diff --git a/lib/infer_pack/attentions.py b/lib/infer_pack/attentions.py
new file mode 100644
index 0000000000000000000000000000000000000000..05501be1871643f78dddbeaa529c96667031a8db
--- /dev/null
+++ b/lib/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 lib.infer_pack import commons
+from lib.infer_pack import modules
+from lib.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/lib/infer_pack/commons.py b/lib/infer_pack/commons.py
new file mode 100644
index 0000000000000000000000000000000000000000..54470986f37825b35d90d7efa7437d1c26b87215
--- /dev/null
+++ b/lib/infer_pack/commons.py
@@ -0,0 +1,166 @@
+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/lib/infer_pack/models.py b/lib/infer_pack/models.py
new file mode 100644
index 0000000000000000000000000000000000000000..3665d03bc0514a6ed07d3372ea24717dae1e0a65
--- /dev/null
+++ b/lib/infer_pack/models.py
@@ -0,0 +1,1142 @@
+import math, pdb, os
+from time import time as ttime
+import torch
+from torch import nn
+from torch.nn import functional as F
+from lib.infer_pack import modules
+from lib.infer_pack import attentions
+from lib.infer_pack import commons
+from lib.infer_pack.commons import init_weights, get_padding
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from lib.infer_pack.commons import init_weights
+import numpy as np
+from lib.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 TextEncoder768(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(768, 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:
+            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,
+        sr,
+        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()
+
+
+sr2sr = {
+    "32k": 32000,
+    "40k": 40000,
+    "48k": 48000,
+}
+
+
+class SynthesizerTrnMs256NSFsid(nn.Module):
+    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,
+        sr,
+        **kwargs
+    ):
+        super().__init__()
+        if type(sr) == type("strr"):
+            sr = sr2sr[sr]
+        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.hop_length = hop_length#
+        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=gin_channels,
+            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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, ds
+    ):  # 这里ds是id，[bs,1]
+        # print(1,pitch.shape)#[bs,t]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        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=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        # print(-1,pitchf.shape,ids_slice,self.segment_size,self.hop_length,self.segment_size//self.hop_length)
+        pitchf = commons.slice_segments2(pitchf, ids_slice, self.segment_size)
+        # print(-2,pitchf.shape,z_slice.shape)
+        o = self.dec(z_slice, pitchf, g=g)
+        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, sid, rate=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        if rate:
+            head = int(z_p.shape[2] * rate)
+            z_p = z_p[:, :, -head:]
+            x_mask = x_mask[:, :, -head:]
+            nsff0 = nsff0[:, -head:]
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec(z * x_mask, nsff0, g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class SynthesizerTrnMs768NSFsid(nn.Module):
+    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,
+        sr,
+        **kwargs
+    ):
+        super().__init__()
+        if type(sr) == type("strr"):
+            sr = sr2sr[sr]
+        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.hop_length = hop_length#
+        self.spk_embed_dim = spk_embed_dim
+        self.enc_p = TextEncoder768(
+            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=gin_channels,
+            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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, ds
+    ):  # 这里ds是id，[bs,1]
+        # print(1,pitch.shape)#[bs,t]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        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=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        # print(-1,pitchf.shape,ids_slice,self.segment_size,self.hop_length,self.segment_size//self.hop_length)
+        pitchf = commons.slice_segments2(pitchf, ids_slice, self.segment_size)
+        # print(-2,pitchf.shape,z_slice.shape)
+        o = self.dec(z_slice, pitchf, g=g)
+        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, sid, rate=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        if rate:
+            head = int(z_p.shape[2] * rate)
+            z_p = z_p[:, :, -head:]
+            x_mask = x_mask[:, :, -head:]
+            nsff0 = nsff0[:, -head:]
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec(z * x_mask, nsff0, g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class SynthesizerTrnMs256NSFsid_nono(nn.Module):
+    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,
+        sr=None,
+        **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.hop_length = hop_length#
+        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,
+            f0=False,
+        )
+        self.dec = Generator(
+            inter_channels,
+            resblock,
+            resblock_kernel_sizes,
