrained/f0G48k.pth b/pretrained/f0G48k.pth
new file mode 100644
index 0000000000000000000000000000000000000000..590e8219fa63743f9f98072e88841920f03c64b6
--- /dev/null
+++ b/pretrained/f0G48k.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f36a4b1e1c3009ee72e5e6c572e0418d8585094e94b00ca9bd5c2b6c8a6c404
+size 73008619
diff --git a/slicer.py b/slicer.py
new file mode 100644
index 0000000000000000000000000000000000000000..11c441b8e2b3d19378b17b8ab5dc7a6ec8d57351
--- /dev/null
+++ b/slicer.py
@@ -0,0 +1,151 @@
+import os.path
+from argparse import ArgumentParser
+import time
+
+import librosa
+import numpy as np
+import soundfile
+from scipy.ndimage import maximum_filter1d, uniform_filter1d
+
+
+def timeit(func):
+ def run(*args, **kwargs):
+ t = time.time()
+ res = func(*args, **kwargs)
+ print('executing \'%s\' costed %.3fs' % (func.__name__, time.time() - t))
+ return res
+ return run
+
+
+# @timeit
+def _window_maximum(arr, win_sz):
+ return maximum_filter1d(arr, size=win_sz)[win_sz // 2: win_sz // 2 + arr.shape[0] - win_sz + 1]
+
+
+# @timeit
+def _window_rms(arr, win_sz):
+ filtered = np.sqrt(uniform_filter1d(np.power(arr, 2), win_sz) - np.power(uniform_filter1d(arr, win_sz), 2))
+ return filtered[win_sz // 2: win_sz // 2 + arr.shape[0] - win_sz + 1]
+
+
+def level2db(levels, eps=1e-12):
+ return 20 * np.log10(np.clip(levels, a_min=eps, a_max=1))
+
+
+def _apply_slice(audio, begin, end):
+ if len(audio.shape) > 1:
+ return audio[:, begin: end]
+ else:
+ return audio[begin: end]
+
+
+class Slicer:
+ def __init__(self,
+ sr: int,
+ db_threshold: float = -40,
+ min_length: int = 5000,
+ win_l: int = 300,
+ win_s: int = 20,
+ max_silence_kept: int = 500):
+ self.db_threshold = db_threshold
+ self.min_samples = round(sr * min_length / 1000)
+ self.win_ln = round(sr * win_l / 1000)
+ self.win_sn = round(sr * win_s / 1000)
+ self.max_silence = round(sr * max_silence_kept / 1000)
+ if not self.min_samples >= self.win_ln >= self.win_sn:
+ raise ValueError('The following condition must be satisfied: min_length >= win_l >= win_s')
+ if not self.max_silence >= self.win_sn:
+ raise ValueError('The following condition must be satisfied: max_silence_kept >= win_s')
+
+ @timeit
+ def slice(self, audio):
+ if len(audio.shape) > 1:
+ samples = librosa.to_mono(audio)
+ else:
+ samples = audio
+ if samples.shape[0] <= self.min_samples:
+ return [audio]
+ # get absolute amplitudes
+ abs_amp = np.abs(samples - np.mean(samples))
+ # calculate local maximum with large window
+ win_max_db = level2db(_window_maximum(abs_amp, win_sz=self.win_ln))
+ sil_tags = []
+ left = right = 0
+ while right < win_max_db.shape[0]:
+ if win_max_db[right] < self.db_threshold:
+ right += 1
+ elif left == right:
+ left += 1
+ right += 1
+ else:
+ if left == 0:
+ split_loc_l = left
+ else:
+ sil_left_n = min(self.max_silence, (right + self.win_ln - left) // 2)
+ rms_db_left = level2db(_window_rms(samples[left: left + sil_left_n], win_sz=self.win_sn))
+ split_win_l = left + np.argmin(rms_db_left)
+ split_loc_l = split_win_l + np.argmin(abs_amp[split_win_l: split_win_l + self.win_sn])
+ if len(sil_tags) != 0 and split_loc_l - sil_tags[-1][1] < self.min_samples and right < win_max_db.shape[0] - 1:
+ right += 1
+ left = right
+ continue
+ if right == win_max_db.shape[0] - 1:
+ split_loc_r = right + self.win_ln
+ else:
+ sil_right_n = min(self.max_silence, (right + self.win_ln - left) // 2)
+ rms_db_right = level2db(_window_rms(samples[right + self.win_ln - sil_right_n: right + self.win_ln], win_sz=self.win_sn))
+ split_win_r = right + self.win_ln - sil_right_n + np.argmin(rms_db_right)
+ split_loc_r = split_win_r + np.argmin(abs_amp[split_win_r: split_win_r + self.win_sn])
+ sil_tags.append((split_loc_l, split_loc_r))
+ right += 1
+ left = right
+ if left != right:
+ sil_left_n = min(self.max_silence, (right + self.win_ln - left) // 2)
+ rms_db_left = level2db(_window_rms(samples[left: left + sil_left_n], win_sz=self.win_sn))
+ split_win_l = left + np.argmin(rms_db_left)
+ split_loc_l = split_win_l + np.argmin(abs_amp[split_win_l: split_win_l + self.win_sn])
+ sil_tags.append((split_loc_l, samples.shape[0]))
+ if len(sil_tags) == 0:
+ return [audio]
+ else:
+ chunks = []
+ if sil_tags[0][0] > 0:
+ chunks.append(_apply_slice(audio, 0, sil_tags[0][0]))
+ for i in range(0, len(sil_tags) - 1):
+ chunks.append(_apply_slice(audio, sil_tags[i][1], sil_tags[i + 1][0]))
+ if sil_tags[-1][1] < samples.shape[0] - 1:
+ chunks.append(_apply_slice(audio, sil_tags[-1][1], samples.shape[0]))
+ return chunks
+
+
+def main():
+ parser = ArgumentParser()
+ parser.add_argument('audio', type=str, help='The audio to be sliced')
+ parser.add_argument('--out', type=str, help='Output directory of the sliced audio clips')
+ parser.add_argument('--db_thresh', type=float, required=False, default=-40, help='The dB threshold for silence detection')
+ parser.add_argument('--min_len', type=int, required=False, default=5000, help='The minimum milliseconds required for each sliced audio clip')
+ parser.add_argument('--win_l', type=int, required=False, default=300, help='Size of the large sliding window, presented in milliseconds')
+ parser.add_argument('--win_s', type=int, required=False, default=20, help='Size of the small sliding window, presented in milliseconds')
+ parser.add_argument('--max_sil_kept', type=int, required=False, default=500, help='The maximum silence length kept around the sliced audio, presented in milliseconds')
+ args = parser.parse_args()
+ out = args.out
+ if out is None:
+ out = os.path.dirname(os.path.abspath(args.audio))
+ audio, sr = librosa.load(args.audio, sr=None)
+ slicer = Slicer(
+ sr=sr,
+ db_threshold=args.db_thresh,
+ min_length=args.min_len,
+ win_l=args.win_l,
+ win_s=args.win_s,
+ max_silence_kept=args.max_sil_kept
+ )
+ chunks = slicer.slice(audio)
+ if not os.path.exists(args.out):
+ os.makedirs(args.out)
+ for i, chunk in enumerate(chunks):
+ soundfile.write(os.path.join(out, f'%s_%d.wav' % (os.path.basename(args.audio).rsplit('.', maxsplit=1)[0], i)), chunk, sr)
+
+
+if __name__ == '__main__':
+ main()
\ No newline at end of file
diff --git a/trainset_preprocess_pipeline.py b/trainset_preprocess_pipeline.py
new file mode 100644
index 0000000000000000000000000000000000000000..4fcead60ae652be4f7e07c0bcfa91e71089ab08a
--- /dev/null
+++ b/trainset_preprocess_pipeline.py
@@ -0,0 +1,63 @@
+import numpy as np,ffmpeg,os,traceback
+from slicer import Slicer
+slicer = Slicer(
+ sr=40000,
+ db_threshold=-32,
+ min_length=800,
+ win_l=400,
+ win_s=20,
+ max_silence_kept=150
+)
+
+
+
+
+def p0_load_audio(file, sr):#str-ing
+ try:
+ out, _ = (
+ ffmpeg.input(file, threads=0)
+ .output("-", format="s16le", acodec="pcm_s16le", ac=1, ar=sr)
+ .run(cmd=["ffmpeg", "-nostdin"], capture_stdout=True, capture_stderr=True)
+ )
+ except ffmpeg.Error as e:
+ raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e
+ return np.frombuffer(out, np.int16).flatten().astype(np.float32) / 32768.0
+
+def p1_trim_audio(slicer,audio):return slicer.slice(audio)
+
+def p2_avg_cut(audio,sr,per=3.7,overlap=0.3,tail=4):
+ i = 0
+ audios=[]
+ while (1):
+ start = int(sr * (per - overlap) * i)
+ i += 1
+ if (len(audio[start:]) > tail * sr):
+ audios.append(audio[start:start + int(per * sr)])
+ else:
+ audios.append(audio[start:])
+ break
+ return audios
+
+def p2b_get_vol(audio):return np.square(audio).mean()
+
+def p3_norm(audio,alpha=0.8,maxx=0.95):return audio / np.abs(audio).max() * (maxx * alpha) + (1-alpha) * audio
+
+def pipeline(inp_root,sr1=40000,sr2=16000,if_trim=True,if_avg_cut=True,if_norm=True,save_root1=None,save_root2=None):
+ if(save_root1==None and save_root2==None):return "No save root."
+ name2vol={}
+ infos=[]
+ names=[]
+ for name in os.listdir(inp_root):
+ try:
+ inp_path=os.path.join(inp_root,name)
+ audio=p0_load_audio(inp_path)
+ except:
+ infos.append("%s\t%s"%(name,traceback.format_exc()))
+ continue
+ if(if_trim==True):res1s=p1_trim_audio(audio)
+ else:res1s=[audio]
+ for i0,res1 in res1s:
+ if(if_avg_cut==True):res2=p2_avg_cut(res1)
+ else:res2=[res1]
+
+
diff --git a/uvr5_pack/__pycache__/utils.cpython-39.pyc b/uvr5_pack/__pycache__/utils.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..f7304c0fde6040dc7c209a536239347d9e0eb02c
Binary files /dev/null and b/uvr5_pack/__pycache__/utils.cpython-39.pyc differ
diff --git a/uvr5_pack/lib_v5/__pycache__/layers_123821KB.cpython-39.pyc b/uvr5_pack/lib_v5/__pycache__/layers_123821KB.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a8ec492b8145c9d7d3468988c8bc34c6ec6771c6
Binary files /dev/null and b/uvr5_pack/lib_v5/__pycache__/layers_123821KB.cpython-39.pyc differ
diff --git a/uvr5_pack/lib_v5/__pycache__/model_param_init.cpython-39.pyc b/uvr5_pack/lib_v5/__pycache__/model_param_init.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c2a87470350c935e3f093da5d31d90aafd1b7d74
Binary files /dev/null and b/uvr5_pack/lib_v5/__pycache__/model_param_init.cpython-39.pyc differ
diff --git a/uvr5_pack/lib_v5/__pycache__/nets_61968KB.cpython-39.pyc b/uvr5_pack/lib_v5/__pycache__/nets_61968KB.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..b13e91d3a4cda898f46504bc3a3b384f4c81a639
Binary files /dev/null and b/uvr5_pack/lib_v5/__pycache__/nets_61968KB.cpython-39.pyc differ
diff --git a/uvr5_pack/lib_v5/__pycache__/spec_utils.cpython-39.pyc b/uvr5_pack/lib_v5/__pycache__/spec_utils.cpython-39.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6e215f04cdf5819cfe690b10cf22b54a44bcdf01
Binary files /dev/null and b/uvr5_pack/lib_v5/__pycache__/spec_utils.cpython-39.pyc differ
diff --git a/uvr5_pack/lib_v5/dataset.py b/uvr5_pack/lib_v5/dataset.py
new file mode 100644
index 0000000000000000000000000000000000000000..59454aaa185ecee802f48bf1167752dfcb3aa6c3
--- /dev/null
+++ b/uvr5_pack/lib_v5/dataset.py
@@ -0,0 +1,170 @@
+import os
+import random
+
+import numpy as np
+import torch
+import torch.utils.data
+from tqdm import tqdm
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class VocalRemoverValidationSet(torch.utils.data.Dataset):
+
+ def __init__(self, patch_list):
+ self.patch_list = patch_list
+
+ def __len__(self):
+ return len(self.patch_list)
+
+ def __getitem__(self, idx):
+ path = self.patch_list[idx]
+ data = np.load(path)
+
+ X, y = data['X'], data['y']
+
+ X_mag = np.abs(X)
+ y_mag = np.abs(y)
+
+ return X_mag, y_mag
+
+
+def make_pair(mix_dir, inst_dir):
+ input_exts = ['.wav', '.m4a', '.mp3', '.mp4', '.flac']
+
+ X_list = sorted([
+ os.path.join(mix_dir, fname)
+ for fname in os.listdir(mix_dir)
+ if os.path.splitext(fname)[1] in input_exts])
+ y_list = sorted([
+ os.path.join(inst_dir, fname)
+ for fname in os.listdir(inst_dir)
+ if os.path.splitext(fname)[1] in input_exts])
+
+ filelist = list(zip(X_list, y_list))
+
+ return filelist
+
+
+def train_val_split(dataset_dir, split_mode, val_rate, val_filelist):
+ if split_mode == 'random':
+ filelist = make_pair(
+ os.path.join(dataset_dir, 'mixtures'),
+ os.path.join(dataset_dir, 'instruments'))
+
+ random.shuffle(filelist)
+
+ if len(val_filelist) == 0:
+ val_size = int(len(filelist) * val_rate)
+ train_filelist = filelist[:-val_size]
+ val_filelist = filelist[-val_size:]
+ else:
+ train_filelist = [
+ pair for pair in filelist
+ if list(pair) not in val_filelist]
+ elif split_mode == 'subdirs':
+ if len(val_filelist) != 0:
+ raise ValueError('The `val_filelist` option is not available in `subdirs` mode')
+
+ train_filelist = make_pair(
+ os.path.join(dataset_dir, 'training/mixtures'),
+ os.path.join(dataset_dir, 'training/instruments'))
+
+ val_filelist = make_pair(
+ os.path.join(dataset_dir, 'validation/mixtures'),
+ os.path.join(dataset_dir, 'validation/instruments'))
+
+ return train_filelist, val_filelist
+
+
+def augment(X, y, reduction_rate, reduction_mask, mixup_rate, mixup_alpha):
+ perm = np.random.permutation(len(X))
+ for i, idx in enumerate(tqdm(perm)):
+ if np.random.uniform() < reduction_rate:
+ y[idx] = spec_utils.reduce_vocal_aggressively(X[idx], y[idx], reduction_mask)
+
+ if np.random.uniform() < 0.5:
+ # swap channel
+ X[idx] = X[idx, ::-1]
+ y[idx] = y[idx, ::-1]
+ if np.random.uniform() < 0.02:
+ # mono
+ X[idx] = X[idx].mean(axis=0, keepdims=True)
+ y[idx] = y[idx].mean(axis=0, keepdims=True)
+ if np.random.uniform() < 0.02:
+ # inst
+ X[idx] = y[idx]
+
+ if np.random.uniform() < mixup_rate and i < len(perm) - 1:
+ lam = np.random.beta(mixup_alpha, mixup_alpha)