+            resblock_dilation_sizes,
+            upsample_rates,
+            upsample_initial_channel,
+            upsample_kernel_sizes,
+            gin_channels=gin_channels,
+        )
+        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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, y, y_lengths, ds):  # 这里ds是id，[bs,1]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        o = self.dec(z_slice, g=g)
+        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, sid, rate=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        if rate:
+            head = int(z_p.shape[2] * rate)
+            z_p = z_p[:, :, -head:]
+            x_mask = x_mask[:, :, -head:]
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec(z * x_mask, g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class SynthesizerTrnMs768NSFsid_nono(nn.Module):
+    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,
+        sr=None,
+        **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.hop_length = hop_length#
+        self.spk_embed_dim = spk_embed_dim
+        self.enc_p = TextEncoder768(
+            inter_channels,
+            hidden_channels,
+            filter_channels,
+            n_heads,
+            n_layers,
+            kernel_size,
+            p_dropout,
+            f0=False,
+        )
+        self.dec = Generator(
+            inter_channels,
+            resblock,
+            resblock_kernel_sizes,
+            resblock_dilation_sizes,
+            upsample_rates,
+            upsample_initial_channel,
+            upsample_kernel_sizes,
+            gin_channels=gin_channels,
+        )
+        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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, y, y_lengths, ds):  # 这里ds是id，[bs,1]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        o = self.dec(z_slice, g=g)
+        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, sid, rate=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        if rate:
+            head = int(z_p.shape[2] * rate)
+            z_p = z_p[:, :, -head:]
+            x_mask = x_mask[:, :, -head:]
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec(z * x_mask, g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class MultiPeriodDiscriminator(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminator, self).__init__()
+        periods = [2, 3, 5, 7, 11, 17]
+        # periods = [3, 5, 7, 11, 17, 23, 37]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []  #
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            # for j in range(len(fmap_r)):
+            #     print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class MultiPeriodDiscriminatorV2(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminatorV2, self).__init__()
+        # periods = [2, 3, 5, 7, 11, 17]
+        periods = [2, 3, 5, 7, 11, 17, 23, 37]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []  #
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            # for j in range(len(fmap_r)):
+            #     print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class DiscriminatorS(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(DiscriminatorS, self).__init__()
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList(
+            [
+                norm_f(Conv1d(1, 16, 15, 1, padding=7)),
+                norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)),
+                norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)),
+                norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
+                norm_f(Conv1d(1024, 1024, 41, 4, groups=256, padding=20)),
+                norm_f(Conv1d(1024, 1024, 5, 1, padding=2)),
+            ]
+        )
+        self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1))
+
+    def forward(self, x):
+        fmap = []
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class DiscriminatorP(torch.nn.Module):
+    def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
+        super(DiscriminatorP, self).__init__()
+        self.period = period
+        self.use_spectral_norm = use_spectral_norm
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList(
+            [
+                norm_f(
+                    Conv2d(
+                        1,
+                        32,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        32,
+                        128,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        128,
+                        512,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        512,
+                        1024,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        1024,
+                        1024,
+                        (kernel_size, 1),
+                        1,
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+            ]
+        )
+        self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))
+
+    def forward(self, x):
+        fmap = []
+
+        # 1d to 2d
+        b, c, t = x.shape
+        if t % self.period != 0:  # pad first
+            n_pad = self.period - (t % self.period)
+            x = F.pad(x, (0, n_pad), "reflect")