+ X[idx] = lam * X[idx] + (1 - lam) * X[perm[i + 1]]
+ y[idx] = lam * y[idx] + (1 - lam) * y[perm[i + 1]]
+
+ return X, y
+
+
+def make_padding(width, cropsize, offset):
+ left = offset
+ roi_size = cropsize - left * 2
+ if roi_size == 0:
+ roi_size = cropsize
+ right = roi_size - (width % roi_size) + left
+
+ return left, right, roi_size
+
+
+def make_training_set(filelist, cropsize, patches, sr, hop_length, n_fft, offset):
+ len_dataset = patches * len(filelist)
+
+ X_dataset = np.zeros(
+ (len_dataset, 2, n_fft // 2 + 1, cropsize), dtype=np.complex64)
+ y_dataset = np.zeros(
+ (len_dataset, 2, n_fft // 2 + 1, cropsize), dtype=np.complex64)
+
+ for i, (X_path, y_path) in enumerate(tqdm(filelist)):
+ X, y = spec_utils.cache_or_load(X_path, y_path, sr, hop_length, n_fft)
+ coef = np.max([np.abs(X).max(), np.abs(y).max()])
+ X, y = X / coef, y / coef
+
+ l, r, roi_size = make_padding(X.shape[2], cropsize, offset)
+ X_pad = np.pad(X, ((0, 0), (0, 0), (l, r)), mode='constant')
+ y_pad = np.pad(y, ((0, 0), (0, 0), (l, r)), mode='constant')
+
+ starts = np.random.randint(0, X_pad.shape[2] - cropsize, patches)
+ ends = starts + cropsize
+ for j in range(patches):
+ idx = i * patches + j
+ X_dataset[idx] = X_pad[:, :, starts[j]:ends[j]]
+ y_dataset[idx] = y_pad[:, :, starts[j]:ends[j]]
+
+ return X_dataset, y_dataset
+
+
+def make_validation_set(filelist, cropsize, sr, hop_length, n_fft, offset):
+ patch_list = []
+ patch_dir = 'cs{}_sr{}_hl{}_nf{}_of{}'.format(cropsize, sr, hop_length, n_fft, offset)
+ os.makedirs(patch_dir, exist_ok=True)
+
+ for i, (X_path, y_path) in enumerate(tqdm(filelist)):
+ basename = os.path.splitext(os.path.basename(X_path))[0]
+
+ X, y = spec_utils.cache_or_load(X_path, y_path, sr, hop_length, n_fft)
+ coef = np.max([np.abs(X).max(), np.abs(y).max()])
+ X, y = X / coef, y / coef
+
+ l, r, roi_size = make_padding(X.shape[2], cropsize, offset)
+ X_pad = np.pad(X, ((0, 0), (0, 0), (l, r)), mode='constant')
+ y_pad = np.pad(y, ((0, 0), (0, 0), (l, r)), mode='constant')
+
+ len_dataset = int(np.ceil(X.shape[2] / roi_size))
+ for j in range(len_dataset):
+ outpath = os.path.join(patch_dir, '{}_p{}.npz'.format(basename, j))
+ start = j * roi_size
+ if not os.path.exists(outpath):
+ np.savez(
+ outpath,
+ X=X_pad[:, :, start:start + cropsize],
+ y=y_pad[:, :, start:start + cropsize])
+ patch_list.append(outpath)
+
+ return VocalRemoverValidationSet(patch_list)
diff --git a/uvr5_pack/lib_v5/layers.py b/uvr5_pack/lib_v5/layers.py
new file mode 100644
index 0000000000000000000000000000000000000000..ca64106800f4ee3d250b23c9a77482764ebba80e
--- /dev/null
+++ b/uvr5_pack/lib_v5/layers.py
@@ -0,0 +1,116 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class Conv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(Conv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class SeperableConv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(SeperableConv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nin,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ groups=nin,
+ bias=False),
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=1,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class Encoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.LeakyReLU):
+ super(Encoder, self).__init__()
+ self.conv1 = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.conv2 = Conv2DBNActiv(nout, nout, ksize, stride, pad, activ=activ)
+
+ def __call__(self, x):
+ skip = self.conv1(x)
+ h = self.conv2(skip)
+
+ return h, skip
+
+
+class Decoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.ReLU, dropout=False):
+ super(Decoder, self).__init__()
+ self.conv = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.dropout = nn.Dropout2d(0.1) if dropout else None
+
+ def __call__(self, x, skip=None):
+ x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
+ if skip is not None:
+ skip = spec_utils.crop_center(skip, x)
+ x = torch.cat([x, skip], dim=1)
+ h = self.conv(x)
+
+ if self.dropout is not None:
+ h = self.dropout(h)
+
+ return h
+
+
+class ASPPModule(nn.Module):
+
+ def __init__(self, nin, nout, dilations=(4, 8, 16), activ=nn.ReLU):
+ super(ASPPModule, self).__init__()
+ self.conv1 = nn.Sequential(
+ nn.AdaptiveAvgPool2d((1, None)),
+ Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ )
+ self.conv2 = Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ self.conv3 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[0], dilations[0], activ=activ)
+ self.conv4 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[1], dilations[1], activ=activ)
+ self.conv5 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.bottleneck = nn.Sequential(
+ Conv2DBNActiv(nin * 5, nout, 1, 1, 0, activ=activ),
+ nn.Dropout2d(0.1)
+ )
+
+ def forward(self, x):
+ _, _, h, w = x.size()
+ feat1 = F.interpolate(self.conv1(x), size=(h, w), mode='bilinear', align_corners=True)
+ feat2 = self.conv2(x)
+ feat3 = self.conv3(x)
+ feat4 = self.conv4(x)
+ feat5 = self.conv5(x)
+ out = torch.cat((feat1, feat2, feat3, feat4, feat5), dim=1)
+ bottle = self.bottleneck(out)
+ return bottle
diff --git a/uvr5_pack/lib_v5/layers_123812KB .py b/uvr5_pack/lib_v5/layers_123812KB .py
new file mode 100644
index 0000000000000000000000000000000000000000..ca64106800f4ee3d250b23c9a77482764ebba80e
--- /dev/null
+++ b/uvr5_pack/lib_v5/layers_123812KB .py
@@ -0,0 +1,116 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class Conv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(Conv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class SeperableConv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(SeperableConv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nin,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ groups=nin,
+ bias=False),
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=1,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class Encoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.LeakyReLU):
+ super(Encoder, self).__init__()
+ self.conv1 = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.conv2 = Conv2DBNActiv(nout, nout, ksize, stride, pad, activ=activ)
+
+ def __call__(self, x):
+ skip = self.conv1(x)
+ h = self.conv2(skip)
+
+ return h, skip
+
+
+class Decoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.ReLU, dropout=False):
+ super(Decoder, self).__init__()
+ self.conv = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.dropout = nn.Dropout2d(0.1) if dropout else None
+
+ def __call__(self, x, skip=None):
+ x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
+ if skip is not None:
+ skip = spec_utils.crop_center(skip, x)
+ x = torch.cat([x, skip], dim=1)
+ h = self.conv(x)
+
+ if self.dropout is not None:
+ h = self.dropout(h)
+
+ return h
+
+
+class ASPPModule(nn.Module):
+
+ def __init__(self, nin, nout, dilations=(4, 8, 16), activ=nn.ReLU):
+ super(ASPPModule, self).__init__()
+ self.conv1 = nn.Sequential(
+ nn.AdaptiveAvgPool2d((1, None)),
+ Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ )
+ self.conv2 = Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ self.conv3 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[0], dilations[0], activ=activ)
+ self.conv4 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[1], dilations[1], activ=activ)
+ self.conv5 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.bottleneck = nn.Sequential(
+ Conv2DBNActiv(nin * 5, nout, 1, 1, 0, activ=activ),
+ nn.Dropout2d(0.1)
+ )
+
+ def forward(self, x):
+ _, _, h, w = x.size()
+ feat1 = F.interpolate(self.conv1(x), size=(h, w), mode='bilinear', align_corners=True)
+ feat2 = self.conv2(x)
+ feat3 = self.conv3(x)
+ feat4 = self.conv4(x)
+ feat5 = self.conv5(x)
+ out = torch.cat((feat1, feat2, feat3, feat4, feat5), dim=1)
+ bottle = self.bottleneck(out)
+ return bottle
diff --git a/uvr5_pack/lib_v5/layers_123821KB.py b/uvr5_pack/lib_v5/layers_123821KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..ca64106800f4ee3d250b23c9a77482764ebba80e
--- /dev/null
+++ b/uvr5_pack/lib_v5/layers_123821KB.py
@@ -0,0 +1,116 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class Conv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(Conv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class SeperableConv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(SeperableConv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nin,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ groups=nin,
+ bias=False),
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=1,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class Encoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.LeakyReLU):
+ super(Encoder, self).__init__()
+ self.conv1 = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.conv2 = Conv2DBNActiv(nout, nout, ksize, stride, pad, activ=activ)
+
+ def __call__(self, x):
+ skip = self.conv1(x)
+ h = self.conv2(skip)
+
+ return h, skip
+
+
+class Decoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.ReLU, dropout=False):
+ super(Decoder, self).__init__()
+ self.conv = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.dropout = nn.Dropout2d(0.1) if dropout else None
+
+ def __call__(self, x, skip=None):
+ x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
+ if skip is not None:
+ skip = spec_utils.crop_center(skip, x)
+ x = torch.cat([x, skip], dim=1)
+ h = self.conv(x)
+
+ if self.dropout is not None:
+ h = self.dropout(h)
+
+ return h
+
+
+class ASPPModule(nn.Module):
+
+ def __init__(self, nin, nout, dilations=(4, 8, 16), activ=nn.ReLU):
+ super(ASPPModule, self).__init__()
+ self.conv1 = nn.Sequential(
+ nn.AdaptiveAvgPool2d((1, None)),
+ Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ )
+ self.conv2 = Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ self.conv3 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[0], dilations[0], activ=activ)
+ self.conv4 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[1], dilations[1], activ=activ)
+ self.conv5 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.bottleneck = nn.Sequential(
+ Conv2DBNActiv(nin * 5, nout, 1, 1, 0, activ=activ),
+ nn.Dropout2d(0.1)
+ )
+
+ def forward(self, x):
+ _, _, h, w = x.size()
+ feat1 = F.interpolate(self.conv1(x), size=(h, w), mode='bilinear', align_corners=True)
+ feat2 = self.conv2(x)
+ feat3 = self.conv3(x)
+ feat4 = self.conv4(x)
+ feat5 = self.conv5(x)
+ out = torch.cat((feat1, feat2, feat3, feat4, feat5), dim=1)
+ bottle = self.bottleneck(out)
+ return bottle
diff --git a/uvr5_pack/lib_v5/layers_33966KB.py b/uvr5_pack/lib_v5/layers_33966KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..0262e002c9f613636ed3401646613ed57c574d7e
--- /dev/null
+++ b/uvr5_pack/lib_v5/layers_33966KB.py
@@ -0,0 +1,122 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class Conv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(Conv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class SeperableConv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(SeperableConv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nin,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ groups=nin,
+ bias=False),
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=1,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class Encoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.LeakyReLU):
+ super(Encoder, self).__init__()
+ self.conv1 = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.conv2 = Conv2DBNActiv(nout, nout, ksize, stride, pad, activ=activ)
+
+ def __call__(self, x):
+ skip = self.conv1(x)
+ h = self.conv2(skip)
+
+ return h, skip
+
+
+class Decoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.ReLU, dropout=False):
+ super(Decoder, self).__init__()
+ self.conv = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.dropout = nn.Dropout2d(0.1) if dropout else None
+
+ def __call__(self, x, skip=None):
+ x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
+ if skip is not None:
+ skip = spec_utils.crop_center(skip, x)
+ x = torch.cat([x, skip], dim=1)
+ h = self.conv(x)
+
+ if self.dropout is not None:
+ h = self.dropout(h)
+
+ return h
+
+
+class ASPPModule(nn.Module):
+
+ def __init__(self, nin, nout, dilations=(4, 8, 16, 32, 64), activ=nn.ReLU):
+ super(ASPPModule, self).__init__()
+ self.conv1 = nn.Sequential(
+ nn.AdaptiveAvgPool2d((1, None)),
+ Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ )
+ self.conv2 = Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ self.conv3 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[0], dilations[0], activ=activ)