+            t = t + n_pad
+        x = x.view(b, c, t // self.period, self.period)
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
diff --git a/lib/infer_pack/models_dml.py b/lib/infer_pack/models_dml.py
new file mode 100644
index 0000000000000000000000000000000000000000..958d7b29259763d2fea94caf8ba7e314c4a77d05
--- /dev/null
+++ b/lib/infer_pack/models_dml.py
@@ -0,0 +1,1124 @@
+import math, pdb, os
+from time import time as ttime
+import torch
+from torch import nn
+from torch.nn import functional as F
+from lib.infer_pack import modules
+from lib.infer_pack import attentions
+from lib.infer_pack import commons
+from lib.infer_pack.commons import init_weights, get_padding
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from lib.infer_pack.commons import init_weights
+import numpy as np
+from lib.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 TextEncoder768(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(768, 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.float()
+
+    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:
+            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,
+        sr,
+        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()
+
+
+sr2sr = {
+    "32k": 32000,
+    "40k": 40000,
+    "48k": 48000,
+}
+
+
+class SynthesizerTrnMs256NSFsid(nn.Module):
+    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,
+        sr,
+        **kwargs
+    ):
+        super().__init__()
+        if type(sr) == type("strr"):
+            sr = sr2sr[sr]
+        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.hop_length = hop_length#
+        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=gin_channels,
+            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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, ds
+    ):  # 这里ds是id，[bs,1]
+        # print(1,pitch.shape)#[bs,t]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        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=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        # print(-1,pitchf.shape,ids_slice,self.segment_size,self.hop_length,self.segment_size//self.hop_length)
+        pitchf = commons.slice_segments2(pitchf, ids_slice, self.segment_size)
+        # print(-2,pitchf.shape,z_slice.shape)
+        o = self.dec(z_slice, pitchf, g=g)
+        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, sid, max_len=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec((z * x_mask)[:, :, :max_len], nsff0, g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class SynthesizerTrnMs768NSFsid(nn.Module):
+    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,
+        sr,
+        **kwargs
+    ):
+        super().__init__()
+        if type(sr) == type("strr"):
+            sr = sr2sr[sr]
+        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.hop_length = hop_length#
+        self.spk_embed_dim = spk_embed_dim
+        self.enc_p = TextEncoder768(
+            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=gin_channels,
+            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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, ds
+    ):  # 这里ds是id，[bs,1]
+        # print(1,pitch.shape)#[bs,t]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        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=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        # print(-1,pitchf.shape,ids_slice,self.segment_size,self.hop_length,self.segment_size//self.hop_length)
+        pitchf = commons.slice_segments2(pitchf, ids_slice, self.segment_size)
+        # print(-2,pitchf.shape,z_slice.shape)
+        o = self.dec(z_slice, pitchf, g=g)
+        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, sid, max_len=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec((z * x_mask)[:, :, :max_len], nsff0, g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class SynthesizerTrnMs256NSFsid_nono(nn.Module):
+    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,
+        sr=None,
+        **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.hop_length = hop_length#
+        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,
+            f0=False,
+        )
+        self.dec = Generator(
+            inter_channels,
+            resblock,
+            resblock_kernel_sizes,
+            resblock_dilation_sizes,
+            upsample_rates,
+            upsample_initial_channel,
+            upsample_kernel_sizes,
+            gin_channels=gin_channels,
+        )
+        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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, y, y_lengths, ds):  # 这里ds是id，[bs,1]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        o = self.dec(z_slice, g=g)
+        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, sid, max_len=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec((z * x_mask)[:, :, :max_len], g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class SynthesizerTrnMs768NSFsid_nono(nn.Module):
+    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,
+        sr=None,
+        **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.hop_length = hop_length#
+        self.spk_embed_dim = spk_embed_dim
+        self.enc_p = TextEncoder768(
+            inter_channels,
+            hidden_channels,
+            filter_channels,
+            n_heads,
+            n_layers,
+            kernel_size,
+            p_dropout,
+            f0=False,
+        )
+        self.dec = Generator(
+            inter_channels,
+            resblock,
+            resblock_kernel_sizes,
+            resblock_dilation_sizes,
+            upsample_rates,
+            upsample_initial_channel,
+            upsample_kernel_sizes,
+            gin_channels=gin_channels,
+        )
+        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.Embedding(self.spk_embed_dim, gin_channels)
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    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, y, y_lengths, ds):  # 这里ds是id，[bs,1]
+        g = self.emb_g(ds).unsqueeze(-1)  # [b, 256, 1]##1是t，广播的
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+        z_p = self.flow(z, y_mask, g=g)
+        z_slice, ids_slice = commons.rand_slice_segments(
+            z, y_lengths, self.segment_size
+        )
+        o = self.dec(z_slice, g=g)
+        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, sid, max_len=None):
+        g = self.emb_g(sid).unsqueeze(-1)
+        m_p, logs_p, x_mask = self.enc_p(phone, None, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * torch.randn_like(m_p) * 0.66666) * x_mask
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec((z * x_mask)[:, :, :max_len], g=g)
+        return o, x_mask, (z, z_p, m_p, logs_p)
+
+
+class MultiPeriodDiscriminator(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminator, self).__init__()
+        periods = [2, 3, 5, 7, 11, 17]
+        # periods = [3, 5, 7, 11, 17, 23, 37]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []  #
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            # for j in range(len(fmap_r)):
+            #     print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class MultiPeriodDiscriminatorV2(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminatorV2, self).__init__()
+        # periods = [2, 3, 5, 7, 11, 17]
+        periods = [2, 3, 5, 7, 11, 17, 23, 37]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []  #
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            # for j in range(len(fmap_r)):
+            #     print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class DiscriminatorS(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(DiscriminatorS, self).__init__()
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList(
+            [
+                norm_f(Conv1d(1, 16, 15, 1, padding=7)),
+                norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)),
+                norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)),
+                norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
+                norm_f(Conv1d(1024, 1024, 41, 4, groups=256, padding=20)),
+                norm_f(Conv1d(1024, 1024, 5, 1, padding=2)),
+            ]
+        )
+        self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1))
+
+    def forward(self, x):
+        fmap = []
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class DiscriminatorP(torch.nn.Module):
+    def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
+        super(DiscriminatorP, self).__init__()
+        self.period = period
+        self.use_spectral_norm = use_spectral_norm
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList(
+            [
+                norm_f(
+                    Conv2d(
+                        1,
+                        32,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        32,
+                        128,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        128,
+                        512,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        512,
+                        1024,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        1024,
+                        1024,
+                        (kernel_size, 1),
+                        1,
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+            ]
+        )
+        self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))
+
+    def forward(self, x):
+        fmap = []
+
+        # 1d to 2d
+        b, c, t = x.shape
+        if t % self.period != 0:  # pad first
+            n_pad = self.period - (t % self.period)
+            x = F.pad(x, (0, n_pad), "reflect")
+            t = t + n_pad
+        x = x.view(b, c, t // self.period, self.period)
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
diff --git a/lib/infer_pack/models_onnx.py b/lib/infer_pack/models_onnx.py
new file mode 100644
index 0000000000000000000000000000000000000000..963e67b29f828e9fdd096397952054fe77cf3d10
--- /dev/null
+++ b/lib/infer_pack/models_onnx.py
@@ -0,0 +1,819 @@
+import math, pdb, os
+from time import time as ttime
+import torch
+from torch import nn
+from torch.nn import functional as F
+from lib.infer_pack import modules
+from lib.infer_pack import attentions
+from lib.infer_pack import commons
+from lib.infer_pack.commons import init_weights, get_padding