+ self.conv4 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[1], dilations[1], activ=activ)
+ self.conv5 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.conv6 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.conv7 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.bottleneck = nn.Sequential(
+ Conv2DBNActiv(nin * 7, nout, 1, 1, 0, activ=activ),
+ nn.Dropout2d(0.1)
+ )
+
+ def forward(self, x):
+ _, _, h, w = x.size()
+ feat1 = F.interpolate(self.conv1(x), size=(h, w), mode='bilinear', align_corners=True)
+ feat2 = self.conv2(x)
+ feat3 = self.conv3(x)
+ feat4 = self.conv4(x)
+ feat5 = self.conv5(x)
+ feat6 = self.conv6(x)
+ feat7 = self.conv7(x)
+ out = torch.cat((feat1, feat2, feat3, feat4, feat5, feat6, feat7), dim=1)
+ bottle = self.bottleneck(out)
+ return bottle
diff --git a/uvr5_pack/lib_v5/layers_537227KB.py b/uvr5_pack/lib_v5/layers_537227KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..0262e002c9f613636ed3401646613ed57c574d7e
--- /dev/null
+++ b/uvr5_pack/lib_v5/layers_537227KB.py
@@ -0,0 +1,122 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class Conv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(Conv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class SeperableConv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(SeperableConv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nin,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ groups=nin,
+ bias=False),
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=1,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class Encoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.LeakyReLU):
+ super(Encoder, self).__init__()
+ self.conv1 = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.conv2 = Conv2DBNActiv(nout, nout, ksize, stride, pad, activ=activ)
+
+ def __call__(self, x):
+ skip = self.conv1(x)
+ h = self.conv2(skip)
+
+ return h, skip
+
+
+class Decoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.ReLU, dropout=False):
+ super(Decoder, self).__init__()
+ self.conv = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.dropout = nn.Dropout2d(0.1) if dropout else None
+
+ def __call__(self, x, skip=None):
+ x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
+ if skip is not None:
+ skip = spec_utils.crop_center(skip, x)
+ x = torch.cat([x, skip], dim=1)
+ h = self.conv(x)
+
+ if self.dropout is not None:
+ h = self.dropout(h)
+
+ return h
+
+
+class ASPPModule(nn.Module):
+
+ def __init__(self, nin, nout, dilations=(4, 8, 16, 32, 64), activ=nn.ReLU):
+ super(ASPPModule, self).__init__()
+ self.conv1 = nn.Sequential(
+ nn.AdaptiveAvgPool2d((1, None)),
+ Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ )
+ self.conv2 = Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ self.conv3 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[0], dilations[0], activ=activ)
+ self.conv4 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[1], dilations[1], activ=activ)
+ self.conv5 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.conv6 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.conv7 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.bottleneck = nn.Sequential(
+ Conv2DBNActiv(nin * 7, nout, 1, 1, 0, activ=activ),
+ nn.Dropout2d(0.1)
+ )
+
+ def forward(self, x):
+ _, _, h, w = x.size()
+ feat1 = F.interpolate(self.conv1(x), size=(h, w), mode='bilinear', align_corners=True)
+ feat2 = self.conv2(x)
+ feat3 = self.conv3(x)
+ feat4 = self.conv4(x)
+ feat5 = self.conv5(x)
+ feat6 = self.conv6(x)
+ feat7 = self.conv7(x)
+ out = torch.cat((feat1, feat2, feat3, feat4, feat5, feat6, feat7), dim=1)
+ bottle = self.bottleneck(out)
+ return bottle
diff --git a/uvr5_pack/lib_v5/layers_537238KB.py b/uvr5_pack/lib_v5/layers_537238KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..0262e002c9f613636ed3401646613ed57c574d7e
--- /dev/null
+++ b/uvr5_pack/lib_v5/layers_537238KB.py
@@ -0,0 +1,122 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class Conv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(Conv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class SeperableConv2DBNActiv(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, dilation=1, activ=nn.ReLU):
+ super(SeperableConv2DBNActiv, self).__init__()
+ self.conv = nn.Sequential(
+ nn.Conv2d(
+ nin, nin,
+ kernel_size=ksize,
+ stride=stride,
+ padding=pad,
+ dilation=dilation,
+ groups=nin,
+ bias=False),
+ nn.Conv2d(
+ nin, nout,
+ kernel_size=1,
+ bias=False),
+ nn.BatchNorm2d(nout),
+ activ()
+ )
+
+ def __call__(self, x):
+ return self.conv(x)
+
+
+class Encoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.LeakyReLU):
+ super(Encoder, self).__init__()
+ self.conv1 = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.conv2 = Conv2DBNActiv(nout, nout, ksize, stride, pad, activ=activ)
+
+ def __call__(self, x):
+ skip = self.conv1(x)
+ h = self.conv2(skip)
+
+ return h, skip
+
+
+class Decoder(nn.Module):
+
+ def __init__(self, nin, nout, ksize=3, stride=1, pad=1, activ=nn.ReLU, dropout=False):
+ super(Decoder, self).__init__()
+ self.conv = Conv2DBNActiv(nin, nout, ksize, 1, pad, activ=activ)
+ self.dropout = nn.Dropout2d(0.1) if dropout else None
+
+ def __call__(self, x, skip=None):
+ x = F.interpolate(x, scale_factor=2, mode='bilinear', align_corners=True)
+ if skip is not None:
+ skip = spec_utils.crop_center(skip, x)
+ x = torch.cat([x, skip], dim=1)
+ h = self.conv(x)
+
+ if self.dropout is not None:
+ h = self.dropout(h)
+
+ return h
+
+
+class ASPPModule(nn.Module):
+
+ def __init__(self, nin, nout, dilations=(4, 8, 16, 32, 64), activ=nn.ReLU):
+ super(ASPPModule, self).__init__()
+ self.conv1 = nn.Sequential(
+ nn.AdaptiveAvgPool2d((1, None)),
+ Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ )
+ self.conv2 = Conv2DBNActiv(nin, nin, 1, 1, 0, activ=activ)
+ self.conv3 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[0], dilations[0], activ=activ)
+ self.conv4 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[1], dilations[1], activ=activ)
+ self.conv5 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.conv6 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.conv7 = SeperableConv2DBNActiv(
+ nin, nin, 3, 1, dilations[2], dilations[2], activ=activ)
+ self.bottleneck = nn.Sequential(
+ Conv2DBNActiv(nin * 7, nout, 1, 1, 0, activ=activ),
+ nn.Dropout2d(0.1)
+ )
+
+ def forward(self, x):
+ _, _, h, w = x.size()
+ feat1 = F.interpolate(self.conv1(x), size=(h, w), mode='bilinear', align_corners=True)
+ feat2 = self.conv2(x)
+ feat3 = self.conv3(x)
+ feat4 = self.conv4(x)
+ feat5 = self.conv5(x)
+ feat6 = self.conv6(x)
+ feat7 = self.conv7(x)
+ out = torch.cat((feat1, feat2, feat3, feat4, feat5, feat6, feat7), dim=1)
+ bottle = self.bottleneck(out)
+ return bottle
diff --git a/uvr5_pack/lib_v5/model_param_init.py b/uvr5_pack/lib_v5/model_param_init.py
new file mode 100644
index 0000000000000000000000000000000000000000..514294d64936a2774aeead01e2afeb4151d7a1d3
--- /dev/null
+++ b/uvr5_pack/lib_v5/model_param_init.py
@@ -0,0 +1,60 @@
+import json
+import os
+import pathlib
+
+default_param = {}
+default_param['bins'] = 768
+default_param['unstable_bins'] = 9 # training only
+default_param['reduction_bins'] = 762 # training only
+default_param['sr'] = 44100
+default_param['pre_filter_start'] = 757
+default_param['pre_filter_stop'] = 768
+default_param['band'] = {}
+
+
+default_param['band'][1] = {
+ 'sr': 11025,
+ 'hl': 128,
+ 'n_fft': 960,
+ 'crop_start': 0,
+ 'crop_stop': 245,
+ 'lpf_start': 61, # inference only
+ 'res_type': 'polyphase'
+}
+
+default_param['band'][2] = {
+ 'sr': 44100,
+ 'hl': 512,
+ 'n_fft': 1536,
+ 'crop_start': 24,
+ 'crop_stop': 547,
+ 'hpf_start': 81, # inference only
+ 'res_type': 'sinc_best'
+}
+
+
+def int_keys(d):
+ r = {}
+ for k, v in d:
+ if k.isdigit():
+ k = int(k)
+ r[k] = v
+ return r
+
+
+class ModelParameters(object):
+ def __init__(self, config_path=''):
+ if '.pth' == pathlib.Path(config_path).suffix:
+ import zipfile
+
+ with zipfile.ZipFile(config_path, 'r') as zip:
+ self.param = json.loads(zip.read('param.json'), object_pairs_hook=int_keys)
+ elif '.json' == pathlib.Path(config_path).suffix:
+ with open(config_path, 'r') as f:
+ self.param = json.loads(f.read(), object_pairs_hook=int_keys)
+ else:
+ self.param = default_param
+
+ for k in ['mid_side', 'mid_side_b', 'mid_side_b2', 'stereo_w', 'stereo_n', 'reverse']:
+ if not k in self.param:
+ self.param[k] = False
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr16000_hl512.json b/uvr5_pack/lib_v5/modelparams/1band_sr16000_hl512.json
new file mode 100644
index 0000000000000000000000000000000000000000..72cb4499867ad2827185e85687f06fb73d33eced
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr16000_hl512.json
@@ -0,0 +1,19 @@
+{
+ "bins": 1024,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 16000,
+ "hl": 512,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 1024,
+ "hpf_start": -1,
+ "res_type": "sinc_best"
+ }
+ },
+ "sr": 16000,
+ "pre_filter_start": 1023,
+ "pre_filter_stop": 1024
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json b/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c00ecf0a105e55a6a86a3c32db301a2635b5b41
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json
@@ -0,0 +1,19 @@
+{
+ "bins": 1024,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 32000,
+ "hl": 512,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 1024,
+ "hpf_start": -1,
+ "res_type": "kaiser_fast"
+ }
+ },
+ "sr": 32000,
+ "pre_filter_start": 1000,
+ "pre_filter_stop": 1021
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr33075_hl384.json b/uvr5_pack/lib_v5/modelparams/1band_sr33075_hl384.json
new file mode 100644
index 0000000000000000000000000000000000000000..55666ac9a8d0547751fb4b4d3bffb1ee2c956913
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr33075_hl384.json
@@ -0,0 +1,19 @@
+{
+ "bins": 1024,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 33075,
+ "hl": 384,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 1024,
+ "hpf_start": -1,
+ "res_type": "sinc_best"
+ }
+ },
+ "sr": 33075,
+ "pre_filter_start": 1000,
+ "pre_filter_stop": 1021
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl1024.json b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl1024.json
new file mode 100644
index 0000000000000000000000000000000000000000..665abe20eb3cc39fe0f8493dad8f25f6ef634a14
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl1024.json
@@ -0,0 +1,19 @@
+{
+ "bins": 1024,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 44100,
+ "hl": 1024,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 1024,
+ "hpf_start": -1,
+ "res_type": "sinc_best"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 1023,
+ "pre_filter_stop": 1024
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl256.json b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl256.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e8b16f89b0231d06eabe8d2f7c2670c7caa2272
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl256.json
@@ -0,0 +1,19 @@
+{
+ "bins": 256,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 44100,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 0,
+ "crop_stop": 256,
+ "hpf_start": -1,
+ "res_type": "sinc_best"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 256,
+ "pre_filter_stop": 256
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512.json b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b38fcaf60ba204e03a47f5bd3f5bcfe75e1983a
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512.json
@@ -0,0 +1,19 @@
+{
+ "bins": 1024,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 1024,
+ "hpf_start": -1,
+ "res_type": "sinc_best"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 1023,
+ "pre_filter_stop": 1024
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512_cut.json b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512_cut.json
new file mode 100644
index 0000000000000000000000000000000000000000..630df3524e340f43a1ddb7b33ff02cc91fc1cb47
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512_cut.json
@@ -0,0 +1,19 @@
+{
+ "bins": 1024,
+ "unstable_bins": 0,
+ "reduction_bins": 0,
+ "band": {
+ "1": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 700,
+ "hpf_start": -1,
+ "res_type": "sinc_best"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 1023,
+ "pre_filter_stop": 700