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from lib.infer_pack.commons import init_weights
+import numpy as np
+from lib.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 TextEncoder768(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(768, 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:
+            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,
+        sr,
+        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()
+
+
+sr2sr = {
+    "32k": 32000,
+    "40k": 40000,
+    "48k": 48000,
+}
+
+
+class SynthesizerTrnMsNSFsidM(nn.Module):
+    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,
+        sr,
+        version,
+        **kwargs
+    ):
+        super().__init__()
+        if type(sr) == type("strr"):
+            sr = sr2sr[sr]
+        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.hop_length = hop_length#
+        self.spk_embed_dim = spk_embed_dim
+        if version == "v1":
+            self.enc_p = TextEncoder256(
+                inter_channels,
+                hidden_channels,
+                filter_channels,
+                n_heads,
+                n_layers,
+                kernel_size,
+                p_dropout,
+            )
+        else:
+            self.enc_p = TextEncoder768(
+                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=gin_channels,
+            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.Embedding(self.spk_embed_dim, gin_channels)
+        self.speaker_map = None
+        print("gin_channels:", gin_channels, "self.spk_embed_dim:", self.spk_embed_dim)
+
+    def remove_weight_norm(self):
+        self.dec.remove_weight_norm()
+        self.flow.remove_weight_norm()
+        self.enc_q.remove_weight_norm()
+
+    def construct_spkmixmap(self, n_speaker):
+        self.speaker_map = torch.zeros((n_speaker, 1, 1, self.gin_channels))
+        for i in range(n_speaker):
+            self.speaker_map[i] = self.emb_g(torch.LongTensor([[i]]))
+        self.speaker_map = self.speaker_map.unsqueeze(0)
+
+    def forward(self, phone, phone_lengths, pitch, nsff0, g, rnd, max_len=None):
+        if self.speaker_map is not None:  # [N, S]  *  [S, B, 1, H]
+            g = g.reshape((g.shape[0], g.shape[1], 1, 1, 1))  # [N, S, B, 1, 1]
+            g = g * self.speaker_map  # [N, S, B, 1, H]
+            g = torch.sum(g, dim=1)  # [N, 1, B, 1, H]
+            g = g.transpose(0, -1).transpose(0, -2).squeeze(0)  # [B, H, N]
+        else:
+            g = g.unsqueeze(0)
+            g = self.emb_g(g).transpose(1, 2)
+
+        m_p, logs_p, x_mask = self.enc_p(phone, pitch, phone_lengths)
+        z_p = (m_p + torch.exp(logs_p) * rnd) * x_mask
+        z = self.flow(z_p, x_mask, g=g, reverse=True)
+        o = self.dec((z * x_mask)[:, :, :max_len], nsff0, g=g)
+        return o
+
+
+class MultiPeriodDiscriminator(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminator, self).__init__()
+        periods = [2, 3, 5, 7, 11, 17]
+        # periods = [3, 5, 7, 11, 17, 23, 37]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []  #
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            # for j in range(len(fmap_r)):
+            #     print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class MultiPeriodDiscriminatorV2(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminatorV2, self).__init__()
+        # periods = [2, 3, 5, 7, 11, 17]
+        periods = [2, 3, 5, 7, 11, 17, 23, 37]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [
+            DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods
+        ]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []  #
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            # for j in range(len(fmap_r)):
+            #     print(i,j,y.shape,y_hat.shape,fmap_r[j].shape,fmap_g[j].shape)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+class DiscriminatorS(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(DiscriminatorS, self).__init__()
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList(
+            [
+                norm_f(Conv1d(1, 16, 15, 1, padding=7)),
+                norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)),
+                norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)),
+                norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
+                norm_f(Conv1d(1024, 1024, 41, 4, groups=256, padding=20)),
+                norm_f(Conv1d(1024, 1024, 5, 1, padding=2)),
+            ]
+        )
+        self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1))
+
+    def forward(self, x):
+        fmap = []
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class DiscriminatorP(torch.nn.Module):
+    def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
+        super(DiscriminatorP, self).__init__()
+        self.period = period
+        self.use_spectral_norm = use_spectral_norm
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList(
+            [
+                norm_f(
+                    Conv2d(
+                        1,
+                        32,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        32,