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/2band_32000.json b/uvr5_pack/lib_v5/modelparams/2band_32000.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab9cf1150a818eb6252105408311be0a40d423b3
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/2band_32000.json
@@ -0,0 +1,30 @@
+{
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 705,
+ "band": {
+ "1": {
+ "sr": 6000,
+ "hl": 66,
+ "n_fft": 512,
+ "crop_start": 0,
+ "crop_stop": 240,
+ "lpf_start": 60,
+ "lpf_stop": 118,
+ "res_type": "sinc_fastest"
+ },
+ "2": {
+ "sr": 32000,
+ "hl": 352,
+ "n_fft": 1024,
+ "crop_start": 22,
+ "crop_stop": 505,
+ "hpf_start": 44,
+ "hpf_stop": 23,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 32000,
+ "pre_filter_start": 710,
+ "pre_filter_stop": 731
+}
diff --git a/uvr5_pack/lib_v5/modelparams/2band_44100_lofi.json b/uvr5_pack/lib_v5/modelparams/2band_44100_lofi.json
new file mode 100644
index 0000000000000000000000000000000000000000..7faa216d7b49aeece24123dbdd868847a1dbc03c
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/2band_44100_lofi.json
@@ -0,0 +1,30 @@
+{
+ "bins": 512,
+ "unstable_bins": 7,
+ "reduction_bins": 510,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 160,
+ "n_fft": 768,
+ "crop_start": 0,
+ "crop_stop": 192,
+ "lpf_start": 41,
+ "lpf_stop": 139,
+ "res_type": "sinc_fastest"
+ },
+ "2": {
+ "sr": 44100,
+ "hl": 640,
+ "n_fft": 1024,
+ "crop_start": 10,
+ "crop_stop": 320,
+ "hpf_start": 47,
+ "hpf_stop": 15,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 510,
+ "pre_filter_stop": 512
+}
diff --git a/uvr5_pack/lib_v5/modelparams/2band_48000.json b/uvr5_pack/lib_v5/modelparams/2band_48000.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e78175052b09cb1a32345e54006475992712f9a
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/2band_48000.json
@@ -0,0 +1,30 @@
+{
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 705,
+ "band": {
+ "1": {
+ "sr": 6000,
+ "hl": 66,
+ "n_fft": 512,
+ "crop_start": 0,
+ "crop_stop": 240,
+ "lpf_start": 60,
+ "lpf_stop": 240,
+ "res_type": "sinc_fastest"
+ },
+ "2": {
+ "sr": 48000,
+ "hl": 528,
+ "n_fft": 1536,
+ "crop_start": 22,
+ "crop_stop": 505,
+ "hpf_start": 82,
+ "hpf_stop": 22,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 48000,
+ "pre_filter_start": 710,
+ "pre_filter_stop": 731
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/3band_44100.json b/uvr5_pack/lib_v5/modelparams/3band_44100.json
new file mode 100644
index 0000000000000000000000000000000000000000..d881d767ff83fbac0e18dfe2587ef16925b29b3c
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/3band_44100.json
@@ -0,0 +1,42 @@
+{
+ "bins": 768,
+ "unstable_bins": 5,
+ "reduction_bins": 733,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 768,
+ "crop_start": 0,
+ "crop_stop": 278,
+ "lpf_start": 28,
+ "lpf_stop": 140,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 768,
+ "crop_start": 14,
+ "crop_stop": 322,
+ "hpf_start": 70,
+ "hpf_stop": 14,
+ "lpf_start": 283,
+ "lpf_stop": 314,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 131,
+ "crop_stop": 313,
+ "hpf_start": 154,
+ "hpf_stop": 141,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 757,
+ "pre_filter_stop": 768
+}
diff --git a/uvr5_pack/lib_v5/modelparams/3band_44100_mid.json b/uvr5_pack/lib_v5/modelparams/3band_44100_mid.json
new file mode 100644
index 0000000000000000000000000000000000000000..77ec198573b19f36519a028a509767d30764c0e2
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/3band_44100_mid.json
@@ -0,0 +1,43 @@
+{
+ "mid_side": true,
+ "bins": 768,
+ "unstable_bins": 5,
+ "reduction_bins": 733,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 768,
+ "crop_start": 0,
+ "crop_stop": 278,
+ "lpf_start": 28,
+ "lpf_stop": 140,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 768,
+ "crop_start": 14,
+ "crop_stop": 322,
+ "hpf_start": 70,
+ "hpf_stop": 14,
+ "lpf_start": 283,
+ "lpf_stop": 314,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 131,
+ "crop_stop": 313,
+ "hpf_start": 154,
+ "hpf_stop": 141,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 757,
+ "pre_filter_stop": 768
+}
diff --git a/uvr5_pack/lib_v5/modelparams/3band_44100_msb2.json b/uvr5_pack/lib_v5/modelparams/3band_44100_msb2.json
new file mode 100644
index 0000000000000000000000000000000000000000..85ee8a7d44541c9176e85ea3dce8728d34990938
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/3band_44100_msb2.json
@@ -0,0 +1,43 @@
+{
+ "mid_side_b2": true,
+ "bins": 640,
+ "unstable_bins": 7,
+ "reduction_bins": 565,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 108,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 187,
+ "lpf_start": 92,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 22050,
+ "hl": 216,
+ "n_fft": 768,
+ "crop_start": 0,
+ "crop_stop": 212,
+ "hpf_start": 68,
+ "hpf_stop": 34,
+ "lpf_start": 174,
+ "lpf_stop": 209,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 44100,
+ "hl": 432,
+ "n_fft": 640,
+ "crop_start": 66,
+ "crop_stop": 307,
+ "hpf_start": 86,
+ "hpf_stop": 72,
+ "res_type": "kaiser_fast"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 639,
+ "pre_filter_stop": 640
+}
diff --git a/uvr5_pack/lib_v5/modelparams/4band_44100.json b/uvr5_pack/lib_v5/modelparams/4band_44100.json
new file mode 100644
index 0000000000000000000000000000000000000000..df123754204372aa50d464fbe9102a401f48cc73
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_44100.json
@@ -0,0 +1,54 @@
+{
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 668,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 186,
+ "lpf_start": 37,
+ "lpf_stop": 73,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 512,
+ "crop_start": 4,
+ "crop_stop": 185,
+ "hpf_start": 36,
+ "hpf_stop": 18,
+ "lpf_start": 93,
+ "lpf_stop": 185,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 46,
+ "crop_stop": 186,
+ "hpf_start": 93,
+ "hpf_stop": 46,
+ "lpf_start": 164,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 121,
+ "crop_stop": 382,
+ "hpf_start": 138,
+ "hpf_stop": 123,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 740,
+ "pre_filter_stop": 768
+}
diff --git a/uvr5_pack/lib_v5/modelparams/4band_44100_mid.json b/uvr5_pack/lib_v5/modelparams/4band_44100_mid.json
new file mode 100644
index 0000000000000000000000000000000000000000..e91b699eb63d3382c3b9e9edf46d40ed91d6122b
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_44100_mid.json
@@ -0,0 +1,55 @@
+{
+ "bins": 768,
+ "unstable_bins": 7,
+ "mid_side": true,
+ "reduction_bins": 668,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 186,
+ "lpf_start": 37,
+ "lpf_stop": 73,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 512,
+ "crop_start": 4,
+ "crop_stop": 185,
+ "hpf_start": 36,
+ "hpf_stop": 18,
+ "lpf_start": 93,
+ "lpf_stop": 185,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 46,
+ "crop_stop": 186,
+ "hpf_start": 93,
+ "hpf_stop": 46,
+ "lpf_start": 164,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 121,
+ "crop_stop": 382,
+ "hpf_start": 138,
+ "hpf_stop": 123,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 740,
+ "pre_filter_stop": 768
+}
diff --git a/uvr5_pack/lib_v5/modelparams/4band_44100_msb.json b/uvr5_pack/lib_v5/modelparams/4band_44100_msb.json
new file mode 100644
index 0000000000000000000000000000000000000000..f852f280ec9d98fc1b65cec688290eaafec61b84
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_44100_msb.json
@@ -0,0 +1,55 @@
+{
+ "mid_side_b": true,
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 668,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 186,
+ "lpf_start": 37,
+ "lpf_stop": 73,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 512,
+ "crop_start": 4,
+ "crop_stop": 185,
+ "hpf_start": 36,
+ "hpf_stop": 18,
+ "lpf_start": 93,
+ "lpf_stop": 185,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 46,
+ "crop_stop": 186,
+ "hpf_start": 93,
+ "hpf_stop": 46,
+ "lpf_start": 164,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 121,
+ "crop_stop": 382,
+ "hpf_start": 138,
+ "hpf_stop": 123,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 740,
+ "pre_filter_stop": 768
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/4band_44100_msb2.json b/uvr5_pack/lib_v5/modelparams/4band_44100_msb2.json
new file mode 100644
index 0000000000000000000000000000000000000000..f852f280ec9d98fc1b65cec688290eaafec61b84
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_44100_msb2.json
@@ -0,0 +1,55 @@
+{
+ "mid_side_b": true,
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 668,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 186,
+ "lpf_start": 37,
+ "lpf_stop": 73,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 512,
+ "crop_start": 4,
+ "crop_stop": 185,
+ "hpf_start": 36,
+ "hpf_stop": 18,
+ "lpf_start": 93,
+ "lpf_stop": 185,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 46,
+ "crop_stop": 186,
+ "hpf_start": 93,
+ "hpf_stop": 46,
+ "lpf_start": 164,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 121,
+ "crop_stop": 382,
+ "hpf_start": 138,
+ "hpf_stop": 123,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 740,
+ "pre_filter_stop": 768
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/4band_44100_reverse.json b/uvr5_pack/lib_v5/modelparams/4band_44100_reverse.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a07d5541bd83dc1caa20b531c3b43a2ffccac88
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_44100_reverse.json
@@ -0,0 +1,55 @@
+{
+ "reverse": true,
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 668,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 186,
+ "lpf_start": 37,
+ "lpf_stop": 73,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 512,
+ "crop_start": 4,
+ "crop_stop": 185,
+ "hpf_start": 36,
+ "hpf_stop": 18,
+ "lpf_start": 93,
+ "lpf_stop": 185,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 46,
+ "crop_stop": 186,
+ "hpf_start": 93,
+ "hpf_stop": 46,
+ "lpf_start": 164,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 121,
+ "crop_stop": 382,
+ "hpf_start": 138,
+ "hpf_stop": 123,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 740,
+ "pre_filter_stop": 768
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/4band_44100_sw.json b/uvr5_pack/lib_v5/modelparams/4band_44100_sw.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba0cf342106de793e6ec3e876854c7fd451fbf76
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_44100_sw.json
@@ -0,0 +1,55 @@
+{
+ "stereo_w": true,
+ "bins": 768,
+ "unstable_bins": 7,
+ "reduction_bins": 668,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 1024,
+ "crop_start": 0,
+ "crop_stop": 186,
+ "lpf_start": 37,
+ "lpf_stop": 73,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 11025,
+ "hl": 128,
+ "n_fft": 512,
+ "crop_start": 4,
+ "crop_stop": 185,
+ "hpf_start": 36,
+ "hpf_stop": 18,
+ "lpf_start": 93,
+ "lpf_stop": 185,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 22050,
+ "hl": 256,
+ "n_fft": 512,
+ "crop_start": 46,
+ "crop_stop": 186,
+ "hpf_start": 93,
+ "hpf_stop": 46,
+ "lpf_start": 164,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 512,
+ "n_fft": 768,
+ "crop_start": 121,
+ "crop_stop": 382,
+ "hpf_start": 138,
+ "hpf_stop": 123,
+ "res_type": "sinc_medium"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 740,
+ "pre_filter_stop": 768
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/4band_v2.json b/uvr5_pack/lib_v5/modelparams/4band_v2.json
new file mode 100644
index 0000000000000000000000000000000000000000..33281a0cf9916fc33558ddfda7a0287a2547faf4
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_v2.json
@@ -0,0 +1,54 @@
+{
+ "bins": 672,
+ "unstable_bins": 8,
+ "reduction_bins": 637,
+ "band": {
+ "1": {
+ "sr": 7350,
+ "hl": 80,
+ "n_fft": 640,
+ "crop_start": 0,
+ "crop_stop": 85,
+ "lpf_start": 25,
+ "lpf_stop": 53,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 7350,
+ "hl": 80,
+ "n_fft": 320,
+ "crop_start": 4,
+ "crop_stop": 87,
+ "hpf_start": 25,
+ "hpf_stop": 12,
+ "lpf_start": 31,
+ "lpf_stop": 62,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 14700,
+ "hl": 160,
+ "n_fft": 512,
+ "crop_start": 17,
+ "crop_stop": 216,
+ "hpf_start": 48,
+ "hpf_stop": 24,
+ "lpf_start": 139,
+ "lpf_stop": 210,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 480,
+ "n_fft": 960,
+ "crop_start": 78,
+ "crop_stop": 383,
+ "hpf_start": 130,
+ "hpf_stop": 86,
+ "res_type": "kaiser_fast"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 668,
+ "pre_filter_stop": 672
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/4band_v2_sn.json b/uvr5_pack/lib_v5/modelparams/4band_v2_sn.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e5c770fe188779bf6b0873190b7a324d6a867b2
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/4band_v2_sn.json
@@ -0,0 +1,55 @@
+{
+ "bins": 672,
+ "unstable_bins": 8,
+ "reduction_bins": 637,
+ "band": {
+ "1": {
+ "sr": 7350,
+ "hl": 80,
+ "n_fft": 640,
+ "crop_start": 0,
+ "crop_stop": 85,
+ "lpf_start": 25,
+ "lpf_stop": 53,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 7350,
+ "hl": 80,
+ "n_fft": 320,
+ "crop_start": 4,
+ "crop_stop": 87,
+ "hpf_start": 25,
+ "hpf_stop": 12,
+ "lpf_start": 31,
+ "lpf_stop": 62,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 14700,
+ "hl": 160,
+ "n_fft": 512,
+ "crop_start": 17,
+ "crop_stop": 216,
+ "hpf_start": 48,
+ "hpf_stop": 24,
+ "lpf_start": 139,
+ "lpf_stop": 210,
+ "res_type": "polyphase"
+ },
+ "4": {
+ "sr": 44100,
+ "hl": 480,
+ "n_fft": 960,
+ "crop_start": 78,
+ "crop_stop": 383,
+ "hpf_start": 130,
+ "hpf_stop": 86,
+ "convert_channels": "stereo_n",
+ "res_type": "kaiser_fast"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 668,
+ "pre_filter_stop": 672
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/modelparams/ensemble.json b/uvr5_pack/lib_v5/modelparams/ensemble.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee69beb46fc82f34619c5e48761e329fcabbbd00
--- /dev/null
+++ b/uvr5_pack/lib_v5/modelparams/ensemble.json
@@ -0,0 +1,43 @@
+{
+ "mid_side_b2": true,
+ "bins": 1280,
+ "unstable_bins": 7,
+ "reduction_bins": 565,
+ "band": {
+ "1": {
+ "sr": 11025,
+ "hl": 108,
+ "n_fft": 2048,
+ "crop_start": 0,
+ "crop_stop": 374,
+ "lpf_start": 92,
+ "lpf_stop": 186,
+ "res_type": "polyphase"
+ },
+ "2": {
+ "sr": 22050,
+ "hl": 216,
+ "n_fft": 1536,
+ "crop_start": 0,
+ "crop_stop": 424,
+ "hpf_start": 68,
+ "hpf_stop": 34,
+ "lpf_start": 348,
+ "lpf_stop": 418,
+ "res_type": "polyphase"
+ },
+ "3": {
+ "sr": 44100,
+ "hl": 432,
+ "n_fft": 1280,
+ "crop_start": 132,
+ "crop_stop": 614,
+ "hpf_start": 172,
+ "hpf_stop": 144,
+ "res_type": "polyphase"
+ }
+ },
+ "sr": 44100,
+ "pre_filter_start": 1280,
+ "pre_filter_stop": 1280
+}
\ No newline at end of file
diff --git a/uvr5_pack/lib_v5/nets.py b/uvr5_pack/lib_v5/nets.py
new file mode 100644
index 0000000000000000000000000000000000000000..70de59ad093872d4004a91af9de75a3cba2b2e81
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets.py
@@ -0,0 +1,113 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers
+from uvr5_pack.lib_v5 import spec_utils
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 16)
+ self.stg1_high_band_net = BaseASPPNet(2, 16)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(18, 8, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(8, 16)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(34, 16, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(16, 32)
+
+ self.out = nn.Conv2d(32, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(16, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(16, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/nets_123812KB.py b/uvr5_pack/lib_v5/nets_123812KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..957c8e179331019e75901c269921ddfcc4fbda5c
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets_123812KB.py
@@ -0,0 +1,112 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers_123821KB as layers
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 32)
+ self.stg1_high_band_net = BaseASPPNet(2, 32)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(34, 16, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(16, 32)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(66, 32, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(32, 64)
+
+ self.out = nn.Conv2d(64, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(32, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(32, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/nets_123821KB.py b/uvr5_pack/lib_v5/nets_123821KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..957c8e179331019e75901c269921ddfcc4fbda5c
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets_123821KB.py
@@ -0,0 +1,112 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers_123821KB as layers
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 32)
+ self.stg1_high_band_net = BaseASPPNet(2, 32)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(34, 16, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(16, 32)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(66, 32, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(32, 64)
+
+ self.out = nn.Conv2d(64, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(32, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(32, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/nets_33966KB.py b/uvr5_pack/lib_v5/nets_33966KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..7cc8262c7c6d404f6b7702a3540c3382e41f50c3
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets_33966KB.py
@@ -0,0 +1,112 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers_33966KB as layers
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16, 32)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 16)
+ self.stg1_high_band_net = BaseASPPNet(2, 16)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(18, 8, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(8, 16)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(34, 16, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(16, 32)
+
+ self.out = nn.Conv2d(32, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(16, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(16, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/nets_537227KB.py b/uvr5_pack/lib_v5/nets_537227KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..3d8006bbffb4186855234acc30fc2108b8544b4e
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets_537227KB.py
@@ -0,0 +1,113 @@
+import torch
+import numpy as np
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers_537238KB as layers
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 64)
+ self.stg1_high_band_net = BaseASPPNet(2, 64)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(66, 32, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(32, 64)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(130, 64, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(64, 128)
+
+ self.out = nn.Conv2d(128, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(64, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(64, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/nets_537238KB.py b/uvr5_pack/lib_v5/nets_537238KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..3d8006bbffb4186855234acc30fc2108b8544b4e
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets_537238KB.py
@@ -0,0 +1,113 @@
+import torch
+import numpy as np
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers_537238KB as layers
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 64)
+ self.stg1_high_band_net = BaseASPPNet(2, 64)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(66, 32, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(32, 64)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(130, 64, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(64, 128)
+
+ self.out = nn.Conv2d(128, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(64, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(64, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/nets_61968KB.py b/uvr5_pack/lib_v5/nets_61968KB.py
new file mode 100644
index 0000000000000000000000000000000000000000..957c8e179331019e75901c269921ddfcc4fbda5c
--- /dev/null
+++ b/uvr5_pack/lib_v5/nets_61968KB.py
@@ -0,0 +1,112 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+from uvr5_pack.lib_v5 import layers_123821KB as layers
+
+
+class BaseASPPNet(nn.Module):
+
+ def __init__(self, nin, ch, dilations=(4, 8, 16)):
+ super(BaseASPPNet, self).__init__()
+ self.enc1 = layers.Encoder(nin, ch, 3, 2, 1)
+ self.enc2 = layers.Encoder(ch, ch * 2, 3, 2, 1)
+ self.enc3 = layers.Encoder(ch * 2, ch * 4, 3, 2, 1)
+ self.enc4 = layers.Encoder(ch * 4, ch * 8, 3, 2, 1)
+
+ self.aspp = layers.ASPPModule(ch * 8, ch * 16, dilations)
+
+ self.dec4 = layers.Decoder(ch * (8 + 16), ch * 8, 3, 1, 1)
+ self.dec3 = layers.Decoder(ch * (4 + 8), ch * 4, 3, 1, 1)
+ self.dec2 = layers.Decoder(ch * (2 + 4), ch * 2, 3, 1, 1)
+ self.dec1 = layers.Decoder(ch * (1 + 2), ch, 3, 1, 1)
+
+ def __call__(self, x):
+ h, e1 = self.enc1(x)
+ h, e2 = self.enc2(h)
+ h, e3 = self.enc3(h)
+ h, e4 = self.enc4(h)
+
+ h = self.aspp(h)
+
+ h = self.dec4(h, e4)
+ h = self.dec3(h, e3)
+ h = self.dec2(h, e2)
+ h = self.dec1(h, e1)
+
+ return h
+
+
+class CascadedASPPNet(nn.Module):
+
+ def __init__(self, n_fft):
+ super(CascadedASPPNet, self).__init__()
+ self.stg1_low_band_net = BaseASPPNet(2, 32)
+ self.stg1_high_band_net = BaseASPPNet(2, 32)
+
+ self.stg2_bridge = layers.Conv2DBNActiv(34, 16, 1, 1, 0)
+ self.stg2_full_band_net = BaseASPPNet(16, 32)
+
+ self.stg3_bridge = layers.Conv2DBNActiv(66, 32, 1, 1, 0)
+ self.stg3_full_band_net = BaseASPPNet(32, 64)
+
+ self.out = nn.Conv2d(64, 2, 1, bias=False)
+ self.aux1_out = nn.Conv2d(32, 2, 1, bias=False)
+ self.aux2_out = nn.Conv2d(32, 2, 1, bias=False)
+
+ self.max_bin = n_fft // 2
+ self.output_bin = n_fft // 2 + 1
+
+ self.offset = 128
+
+ def forward(self, x, aggressiveness=None):
+ mix = x.detach()
+ x = x.clone()
+
+ x = x[:, :, :self.max_bin]
+
+ bandw = x.size()[2] // 2
+ aux1 = torch.cat([
+ self.stg1_low_band_net(x[:, :, :bandw]),
+ self.stg1_high_band_net(x[:, :, bandw:])
+ ], dim=2)
+
+ h = torch.cat([x, aux1], dim=1)
+ aux2 = self.stg2_full_band_net(self.stg2_bridge(h))
+
+ h = torch.cat([x, aux1, aux2], dim=1)
+ h = self.stg3_full_band_net(self.stg3_bridge(h))
+
+ mask = torch.sigmoid(self.out(h))
+ mask = F.pad(
+ input=mask,
+ pad=(0, 0, 0, self.output_bin - mask.size()[2]),
+ mode='replicate')
+
+ if self.training:
+ aux1 = torch.sigmoid(self.aux1_out(aux1))
+ aux1 = F.pad(
+ input=aux1,
+ pad=(0, 0, 0, self.output_bin - aux1.size()[2]),
+ mode='replicate')
+ aux2 = torch.sigmoid(self.aux2_out(aux2))
+ aux2 = F.pad(
+ input=aux2,
+ pad=(0, 0, 0, self.output_bin - aux2.size()[2]),
+ mode='replicate')
+ return mask * mix, aux1 * mix, aux2 * mix
+ else:
+ if aggressiveness:
+ mask[:, :, :aggressiveness['split_bin']] = torch.pow(mask[:, :, :aggressiveness['split_bin']], 1 + aggressiveness['value'] / 3)
+ mask[:, :, aggressiveness['split_bin']:] = torch.pow(mask[:, :, aggressiveness['split_bin']:], 1 + aggressiveness['value'])
+
+ return mask * mix
+
+ def predict(self, x_mag, aggressiveness=None):
+ h = self.forward(x_mag, aggressiveness)
+
+ if self.offset > 0:
+ h = h[:, :, :, self.offset:-self.offset]
+ assert h.size()[3] > 0
+
+ return h
diff --git a/uvr5_pack/lib_v5/spec_utils.py b/uvr5_pack/lib_v5/spec_utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..fe95916f03681ec023d3ebb3e81ef86284ea7cf1
--- /dev/null
+++ b/uvr5_pack/lib_v5/spec_utils.py
@@ -0,0 +1,485 @@
+import os,librosa
+import numpy as np
+import soundfile as sf
+from tqdm import tqdm
+import json,math ,hashlib
+
+def crop_center(h1, h2):
+ h1_shape = h1.size()
+ h2_shape = h2.size()
+
+ if h1_shape[3] == h2_shape[3]:
+ return h1
+ elif h1_shape[3] < h2_shape[3]:
+ raise ValueError('h1_shape[3] must be greater than h2_shape[3]')
+
+ # s_freq = (h2_shape[2] - h1_shape[2]) // 2
+ # e_freq = s_freq + h1_shape[2]
+ s_time = (h1_shape[3] - h2_shape[3]) // 2
+ e_time = s_time + h2_shape[3]
+ h1 = h1[:, :, :, s_time:e_time]
+
+ return h1
+
+
+def wave_to_spectrogram(wave, hop_length, n_fft, mid_side=False, mid_side_b2=False, reverse=False):
+ if reverse:
+ wave_left = np.flip(np.asfortranarray(wave[0]))
+ wave_right = np.flip(np.asfortranarray(wave[1]))
+ elif mid_side:
+ wave_left = np.asfortranarray(np.add(wave[0], wave[1]) / 2)
+ wave_right = np.asfortranarray(np.subtract(wave[0], wave[1]))
+ elif mid_side_b2:
+ wave_left = np.asfortranarray(np.add(wave[1], wave[0] * .5))
+ wave_right = np.asfortranarray(np.subtract(wave[0], wave[1] * .5))
+ else:
+ wave_left = np.asfortranarray(wave[0])
+ wave_right = np.asfortranarray(wave[1])
+
+ spec_left = librosa.stft(wave_left, n_fft, hop_length=hop_length)