+                        128,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        128,
+                        512,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        512,
+                        1024,
+                        (kernel_size, 1),
+                        (stride, 1),
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+                norm_f(
+                    Conv2d(
+                        1024,
+                        1024,
+                        (kernel_size, 1),
+                        1,
+                        padding=(get_padding(kernel_size, 1), 0),
+                    )
+                ),
+            ]
+        )
+        self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))
+
+    def forward(self, x):
+        fmap = []
+
+        # 1d to 2d
+        b, c, t = x.shape
+        if t % self.period != 0:  # pad first
+            n_pad = self.period - (t % self.period)
+            x = F.pad(x, (0, n_pad), "reflect")
+            t = t + n_pad
+        x = x.view(b, c, t // self.period, self.period)
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
diff --git a/lib/infer_pack/modules.py b/lib/infer_pack/modules.py
new file mode 100644
index 0000000000000000000000000000000000000000..c83289df7c79a4810dacd15c050148544ba0b6a9
--- /dev/null
+++ b/lib/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 lib.infer_pack import commons
+from lib.infer_pack.commons import init_weights, get_padding
+from lib.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/lib/infer_pack/modules/F0Predictor/DioF0Predictor.py b/lib/infer_pack/modules/F0Predictor/DioF0Predictor.py
new file mode 100644
index 0000000000000000000000000000000000000000..ee3171bcb7c4a5066560723108b56e055f18be45
--- /dev/null
+++ b/lib/infer_pack/modules/F0Predictor/DioF0Predictor.py
@@ -0,0 +1,90 @@
+from lib.infer_pack.modules.F0Predictor.F0Predictor import F0Predictor
+import pyworld
+import numpy as np
+
+
+class DioF0Predictor(F0Predictor):
+    def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
+        self.hop_length = hop_length
+        self.f0_min = f0_min
+        self.f0_max = f0_max
+        self.sampling_rate = sampling_rate
+
+    def interpolate_f0(self, f0):
+        """
+        对F0进行插值处理
+        """
+
+        data = np.reshape(f0, (f0.size, 1))
+
+        vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
+        vuv_vector[data > 0.0] = 1.0
+        vuv_vector[data <= 0.0] = 0.0
+
+        ip_data = data
+
+        frame_number = data.size
+        last_value = 0.0
+        for i in range(frame_number):
+            if data[i] <= 0.0:
+                j = i + 1
+                for j in range(i + 1, frame_number):
+                    if data[j] > 0.0:
+                        break
+                if j < frame_number - 1:
+                    if last_value > 0.0:
+                        step = (data[j] - data[i - 1]) / float(j - i)
+                        for k in range(i, j):
+                            ip_data[k] = data[i - 1] + step * (k - i + 1)
+                    else:
+                        for k in range(i, j):
+                            ip_data[k] = data[j]
+                else:
+                    for k in range(i, frame_number):
+                        ip_data[k] = last_value
+            else:
+                ip_data[i] = data[i]  # 这里可能存在一个没有必要的拷贝
+                last_value = data[i]
+
+        return ip_data[:, 0], vuv_vector[:, 0]
+
+    def resize_f0(self, x, target_len):
+        source = np.array(x)
+        source[source < 0.001] = np.nan
+        target = np.interp(
+            np.arange(0, len(source) * target_len, len(source)) / target_len,
+            np.arange(0, len(source)),
+            source,
+        )
+        res = np.nan_to_num(target)
+        return res
+
+    def compute_f0(self, wav, p_len=None):
+        if p_len is None:
+            p_len = wav.shape[0] // self.hop_length
+        f0, t = pyworld.dio(
+            wav.astype(np.double),
+            fs=self.sampling_rate,
+            f0_floor=self.f0_min,
+            f0_ceil=self.f0_max,
+            frame_period=1000 * self.hop_length / self.sampling_rate,
+        )
+        f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.sampling_rate)
+        for index, pitch in enumerate(f0):
+            f0[index] = round(pitch, 1)
+        return self.interpolate_f0(self.resize_f0(f0, p_len))[0]
+
+    def compute_f0_uv(self, wav, p_len=None):
+        if p_len is None:
+            p_len = wav.shape[0] // self.hop_length
+        f0, t = pyworld.dio(
+            wav.astype(np.double),
+            fs=self.sampling_rate,
+            f0_floor=self.f0_min,
+            f0_ceil=self.f0_max,
+            frame_period=1000 * self.hop_length / self.sampling_rate,
+        )
+        f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.sampling_rate)
+        for index, pitch in enumerate(f0):
+            f0[index] = round(pitch, 1)
+        return self.interpolate_f0(self.resize_f0(f0, p_len))
diff --git a/lib/infer_pack/modules/F0Predictor/F0Predictor.py b/lib/infer_pack/modules/F0Predictor/F0Predictor.py
new file mode 100644