+ spec_right = librosa.stft(wave_right, n_fft, hop_length=hop_length)
+
+ spec = np.asfortranarray([spec_left, spec_right])
+
+ return spec
+
+
+def wave_to_spectrogram_mt(wave, hop_length, n_fft, mid_side=False, mid_side_b2=False, reverse=False):
+ import threading
+
+ if reverse:
+ wave_left = np.flip(np.asfortranarray(wave[0]))
+ wave_right = np.flip(np.asfortranarray(wave[1]))
+ elif mid_side:
+ wave_left = np.asfortranarray(np.add(wave[0], wave[1]) / 2)
+ wave_right = np.asfortranarray(np.subtract(wave[0], wave[1]))
+ elif mid_side_b2:
+ wave_left = np.asfortranarray(np.add(wave[1], wave[0] * .5))
+ wave_right = np.asfortranarray(np.subtract(wave[0], wave[1] * .5))
+ else:
+ wave_left = np.asfortranarray(wave[0])
+ wave_right = np.asfortranarray(wave[1])
+
+ def run_thread(**kwargs):
+ global spec_left
+ spec_left = librosa.stft(**kwargs)
+
+ thread = threading.Thread(target=run_thread, kwargs={'y': wave_left, 'n_fft': n_fft, 'hop_length': hop_length})
+ thread.start()
+ spec_right = librosa.stft(wave_right, n_fft, hop_length=hop_length)
+ thread.join()
+
+ spec = np.asfortranarray([spec_left, spec_right])
+
+ return spec
+
+
+def combine_spectrograms(specs, mp):
+ l = min([specs[i].shape[2] for i in specs])
+ spec_c = np.zeros(shape=(2, mp.param['bins'] + 1, l), dtype=np.complex64)
+ offset = 0
+ bands_n = len(mp.param['band'])
+
+ for d in range(1, bands_n + 1):
+ h = mp.param['band'][d]['crop_stop'] - mp.param['band'][d]['crop_start']
+ spec_c[:, offset:offset+h, :l] = specs[d][:, mp.param['band'][d]['crop_start']:mp.param['band'][d]['crop_stop'], :l]
+ offset += h
+
+ if offset > mp.param['bins']:
+ raise ValueError('Too much bins')
+
+ # lowpass fiter
+ if mp.param['pre_filter_start'] > 0: # and mp.param['band'][bands_n]['res_type'] in ['scipy', 'polyphase']:
+ if bands_n == 1:
+ spec_c = fft_lp_filter(spec_c, mp.param['pre_filter_start'], mp.param['pre_filter_stop'])
+ else:
+ gp = 1
+ for b in range(mp.param['pre_filter_start'] + 1, mp.param['pre_filter_stop']):
+ g = math.pow(10, -(b - mp.param['pre_filter_start']) * (3.5 - gp) / 20.0)
+ gp = g
+ spec_c[:, b, :] *= g
+
+ return np.asfortranarray(spec_c)
+
+
+def spectrogram_to_image(spec, mode='magnitude'):
+ if mode == 'magnitude':
+ if np.iscomplexobj(spec):
+ y = np.abs(spec)
+ else:
+ y = spec
+ y = np.log10(y ** 2 + 1e-8)
+ elif mode == 'phase':
+ if np.iscomplexobj(spec):
+ y = np.angle(spec)
+ else:
+ y = spec
+
+ y -= y.min()
+ y *= 255 / y.max()
+ img = np.uint8(y)
+
+ if y.ndim == 3:
+ img = img.transpose(1, 2, 0)
+ img = np.concatenate([
+ np.max(img, axis=2, keepdims=True), img
+ ], axis=2)
+
+ return img
+
+
+def reduce_vocal_aggressively(X, y, softmask):
+ v = X - y
+ y_mag_tmp = np.abs(y)
+ v_mag_tmp = np.abs(v)
+
+ v_mask = v_mag_tmp > y_mag_tmp
+ y_mag = np.clip(y_mag_tmp - v_mag_tmp * v_mask * softmask, 0, np.inf)
+
+ return y_mag * np.exp(1.j * np.angle(y))
+
+
+def mask_silence(mag, ref, thres=0.2, min_range=64, fade_size=32):
+ if min_range < fade_size * 2:
+ raise ValueError('min_range must be >= fade_area * 2')
+
+ mag = mag.copy()
+
+ idx = np.where(ref.mean(axis=(0, 1)) < thres)[0]
+ starts = np.insert(idx[np.where(np.diff(idx) != 1)[0] + 1], 0, idx[0])
+ ends = np.append(idx[np.where(np.diff(idx) != 1)[0]], idx[-1])
+ uninformative = np.where(ends - starts > min_range)[0]
+ if len(uninformative) > 0:
+ starts = starts[uninformative]
+ ends = ends[uninformative]
+ old_e = None
+ for s, e in zip(starts, ends):
+ if old_e is not None and s - old_e < fade_size:
+ s = old_e - fade_size * 2
+
+ if s != 0:
+ weight = np.linspace(0, 1, fade_size)
+ mag[:, :, s:s + fade_size] += weight * ref[:, :, s:s + fade_size]
+ else:
+ s -= fade_size
+
+ if e != mag.shape[2]:
+ weight = np.linspace(1, 0, fade_size)
+ mag[:, :, e - fade_size:e] += weight * ref[:, :, e - fade_size:e]
+ else:
+ e += fade_size
+
+ mag[:, :, s + fade_size:e - fade_size] += ref[:, :, s + fade_size:e - fade_size]
+ old_e = e
+
+ return mag
+
+
+def align_wave_head_and_tail(a, b):
+ l = min([a[0].size, b[0].size])
+
+ return a[:l,:l], b[:l,:l]
+
+
+def cache_or_load(mix_path, inst_path, mp):
+ mix_basename = os.path.splitext(os.path.basename(mix_path))[0]
+ inst_basename = os.path.splitext(os.path.basename(inst_path))[0]
+
+ cache_dir = 'mph{}'.format(hashlib.sha1(json.dumps(mp.param, sort_keys=True).encode('utf-8')).hexdigest())
+ mix_cache_dir = os.path.join('cache', cache_dir)
+ inst_cache_dir = os.path.join('cache', cache_dir)
+
+ os.makedirs(mix_cache_dir, exist_ok=True)
+ os.makedirs(inst_cache_dir, exist_ok=True)
+
+ mix_cache_path = os.path.join(mix_cache_dir, mix_basename + '.npy')
+ inst_cache_path = os.path.join(inst_cache_dir, inst_basename + '.npy')
+
+ if os.path.exists(mix_cache_path) and os.path.exists(inst_cache_path):
+ X_spec_m = np.load(mix_cache_path)
+ y_spec_m = np.load(inst_cache_path)
+ else:
+ X_wave, y_wave, X_spec_s, y_spec_s = {}, {}, {}, {}
+
+ for d in range(len(mp.param['band']), 0, -1):
+ bp = mp.param['band'][d]
+
+ if d == len(mp.param['band']): # high-end band
+ X_wave[d], _ = librosa.load(
+ mix_path, bp['sr'], False, dtype=np.float32, res_type=bp['res_type'])
+ y_wave[d], _ = librosa.load(
+ inst_path, bp['sr'], False, dtype=np.float32, res_type=bp['res_type'])
+ else: # lower bands
+ X_wave[d] = librosa.resample(X_wave[d+1], mp.param['band'][d+1]['sr'], bp['sr'], res_type=bp['res_type'])
+ y_wave[d] = librosa.resample(y_wave[d+1], mp.param['band'][d+1]['sr'], bp['sr'], res_type=bp['res_type'])
+
+ X_wave[d], y_wave[d] = align_wave_head_and_tail(X_wave[d], y_wave[d])
+
+ X_spec_s[d] = wave_to_spectrogram(X_wave[d], bp['hl'], bp['n_fft'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse'])
+ y_spec_s[d] = wave_to_spectrogram(y_wave[d], bp['hl'], bp['n_fft'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse'])
+
+ del X_wave, y_wave
+
+ X_spec_m = combine_spectrograms(X_spec_s, mp)
+ y_spec_m = combine_spectrograms(y_spec_s, mp)
+
+ if X_spec_m.shape != y_spec_m.shape:
+ raise ValueError('The combined spectrograms are different: ' + mix_path)
+
+ _, ext = os.path.splitext(mix_path)
+
+ np.save(mix_cache_path, X_spec_m)
+ np.save(inst_cache_path, y_spec_m)
+
+ return X_spec_m, y_spec_m
+
+
+def spectrogram_to_wave(spec, hop_length, mid_side, mid_side_b2, reverse):
+ spec_left = np.asfortranarray(spec[0])
+ spec_right = np.asfortranarray(spec[1])
+
+ wave_left = librosa.istft(spec_left, hop_length=hop_length)
+ wave_right = librosa.istft(spec_right, hop_length=hop_length)
+
+ if reverse:
+ return np.asfortranarray([np.flip(wave_left), np.flip(wave_right)])
+ elif mid_side:
+ return np.asfortranarray([np.add(wave_left, wave_right / 2), np.subtract(wave_left, wave_right / 2)])
+ elif mid_side_b2:
+ return np.asfortranarray([np.add(wave_right / 1.25, .4 * wave_left), np.subtract(wave_left / 1.25, .4 * wave_right)])
+ else:
+ return np.asfortranarray([wave_left, wave_right])
+
+
+def spectrogram_to_wave_mt(spec, hop_length, mid_side, reverse, mid_side_b2):
+ import threading
+
+ spec_left = np.asfortranarray(spec[0])
+ spec_right = np.asfortranarray(spec[1])
+
+ def run_thread(**kwargs):
+ global wave_left
+ wave_left = librosa.istft(**kwargs)
+
+ thread = threading.Thread(target=run_thread, kwargs={'stft_matrix': spec_left, 'hop_length': hop_length})
+ thread.start()
+ wave_right = librosa.istft(spec_right, hop_length=hop_length)
+ thread.join()
+
+ if reverse:
+ return np.asfortranarray([np.flip(wave_left), np.flip(wave_right)])
+ elif mid_side:
+ return np.asfortranarray([np.add(wave_left, wave_right / 2), np.subtract(wave_left, wave_right / 2)])
+ elif mid_side_b2:
+ return np.asfortranarray([np.add(wave_right / 1.25, .4 * wave_left), np.subtract(wave_left / 1.25, .4 * wave_right)])
+ else:
+ return np.asfortranarray([wave_left, wave_right])
+
+
+def cmb_spectrogram_to_wave(spec_m, mp, extra_bins_h=None, extra_bins=None):
+ wave_band = {}
+ bands_n = len(mp.param['band'])
+ offset = 0
+
+ for d in range(1, bands_n + 1):
+ bp = mp.param['band'][d]
+ spec_s = np.ndarray(shape=(2, bp['n_fft'] // 2 + 1, spec_m.shape[2]), dtype=complex)
+ h = bp['crop_stop'] - bp['crop_start']
+ spec_s[:, bp['crop_start']:bp['crop_stop'], :] = spec_m[:, offset:offset+h, :]
+
+ offset += h
+ if d == bands_n: # higher
+ if extra_bins_h: # if --high_end_process bypass
+ max_bin = bp['n_fft'] // 2
+ spec_s[:, max_bin-extra_bins_h:max_bin, :] = extra_bins[:, :extra_bins_h, :]
+ if bp['hpf_start'] > 0:
+ spec_s = fft_hp_filter(spec_s, bp['hpf_start'], bp['hpf_stop'] - 1)
+ if bands_n == 1:
+ wave = spectrogram_to_wave(spec_s, bp['hl'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse'])
+ else:
+ wave = np.add(wave, spectrogram_to_wave(spec_s, bp['hl'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse']))
+ else:
+ sr = mp.param['band'][d+1]['sr']
+ if d == 1: # lower
+ spec_s = fft_lp_filter(spec_s, bp['lpf_start'], bp['lpf_stop'])
+ wave = librosa.resample(spectrogram_to_wave(spec_s, bp['hl'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse']), bp['sr'], sr, res_type="sinc_fastest")
+ else: # mid
+ spec_s = fft_hp_filter(spec_s, bp['hpf_start'], bp['hpf_stop'] - 1)
+ spec_s = fft_lp_filter(spec_s, bp['lpf_start'], bp['lpf_stop'])
+ wave2 = np.add(wave, spectrogram_to_wave(spec_s, bp['hl'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse']))
+ # wave = librosa.core.resample(wave2, bp['sr'], sr, res_type="sinc_fastest")
+ wave = librosa.core.resample(wave2, bp['sr'], sr,res_type='scipy')
+
+ return wave.T
+
+
+def fft_lp_filter(spec, bin_start, bin_stop):
+ g = 1.0
+ for b in range(bin_start, bin_stop):
+ g -= 1 / (bin_stop - bin_start)
+ spec[:, b, :] = g * spec[:, b, :]
+
+ spec[:, bin_stop:, :] *= 0
+
+ return spec
+
+
+def fft_hp_filter(spec, bin_start, bin_stop):
+ g = 1.0
+ for b in range(bin_start, bin_stop, -1):
+ g -= 1 / (bin_start - bin_stop)
+ spec[:, b, :] = g * spec[:, b, :]
+
+ spec[:, 0:bin_stop+1, :] *= 0
+
+ return spec
+
+
+def mirroring(a, spec_m, input_high_end, mp):
+ if 'mirroring' == a:
+ mirror = np.flip(np.abs(spec_m[:, mp.param['pre_filter_start']-10-input_high_end.shape[1]:mp.param['pre_filter_start']-10, :]), 1)
+ mirror = mirror * np.exp(1.j * np.angle(input_high_end))
+
+ return np.where(np.abs(input_high_end) <= np.abs(mirror), input_high_end, mirror)
+
+ if 'mirroring2' == a:
+ mirror = np.flip(np.abs(spec_m[:, mp.param['pre_filter_start']-10-input_high_end.shape[1]:mp.param['pre_filter_start']-10, :]), 1)
+ mi = np.multiply(mirror, input_high_end * 1.7)
+
+ return np.where(np.abs(input_high_end) <= np.abs(mi), input_high_end, mi)
+
+
+def ensembling(a, specs):
+ for i in range(1, len(specs)):
+ if i == 1:
+ spec = specs[0]
+
+ ln = min([spec.shape[2], specs[i].shape[2]])
+ spec = spec[:,:,:ln]
+ specs[i] = specs[i][:,:,:ln]
+
+ if 'min_mag' == a:
+ spec = np.where(np.abs(specs[i]) <= np.abs(spec), specs[i], spec)
+ if 'max_mag' == a:
+ spec = np.where(np.abs(specs[i]) >= np.abs(spec), specs[i], spec)
+
+ return spec
+
+def stft(wave, nfft, hl):
+ wave_left = np.asfortranarray(wave[0])
+ wave_right = np.asfortranarray(wave[1])
+ spec_left = librosa.stft(wave_left, nfft, hop_length=hl)
+ spec_right = librosa.stft(wave_right, nfft, hop_length=hl)
+ spec = np.asfortranarray([spec_left, spec_right])
+
+ return spec
+
+def istft(spec, hl):
+ spec_left = np.asfortranarray(spec[0])
+ spec_right = np.asfortranarray(spec[1])
+
+ wave_left = librosa.istft(spec_left, hop_length=hl)
+ wave_right = librosa.istft(spec_right, hop_length=hl)
+ wave = np.asfortranarray([wave_left, wave_right])
+
+
+if __name__ == "__main__":
+ import cv2
+ import sys
+ import time
+ import argparse
+ from model_param_init import ModelParameters
+
+ p = argparse.ArgumentParser()
+ p.add_argument('--algorithm', '-a', type=str, choices=['invert', 'invert_p', 'min_mag', 'max_mag', 'deep', 'align'], default='min_mag')
+ p.add_argument('--model_params', '-m', type=str, default=os.path.join('modelparams', '1band_sr44100_hl512.json'))
+ p.add_argument('--output_name', '-o', type=str, default='output')
+ p.add_argument('--vocals_only', '-v', action='store_true')
+ p.add_argument('input', nargs='+')
+ args = p.parse_args()
+
+ start_time = time.time()
+
+ if args.algorithm.startswith('invert') and len(args.input) != 2:
+ raise ValueError('There should be two input files.')
+
+ if not args.algorithm.startswith('invert') and len(args.input) < 2:
+ raise ValueError('There must be at least two input files.')
+
+ wave, specs = {}, {}
+ mp = ModelParameters(args.model_params)
+
+ for i in range(len(args.input)):
+ spec = {}
+
+ for d in range(len(mp.param['band']), 0, -1):
+ bp = mp.param['band'][d]
+
+ if d == len(mp.param['band']): # high-end band
+ wave[d], _ = librosa.load(
+ args.input[i], bp['sr'], False, dtype=np.float32, res_type=bp['res_type'])
+
+ if len(wave[d].shape) == 1: # mono to stereo
+ wave[d] = np.array([wave[d], wave[d]])
+ else: # lower bands
+ wave[d] = librosa.resample(wave[d+1], mp.param['band'][d+1]['sr'], bp['sr'], res_type=bp['res_type'])
+
+ spec[d] = wave_to_spectrogram(wave[d], bp['hl'], bp['n_fft'], mp.param['mid_side'], mp.param['mid_side_b2'], mp.param['reverse'])
+
+ specs[i] = combine_spectrograms(spec, mp)
+
+ del wave
+
+ if args.algorithm == 'deep':
+ d_spec = np.where(np.abs(specs[0]) <= np.abs(spec[1]), specs[0], spec[1])
+ v_spec = d_spec - specs[1]
+ sf.write(os.path.join('{}.wav'.format(args.output_name)), cmb_spectrogram_to_wave(v_spec, mp), mp.param['sr'])
+
+ if args.algorithm.startswith('invert'):
+ ln = min([specs[0].shape[2], specs[1].shape[2]])
+ specs[0] = specs[0][:,:,:ln]
+ specs[1] = specs[1][:,:,:ln]
+
+ if 'invert_p' == args.algorithm:
+ X_mag = np.abs(specs[0])
+ y_mag = np.abs(specs[1])
+ max_mag = np.where(X_mag >= y_mag, X_mag, y_mag)
+ v_spec = specs[1] - max_mag * np.exp(1.j * np.angle(specs[0]))
+ else:
+ specs[1] = reduce_vocal_aggressively(specs[0], specs[1], 0.2)
+ v_spec = specs[0] - specs[1]
+
+ if not args.vocals_only:
+ X_mag = np.abs(specs[0])
+ y_mag = np.abs(specs[1])
+ v_mag = np.abs(v_spec)
+
+ X_image = spectrogram_to_image(X_mag)
+ y_image = spectrogram_to_image(y_mag)
+ v_image = spectrogram_to_image(v_mag)
+
+ cv2.imwrite('{}_X.png'.format(args.output_name), X_image)
+ cv2.imwrite('{}_y.png'.format(args.output_name), y_image)
+ cv2.imwrite('{}_v.png'.format(args.output_name), v_image)
+
+ sf.write('{}_X.wav'.format(args.output_name), cmb_spectrogram_to_wave(specs[0], mp), mp.param['sr'])
+ sf.write('{}_y.wav'.format(args.output_name), cmb_spectrogram_to_wave(specs[1], mp), mp.param['sr'])
+
+ sf.write('{}_v.wav'.format(args.output_name), cmb_spectrogram_to_wave(v_spec, mp), mp.param['sr'])
+ else:
+ if not args.algorithm == 'deep':
+ sf.write(os.path.join('ensembled','{}.wav'.format(args.output_name)), cmb_spectrogram_to_wave(ensembling(args.algorithm, specs), mp), mp.param['sr'])
+
+ if args.algorithm == 'align':
+
+ trackalignment = [
+ {
+ 'file1':'"{}"'.format(args.input[0]),
+ 'file2':'"{}"'.format(args.input[1])
+ }
+ ]
+
+ for i,e in tqdm(enumerate(trackalignment), desc="Performing Alignment..."):
+ os.system(f"python lib/align_tracks.py {e['file1']} {e['file2']}")
+
+ #print('Total time: {0:.{1}f}s'.format(time.time() - start_time, 1))
diff --git a/uvr5_pack/utils.py b/uvr5_pack/utils.py
new file mode 100644
index 0000000000000000000000000000000000000000..2848355d0cc5117ad878651454c4d4e2e432f05d
--- /dev/null
+++ b/uvr5_pack/utils.py
@@ -0,0 +1,242 @@
+import torch
+import numpy as np
+from tqdm import tqdm
+
+def make_padding(width, cropsize, offset):
+ left = offset
+ roi_size = cropsize - left * 2
+ if roi_size == 0:
+ roi_size = cropsize
+ right = roi_size - (width % roi_size) + left
+
+ return left, right, roi_size
+def inference(X_spec, device, model, aggressiveness,data):
+ '''
+ data ： dic configs
+ '''
+
+ def _execute(X_mag_pad, roi_size, n_window, device, model, aggressiveness,is_half=True):
+ model.eval()
+ with torch.no_grad():
+ preds = []
+
+ iterations = [n_window]
+
+ total_iterations = sum(iterations)
+ for i in tqdm(range(n_window)):
+ start = i * roi_size
+ X_mag_window = X_mag_pad[None, :, :, start:start + data['window_size']]
+ X_mag_window = torch.from_numpy(X_mag_window)
+ if(is_half==True):X_mag_window=X_mag_window.half()
+ X_mag_window=X_mag_window.to(device)
+
+ pred = model.predict(X_mag_window, aggressiveness)
+
+ pred = pred.detach().cpu().numpy()
+ preds.append(pred[0])
+
+ pred = np.concatenate(preds, axis=2)
+ return pred
+
+ def preprocess(X_spec):
+ X_mag = np.abs(X_spec)
+ X_phase = np.angle(X_spec)
+
+ return X_mag, X_phase
+
+ X_mag, X_phase = preprocess(X_spec)
+
+ coef = X_mag.max()
+ X_mag_pre = X_mag / coef
+
+ n_frame = X_mag_pre.shape[2]
+ pad_l, pad_r, roi_size = make_padding(n_frame,
+ data['window_size'], model.offset)
+ n_window = int(np.ceil(n_frame / roi_size))
+
+ X_mag_pad = np.pad(
+ X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
+
+ if(list(model.state_dict().values())[0].dtype==torch.float16):is_half=True
+ else:is_half=False
+ pred = _execute(X_mag_pad, roi_size, n_window,
+ device, model, aggressiveness,is_half)
+ pred = pred[:, :, :n_frame]
+
+ if data['tta']:
+ pad_l += roi_size // 2
+ pad_r += roi_size // 2
+ n_window += 1
+
+ X_mag_pad = np.pad(
+ X_mag_pre, ((0, 0), (0, 0), (pad_l, pad_r)), mode='constant')
+
+ pred_tta = _execute(X_mag_pad, roi_size, n_window,
+ device, model, aggressiveness,is_half)
+ pred_tta = pred_tta[:, :, roi_size // 2:]
+ pred_tta = pred_tta[:, :, :n_frame]
+
+ return (pred + pred_tta) * 0.5 * coef, X_mag, np.exp(1.j * X_phase)
+ else:
+ return pred * coef, X_mag, np.exp(1.j * X_phase)
+
+
+
+def _get_name_params(model_path , model_hash):
+ ModelName = model_path
+ if model_hash == '47939caf0cfe52a0e81442b85b971dfd':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100')
+ if model_hash == '4e4ecb9764c50a8c414fee6e10395bbe':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_v2.json')
+ param_name_auto=str('4band_v2')
+ if model_hash == 'ca106edd563e034bde0bdec4bb7a4b36':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_v2.json')
+ param_name_auto=str('4band_v2')
+ if model_hash == 'e60a1e84803ce4efc0a6551206cc4b71':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100')
+ if model_hash == 'a82f14e75892e55e994376edbf0c8435':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100')
+ if model_hash == '6dd9eaa6f0420af9f1d403aaafa4cc06':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_v2_sn.json')
+ param_name_auto=str('4band_v2_sn')
+ if model_hash == '08611fb99bd59eaa79ad27c58d137727':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_v2_sn.json')
+ param_name_auto=str('4band_v2_sn')
+ if model_hash == '5c7bbca45a187e81abbbd351606164e5':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100_msb2.json')
+ param_name_auto=str('3band_44100_msb2')
+ if model_hash == 'd6b2cb685a058a091e5e7098192d3233':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100_msb2.json')
+ param_name_auto=str('3band_44100_msb2')
+ if model_hash == 'c1b9f38170a7c90e96f027992eb7c62b':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100')
+ if model_hash == 'c3448ec923fa0edf3d03a19e633faa53':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100')
+ if model_hash == '68aa2c8093d0080704b200d140f59e54':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100.json')
+ param_name_auto=str('3band_44100.json')
+ if model_hash == 'fdc83be5b798e4bd29fe00fe6600e147':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100_mid.json')
+ param_name_auto=str('3band_44100_mid.json')
+ if model_hash == '2ce34bc92fd57f55db16b7a4def3d745':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100_mid.json')
+ param_name_auto=str('3band_44100_mid.json')
+ if model_hash == '52fdca89576f06cf4340b74a4730ee5f':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100.json')
+ if model_hash == '41191165b05d38fc77f072fa9e8e8a30':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100.json')
+ if model_hash == '89e83b511ad474592689e562d5b1f80e':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/2band_32000.json')
+ param_name_auto=str('2band_32000.json')
+ if model_hash == '0b954da81d453b716b114d6d7c95177f':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/2band_32000.json')
+ param_name_auto=str('2band_32000.json')
+
+ #v4 Models
+ if model_hash == '6a00461c51c2920fd68937d4609ed6c8':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr16000_hl512.json')
+ param_name_auto=str('1band_sr16000_hl512')
+ if model_hash == '0ab504864d20f1bd378fe9c81ef37140':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json')
+ param_name_auto=str('1band_sr32000_hl512')
+ if model_hash == '7dd21065bf91c10f7fccb57d7d83b07f':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json')
+ param_name_auto=str('1band_sr32000_hl512')
+ if model_hash == '80ab74d65e515caa3622728d2de07d23':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json')
+ param_name_auto=str('1band_sr32000_hl512')
+ if model_hash == 'edc115e7fc523245062200c00caa847f':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr33075_hl384.json')
+ param_name_auto=str('1band_sr33075_hl384')
+ if model_hash == '28063e9f6ab5b341c5f6d3c67f2045b7':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr33075_hl384.json')
+ param_name_auto=str('1band_sr33075_hl384')
+ if model_hash == 'b58090534c52cbc3e9b5104bad666ef2':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512.json')
+ param_name_auto=str('1band_sr44100_hl512')
+ if model_hash == '0cdab9947f1b0928705f518f3c78ea8f':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512.json')
+ param_name_auto=str('1band_sr44100_hl512')
+ if model_hash == 'ae702fed0238afb5346db8356fe25f13':
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl1024.json')
+ param_name_auto=str('1band_sr44100_hl1024')
+ #User Models
+
+ #1 Band
+ if '1band_sr16000_hl512' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr16000_hl512.json')
+ param_name_auto=str('1band_sr16000_hl512')
+ if '1band_sr32000_hl512' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr32000_hl512.json')
+ param_name_auto=str('1band_sr32000_hl512')
+ if '1band_sr33075_hl384' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr33075_hl384.json')
+ param_name_auto=str('1band_sr33075_hl384')
+ if '1band_sr44100_hl256' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl256.json')
+ param_name_auto=str('1band_sr44100_hl256')
+ if '1band_sr44100_hl512' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl512.json')
+ param_name_auto=str('1band_sr44100_hl512')
+ if '1band_sr44100_hl1024' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/1band_sr44100_hl1024.json')
+ param_name_auto=str('1band_sr44100_hl1024')
+
+ #2 Band
+ if '2band_44100_lofi' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/2band_44100_lofi.json')
+ param_name_auto=str('2band_44100_lofi')
+ if '2band_32000' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/2band_32000.json')
+ param_name_auto=str('2band_32000')
+ if '2band_48000' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/2band_48000.json')
+ param_name_auto=str('2band_48000')
+
+ #3 Band
+ if '3band_44100' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100.json')
+ param_name_auto=str('3band_44100')
+ if '3band_44100_mid' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100_mid.json')
+ param_name_auto=str('3band_44100_mid')
+ if '3band_44100_msb2' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/3band_44100_msb2.json')
+ param_name_auto=str('3band_44100_msb2')
+
+ #4 Band
+ if '4band_44100' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100.json')
+ param_name_auto=str('4band_44100')
+ if '4band_44100_mid' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100_mid.json')
+ param_name_auto=str('4band_44100_mid')
+ if '4band_44100_msb' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100_msb.json')
+ param_name_auto=str('4band_44100_msb')
+ if '4band_44100_msb2' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100_msb2.json')
+ param_name_auto=str('4band_44100_msb2')
+ if '4band_44100_reverse' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100_reverse.json')
+ param_name_auto=str('4band_44100_reverse')
+ if '4band_44100_sw' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_44100_sw.json')
+ param_name_auto=str('4band_44100_sw')
+ if '4band_v2' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_v2.json')
+ param_name_auto=str('4band_v2')
+ if '4band_v2_sn' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/4band_v2_sn.json')
+ param_name_auto=str('4band_v2_sn')
+ if 'tmodelparam' in ModelName:
+ model_params_auto=str('runtime/Lib/site-packages/uvr5_pack/lib_v5/modelparams/tmodelparam.json')
+ param_name_auto=str('User Model Param Set')
+ return param_name_auto , model_params_auto
diff --git "a/uvr5_weights/HP2-\344\272\272\345\243\260vocals+\351\235\236\344\272\272\345\243\260instrumentals.pth" "b/uvr5_weights/HP2-\344\272\272\345\243\260vocals+\351\235\236\344\272\272\345\243\260instrumentals.pth"
new file mode 100644
index 0000000000000000000000000000000000000000..2594ca1392e5e9f7b8d9a8535c7174bf36309a9e
--- /dev/null
+++ "b/uvr5_weights/HP2-\344\272\272\345\243\260vocals+\351\235\236\344\272\272\345\243\260instrumentals.pth"