index 0000000000000000000000000000000000000000..f56e49e7f0e6eab3babf0711cae2933371b9f9cc
--- /dev/null
+++ b/lib/infer_pack/modules/F0Predictor/F0Predictor.py
@@ -0,0 +1,16 @@
+class F0Predictor(object):
+    def compute_f0(self, wav, p_len):
+        """
+        input: wav:[signal_length]
+               p_len:int
+        output: f0:[signal_length//hop_length]
+        """
+        pass
+
+    def compute_f0_uv(self, wav, p_len):
+        """
+        input: wav:[signal_length]
+               p_len:int
+        output: f0:[signal_length//hop_length],uv:[signal_length//hop_length]
+        """
+        pass
diff --git a/lib/infer_pack/modules/F0Predictor/HarvestF0Predictor.py b/lib/infer_pack/modules/F0Predictor/HarvestF0Predictor.py
new file mode 100644
index 0000000000000000000000000000000000000000..b412ba2814e114ca7bb00b6fd6ef217f63d788a3
--- /dev/null
+++ b/lib/infer_pack/modules/F0Predictor/HarvestF0Predictor.py
@@ -0,0 +1,86 @@
+from lib.infer_pack.modules.F0Predictor.F0Predictor import F0Predictor
+import pyworld
+import numpy as np
+
+
+class HarvestF0Predictor(F0Predictor):
+    def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
+        self.hop_length = hop_length
+        self.f0_min = f0_min
+        self.f0_max = f0_max
+        self.sampling_rate = sampling_rate
+
+    def interpolate_f0(self, f0):
+        """
+        对F0进行插值处理
+        """
+
+        data = np.reshape(f0, (f0.size, 1))
+
+        vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
+        vuv_vector[data > 0.0] = 1.0
+        vuv_vector[data <= 0.0] = 0.0
+
+        ip_data = data
+
+        frame_number = data.size
+        last_value = 0.0
+        for i in range(frame_number):
+            if data[i] <= 0.0:
+                j = i + 1
+                for j in range(i + 1, frame_number):
+                    if data[j] > 0.0:
+                        break
+                if j < frame_number - 1:
+                    if last_value > 0.0:
+                        step = (data[j] - data[i - 1]) / float(j - i)
+                        for k in range(i, j):
+                            ip_data[k] = data[i - 1] + step * (k - i + 1)
+                    else:
+                        for k in range(i, j):
+                            ip_data[k] = data[j]
+                else:
+                    for k in range(i, frame_number):
+                        ip_data[k] = last_value
+            else:
+                ip_data[i] = data[i]  # 这里可能存在一个没有必要的拷贝
+                last_value = data[i]
+
+        return ip_data[:, 0], vuv_vector[:, 0]
+
+    def resize_f0(self, x, target_len):
+        source = np.array(x)
+        source[source < 0.001] = np.nan
+        target = np.interp(
+            np.arange(0, len(source) * target_len, len(source)) / target_len,
+            np.arange(0, len(source)),
+            source,
+        )
+        res = np.nan_to_num(target)
+        return res
+
+    def compute_f0(self, wav, p_len=None):
+        if p_len is None:
+            p_len = wav.shape[0] // self.hop_length
+        f0, t = pyworld.harvest(
+            wav.astype(np.double),
+            fs=self.hop_length,
+            f0_ceil=self.f0_max,
+            f0_floor=self.f0_min,
+            frame_period=1000 * self.hop_length / self.sampling_rate,
+        )
+        f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.fs)
+        return self.interpolate_f0(self.resize_f0(f0, p_len))[0]
+
+    def compute_f0_uv(self, wav, p_len=None):
+        if p_len is None:
+            p_len = wav.shape[0] // self.hop_length
+        f0, t = pyworld.harvest(
+            wav.astype(np.double),
+            fs=self.sampling_rate,
+            f0_floor=self.f0_min,
+            f0_ceil=self.f0_max,
+            frame_period=1000 * self.hop_length / self.sampling_rate,
+        )
+        f0 = pyworld.stonemask(wav.astype(np.double), f0, t, self.sampling_rate)
+        return self.interpolate_f0(self.resize_f0(f0, p_len))
diff --git a/lib/infer_pack/modules/F0Predictor/PMF0Predictor.py b/lib/infer_pack/modules/F0Predictor/PMF0Predictor.py
new file mode 100644
index 0000000000000000000000000000000000000000..b2c592527a5966e6f8e79e8c52dc5b414246dcc6
--- /dev/null
+++ b/lib/infer_pack/modules/F0Predictor/PMF0Predictor.py
@@ -0,0 +1,97 @@
+from lib.infer_pack.modules.F0Predictor.F0Predictor import F0Predictor
+import parselmouth
+import numpy as np
+
+
+class PMF0Predictor(F0Predictor):
+    def __init__(self, hop_length=512, f0_min=50, f0_max=1100, sampling_rate=44100):
+        self.hop_length = hop_length
+        self.f0_min = f0_min
+        self.f0_max = f0_max
+        self.sampling_rate = sampling_rate
+
+    def interpolate_f0(self, f0):
+        """
+        对F0进行插值处理
+        """
+
+        data = np.reshape(f0, (f0.size, 1))
+
+        vuv_vector = np.zeros((data.size, 1), dtype=np.float32)
+        vuv_vector[data > 0.0] = 1.0
+        vuv_vector[data <= 0.0] = 0.0
+
+        ip_data = data
+
+        frame_number = data.size
+        last_value = 0.0
+        for i in range(frame_number):
+            if data[i] <= 0.0:
+                j = i + 1
+                for j in range(i + 1, frame_number):
+                    if data[j] > 0.0:
+                        break