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39796caa5db18d7f9382d8ac997ac967bfd85f7761014bb807d2543cc844ef05
+size 63454827
diff --git "a/uvr5_weights/HP5-\344\270\273\346\227\213\345\276\213\344\272\272\345\243\260vocals+\345\205\266\344\273\226instrumentals.pth" "b/uvr5_weights/HP5-\344\270\273\346\227\213\345\276\213\344\272\272\345\243\260vocals+\345\205\266\344\273\226instrumentals.pth"
new file mode 100644
index 0000000000000000000000000000000000000000..37265088b4cff876d7635addd70ffa3f0fb259e1
--- /dev/null
+++ "b/uvr5_weights/HP5-\344\270\273\346\227\213\345\276\213\344\272\272\345\243\260vocals+\345\205\266\344\273\226instrumentals.pth"
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5908891829634926119720241e8573d97cbeb8277110a7512bdb0bd7563258ee
+size 63454827
diff --git a/vc_infer_pipeline.py b/vc_infer_pipeline.py
new file mode 100644
index 0000000000000000000000000000000000000000..6a4c10da721820e631a38b965a79dc05a9708a87
--- /dev/null
+++ b/vc_infer_pipeline.py
@@ -0,0 +1,225 @@
+import numpy as np,parselmouth,torch,pdb
+from time import time as ttime
+import torch.nn.functional as F
+from config import x_pad,x_query,x_center,x_max
+from sklearn.cluster import KMeans
+
+def resize2d(x, target_len,is1):
+ minn=1 if is1==True else 0
+ ss = np.array(x).astype("float32")
+ ss[ss <=minn] = np.nan
+ target = np.interp(np.arange(0, len(ss) * target_len, len(ss)) / target_len, np.arange(0, len(ss)), ss)
+ res = np.nan_to_num(target)
+ return res
+
+class VC(object):
+ def __init__(self,tgt_sr,device,is_half):
+ self.sr=16000#hubert输入采样率
+ self.window=160#每帧点数
+ self.t_pad=self.sr*x_pad#每条前后pad时间
+ self.t_pad_tgt=tgt_sr*x_pad
+ self.t_pad2=self.t_pad*2
+ self.t_query=self.sr*x_query#查询切点前后查询时间
+ self.t_center=self.sr*x_center#查询切点位置
+ self.t_max=self.sr*x_max#免查询时长阈值
+ self.device=device
+ self.is_half=is_half
+
+ def get_f0(self,x, p_len,f0_up_key=0,inp_f0=None):
+ time_step = self.window / self.sr * 1000
+ f0_min = 50
+ f0_max = 1100
+ f0_mel_min = 1127 * np.log(1 + f0_min / 700)
+ f0_mel_max = 1127 * np.log(1 + f0_max / 700)
+ f0 = parselmouth.Sound(x, self.sr).to_pitch_ac(
+ time_step=time_step / 1000, voicing_threshold=0.6,
+ pitch_floor=f0_min, pitch_ceiling=f0_max).selected_array['frequency']
+ pad_size=(p_len - len(f0) + 1) // 2
+ if(pad_size>0 or p_len - len(f0) - pad_size>0):
+ f0 = np.pad(f0,[[pad_size,p_len - len(f0) - pad_size]], mode='constant')
+ f0 *= pow(2, f0_up_key / 12)
+ # with open("test.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
+ tf0=self.sr//self.window#每秒f0点数
+ if (inp_f0 is not None):
+ delta_t=np.round((inp_f0[:,0].max()-inp_f0[:,0].min())*tf0+1).astype("int16")
+ replace_f0=np.interp(list(range(delta_t)), inp_f0[:, 0]*100, inp_f0[:, 1])
+ shape=f0[x_pad*tf0:x_pad*tf0+len(replace_f0)].shape[0]
+ f0[x_pad*tf0:x_pad*tf0+len(replace_f0)]=replace_f0[:shape]
+ # with open("test_opt.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
+ f0bak = f0.copy()
+ f0_mel = 1127 * np.log(1 + f0 / 700)
+ f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - f0_mel_min) * 254 / (f0_mel_max - f0_mel_min) + 1
+ f0_mel[f0_mel <= 1] = 1
+ f0_mel[f0_mel > 255] = 255
+ f0_coarse = np.rint(f0_mel).astype(np.int)
+ return f0_coarse, f0bak#1-0
+
+ def vc(self,model,net_g,dv,audio0,pitch,pitchf,times):
+ feats = torch.from_numpy(audio0)
+ if(self.is_half==True):feats=feats.half()
+ else:feats=feats.float()
+ if feats.dim() == 2: # double channels
+ feats = feats.mean(-1)
+ assert feats.dim() == 1, feats.dim()
+ feats = feats.view(1, -1)
+ padding_mask = torch.BoolTensor(feats.shape).fill_(False)
+
+ inputs = {
+ "source": feats.to(self.device),
+ "padding_mask": padding_mask.to(self.device),
+ "output_layer": 9, # layer 9
+ }
+ t0 = ttime()
+ with torch.no_grad():
+ logits = model.extract_features(**inputs)
+ feats = model.final_proj(logits[0])
+ feats = F.interpolate(feats.permute(0, 2, 1), scale_factor=2).permute(0, 2, 1)
+ t1 = ttime()
+ p_len = audio0.shape[0]//self.window
+ if(feats.shape[1]<p_len):
+ p_len=feats.shape[1]
+ pitch=pitch[:,:p_len]
+ pitchf=pitchf[:,:p_len]
+ p_len=torch.LongTensor([p_len]).to(self.device)
+ with torch.no_grad():
+ audio1 = (net_g.infer(feats, p_len, pitch, pitchf, dv)[0][0, 0] * 32768).data.cpu().float().numpy().astype(np.int16)
+ del feats,p_len,padding_mask
+ torch.cuda.empty_cache()
+ t2 = ttime()
+ times[0] += (t1 - t0)
+ times[2] += (t2 - t1)
+ return audio1
+ def vc_km(self,model,net_g,dv,audio0,pitch,pitchf,times):
+ kmeans = KMeans(500)
+ def get_cluster_result(x):
+ """x: np.array [t, 256]"""
+ return kmeans.predict(x)
+ checkpoint = torch.load("lulu_contentvec_kmeans_500.pt")
+ kmeans.__dict__["n_features_in_"] = checkpoint["n_features_in_"]
+ kmeans.__dict__["_n_threads"] = checkpoint["_n_threads"]
+ kmeans.__dict__["cluster_centers_"] = checkpoint["cluster_centers_"]
+ feats = torch.from_numpy(audio0).float()
+ if feats.dim() == 2: # double channels
+ feats = feats.mean(-1)
+ assert feats.dim() == 1, feats.dim()
+ feats = feats.view(1, -1)
+ padding_mask = torch.BoolTensor(feats.shape).fill_(False)
+ inputs = {
+ "source": feats.half().to(self.device),
+ "padding_mask": padding_mask.to(self.device),
+ "output_layer": 9, # layer 9
+ }
+ torch.cuda.synchronize()
+ t0 = ttime()
+ with torch.no_grad():
+ logits = model.extract_features(**inputs)
+ feats = model.final_proj(logits[0])
+ feats = get_cluster_result(feats.cpu().numpy()[0].astype("float32"))
+ feats = torch.from_numpy(feats).to(self.device)
+ feats = F.interpolate(feats.half().unsqueeze(0).unsqueeze(0), scale_factor=2).long().squeeze(0)
+ t1 = ttime()
+ p_len = audio0.shape[0]//self.window
+ if(feats.shape[1]<p_len):
+ p_len=feats.shape[1]
+ pitch=pitch[:,:p_len]
+ pitchf=pitchf[:,:p_len]
+ p_len=torch.LongTensor([p_len]).to(self.device)
+ with torch.no_grad():
+ audio1 = (net_g.infer(feats, p_len, pitch, pitchf, dv)[0][0, 0] * 32768).data.cpu().float().numpy().astype(np.int16)
+ del feats,p_len,padding_mask
+ torch.cuda.empty_cache()
+ t2 = ttime()
+ times[0] += (t1 - t0)
+ times[2] += (t2 - t1)
+ return audio1
+
+ def pipeline(self,model,net_g,dv,audio,times,f0_up_key,f0_file=None):
+ audio_pad = np.pad(audio, (self.window // 2, self.window // 2), mode='reflect')
+ opt_ts = []
+ if(audio_pad.shape[0]>self.t_max):
+ audio_sum = np.zeros_like(audio)
+ for i in range(self.window): audio_sum += audio_pad[i:i - self.window]
+ for t in range(self.t_center, audio.shape[0],self.t_center):opt_ts.append(t - self.t_query + np.where(np.abs(audio_sum[t - self.t_query:t + self.t_query]) == np.abs(audio_sum[t - self.t_query:t + self.t_query]).min())[0][0])
+ s = 0
+ audio_opt=[]
+ t=None
+ t1=ttime()
+ audio_pad = np.pad(audio, (self.t_pad, self.t_pad), mode='reflect')
+ p_len=audio_pad.shape[0]//self.window
+ inp_f0=None
+ if(hasattr(f0_file,'name') ==True):
+ try:
+ with open(f0_file.name,"r")as f:
+ lines=f.read().strip("\n").split("\n")
+ inp_f0=[]
+ for line in lines:inp_f0.append([float(i)for i in line.split(",")])
+ inp_f0=np.array(inp_f0,dtype="float32")
+ except:
+ traceback.print_exc()
+ pitch, pitchf = self.get_f0(audio_pad, p_len, f0_up_key,inp_f0)
+
+ pitch = pitch[:p_len]
+ pitchf = pitchf[:p_len]
+ # if(inp_f0 is None):
+ # pitch = pitch[:p_len]
+ # pitchf = pitchf[:p_len]
+ # else:
+ # pitch=resize2d(pitch,p_len,is1=True)
+ # pitchf=resize2d(pitchf,p_len,is1=False)
+ pitch = torch.LongTensor(pitch).unsqueeze(0).to(self.device)
+ pitchf = torch.FloatTensor(pitchf).unsqueeze(0).to(self.device)
+ t2=ttime()
+ times[1] += (t2 - t1)
+ for t in opt_ts:
+ t=t//self.window*self.window
+ audio_opt.append(self.vc(model,net_g,dv,audio_pad[s:t+self.t_pad2+self.window],pitch[:,s//self.window:(t+self.t_pad2)//self.window],pitchf[:,s//self.window:(t+self.t_pad2)//self.window],times)[self.t_pad_tgt:-self.t_pad_tgt])
+ s = t
+ audio_opt.append(self.vc(model,net_g,dv,audio_pad[t:],pitch[:,t//self.window:]if t is not None else pitch,pitchf[:,t//self.window:]if t is not None else pitchf,times)[self.t_pad_tgt:-self.t_pad_tgt])
+ audio_opt=np.concatenate(audio_opt)
+ del pitch,pitchf
+ return audio_opt
+ def pipeline_km(self,model,net_g,dv,audio,times,f0_up_key,f0_file=None):
+ audio_pad = np.pad(audio, (self.window // 2, self.window // 2), mode='reflect')
+ opt_ts = []
+ if(audio_pad.shape[0]>self.t_max):
+ audio_sum = np.zeros_like(audio)
+ for i in range(self.window): audio_sum += audio_pad[i:i - self.window]
+ for t in range(self.t_center, audio.shape[0],self.t_center):opt_ts.append(t - self.t_query + np.where(np.abs(audio_sum[t - self.t_query:t + self.t_query]) == np.abs(audio_sum[t - self.t_query:t + self.t_query]).min())[0][0])
+ s = 0
+ audio_opt=[]
+ t=None
+ t1=ttime()
+ audio_pad = np.pad(audio, (self.t_pad, self.t_pad), mode='reflect')
+ p_len=audio_pad.shape[0]//self.window
+ inp_f0=None
+ if(hasattr(f0_file,'name') ==True):
+ try:
+ with open(f0_file.name,"r")as f:
+ lines=f.read().strip("\n").split("\n")
+ inp_f0=[]
+ for line in lines:inp_f0.append([float(i)for i in line.split(",")])
+ inp_f0=np.array(inp_f0,dtype="float32")
+ except:
+ traceback.print_exc()
+ pitch, pitchf = self.get_f0(audio_pad, p_len, f0_up_key,inp_f0)
+
+ pitch = pitch[:p_len]
+ pitchf = pitchf[:p_len]
+ # if(inp_f0 is None):
+ # pitch = pitch[:p_len]
+ # pitchf = pitchf[:p_len]
+ # else:
+ # pitch=resize2d(pitch,p_len,is1=True)
+ # pitchf=resize2d(pitchf,p_len,is1=False)
+ pitch = torch.LongTensor(pitch).unsqueeze(0).to(self.device)
+ pitchf = torch.FloatTensor(pitchf).unsqueeze(0).to(self.device)
+ t2=ttime()
+ times[1] += (t2 - t1)
+ for t in opt_ts:
+ t=t//self.window*self.window
+ audio_opt.append(self.vc_km(model,net_g,dv,audio_pad[s:t+self.t_pad2+self.window],pitch[:,s//self.window:(t+self.t_pad2)//self.window],pitchf[:,s//self.window:(t+self.t_pad2)//self.window],times)[self.t_pad_tgt:-self.t_pad_tgt])
+ s = t
+ audio_opt.append(self.vc_km(model,net_g,dv,audio_pad[t:],pitch[:,t//self.window:]if t is not None else pitch,pitchf[:,t//self.window:]if t is not None else pitchf,times)[self.t_pad_tgt:-self.t_pad_tgt])
+ audio_opt=np.concatenate(audio_opt)
+ del pitch,pitchf
+ return audio_opt
diff --git a/vits_vc_gpu.7z b/vits_vc_gpu.7z
new file mode 100644
index 0000000000000000000000000000000000000000..d872b0e0677f5a7a06f3208e2cc7708d2c8d6a0b
--- /dev/null
+++ b/vits_vc_gpu.7z
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bffb6b65240ca3d25b8b00b8975f149b61b5dd7ee6424ae5a34f4d1be9de2026
+size 1964422951
diff --git a/vits_vc_gpu_train.7z b/vits_vc_gpu_train.7z
new file mode 100644
index 0000000000000000000000000000000000000000..59fd310c2b8632dd1504ea19b7fb2044e876f02d
--- /dev/null
+++ b/vits_vc_gpu_train.7z
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58bc8c102873458a900c63ddfb41ed6a5e5b30937107e1bfdea1613f9b3ed768
+size 3141978949
diff --git "a/weights/\347\231\275\350\217\234357k.pt" "b/weights/\347\231\275\350\217\234357k.pt"
new file mode 100644
index 0000000000000000000000000000000000000000..875ab66d5a5f40a033059b018b8362319a55f513
--- /dev/null
+++ "b/weights/\347\231\275\350\217\234357k.pt"
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d309e8056dff08d33b30854839a9b9c36dfb612bf5971c070f552bde18158a55
+size 72645217
diff --git "a/\344\275\277\347\224\250\351\234\200\351\201\265\345\256\210\347\232\204\345\215\217\350\256\256-LICENSE.txt" "b/\344\275\277\347\224\250\351\234\200\351\201\265\345\256\210\347\232\204\345\215\217\350\256\256-LICENSE.txt"
new file mode 100644
index 0000000000000000000000000000000000000000..0fb0f44ba6eee54a81d1ec613d1dc4e1305f64c4
--- /dev/null
+++ "b/\344\275\277\347\224\250\351\234\200\351\201\265\345\256\210\347\232\204\345\215\217\350\256\256-LICENSE.txt"
@@ -0,0 +1,54 @@
+MIT License
+
+Copyright (c) 2022 lj1995
+
+本软件仅供研究使用，使用软件者、传播软件导出的声音者自负全责。如不认可该条款，则不能使用/引用软件包内所有代码和文件。
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+#################
+ContentVec
+https://github.com/auspicious3000/contentvec/blob/main/LICENSE
+MIT License
+#################
+VITS
+https://github.com/jaywalnut310/vits/blob/main/LICENSE
+MIT License
+#################
+HIFIGAN
+https://github.com/jik876/hifi-gan/blob/master/LICENSE
+MIT License
+#################
+gradio
+https://github.com/gradio-app/gradio/blob/main/LICENSE
+Apache License 2.0
+#################
+ffmpeg
+https://github.com/FFmpeg/FFmpeg/blob/master/COPYING.LGPLv3
+https://github.com/BtbN/FFmpeg-Builds/releases/download/autobuild-2021-02-28-12-32/ffmpeg-n4.3.2-160-gfbb9368226-win64-lgpl-4.3.zip
+LPGLv3 License
+MIT License
+#################
+ultimatevocalremovergui
+https://github.com/Anjok07/ultimatevocalremovergui/blob/master/LICENSE
+https://github.com/yang123qwe/vocal_separation_by_uvr5
+MIT License
+#################
+audio-slicer
+https://github.com/openvpi/audio-slicer/blob/main/LICENSE
+MIT License
\ No newline at end of file