+                if j < frame_number - 1:
+                    if last_value > 0.0:
+                        step = (data[j] - data[i - 1]) / float(j - i)
+                        for k in range(i, j):
+                            ip_data[k] = data[i - 1] + step * (k - i + 1)
+                    else:
+                        for k in range(i, j):
+                            ip_data[k] = data[j]
+                else:
+                    for k in range(i, frame_number):
+                        ip_data[k] = last_value
+            else:
+                ip_data[i] = data[i]  # 这里可能存在一个没有必要的拷贝
+                last_value = data[i]
+
+        return ip_data[:, 0], vuv_vector[:, 0]
+
+    def compute_f0(self, wav, p_len=None):
+        x = wav
+        if p_len is None:
+            p_len = x.shape[0] // self.hop_length
+        else:
+            assert abs(p_len - x.shape[0] // self.hop_length) < 4, "pad length error"
+        time_step = self.hop_length / self.sampling_rate * 1000
+        f0 = (
+            parselmouth.Sound(x, self.sampling_rate)
+            .to_pitch_ac(
+                time_step=time_step / 1000,
+                voicing_threshold=0.6,
+                pitch_floor=self.f0_min,
+                pitch_ceiling=self.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, uv = self.interpolate_f0(f0)
+        return f0
+
+    def compute_f0_uv(self, wav, p_len=None):
+        x = wav
+        if p_len is None:
+            p_len = x.shape[0] // self.hop_length
+        else:
+            assert abs(p_len - x.shape[0] // self.hop_length) < 4, "pad length error"
+        time_step = self.hop_length / self.sampling_rate * 1000
+        f0 = (
+            parselmouth.Sound(x, self.sampling_rate)
+            .to_pitch_ac(
+                time_step=time_step / 1000,
+                voicing_threshold=0.6,
+                pitch_floor=self.f0_min,
+                pitch_ceiling=self.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, uv = self.interpolate_f0(f0)
+        return f0, uv
diff --git a/lib/infer_pack/modules/F0Predictor/__init__.py b/lib/infer_pack/modules/F0Predictor/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/lib/infer_pack/onnx_inference.py b/lib/infer_pack/onnx_inference.py
new file mode 100644
index 0000000000000000000000000000000000000000..6517853be49e61c427cf7cd9b5ed203f6d5f367e
--- /dev/null
+++ b/lib/infer_pack/onnx_inference.py
@@ -0,0 +1,145 @@
+import onnxruntime
+import librosa
+import numpy as np
+import soundfile
+
+
+class ContentVec:
+    def __init__(self, vec_path="pretrained/vec-768-layer-12.onnx", device=None):
+        print("load model(s) from {}".format(vec_path))
+        if device == "cpu" or device is None:
+            providers = ["CPUExecutionProvider"]
+        elif device == "cuda":
+            providers = ["CUDAExecutionProvider", "CPUExecutionProvider"]
+        elif device == "dml":
+            providers = ["DmlExecutionProvider"]
+        else:
+            raise RuntimeError("Unsportted Device")
+        self.model = onnxruntime.InferenceSession(vec_path, providers=providers)
+
+    def __call__(self, wav):
+        return self.forward(wav)
+
+    def forward(self, wav):
+        feats = wav
+        if feats.ndim == 2:  # double channels
+            feats = feats.mean(-1)
+        assert feats.ndim == 1, feats.ndim
+        feats = np.expand_dims(np.expand_dims(feats, 0), 0)
+        onnx_input = {self.model.get_inputs()[0].name: feats}
+        logits = self.model.run(None, onnx_input)[0]
+        return logits.transpose(0, 2, 1)
+
+
+def get_f0_predictor(f0_predictor, hop_length, sampling_rate, **kargs):
+    if f0_predictor == "pm":
+        from lib.infer_pack.modules.F0Predictor.PMF0Predictor import PMF0Predictor
+
+        f0_predictor_object = PMF0Predictor(
+            hop_length=hop_length, sampling_rate=sampling_rate
+        )
+    elif f0_predictor == "harvest":
+        from lib.infer_pack.modules.F0Predictor.HarvestF0Predictor import (
+            HarvestF0Predictor,
+        )
+
+        f0_predictor_object = HarvestF0Predictor(
+            hop_length=hop_length, sampling_rate=sampling_rate
+        )
+    elif f0_predictor == "dio":
+        from lib.infer_pack.modules.F0Predictor.DioF0Predictor import DioF0Predictor
+
+        f0_predictor_object = DioF0Predictor(
+            hop_length=hop_length, sampling_rate=sampling_rate
+        )
+    else:
+        raise Exception("Unknown f0 predictor")
+    return f0_predictor_object
+
+
+class OnnxRVC:
+    def __init__(
+        self,
+        model_path,
+        sr=40000,
+        hop_size=512,
+        vec_path="vec-768-layer-12",
+        device="cpu",
+    ):
+        vec_path = f"pretrained/{vec_path}.onnx"
+        self.vec_model = ContentVec(vec_path, device)
+        if device == "cpu" or device is None:
+            providers = ["CPUExecutionProvider"]
+        elif device == "cuda":
+            providers = ["CUDAExecutionProvider", "CPUExecutionProvider"]
+        elif device == "dml":
+            providers = ["DmlExecutionProvider"]
+        else:
+            raise RuntimeError("Unsportted Device")
+        self.model = onnxruntime.InferenceSession(model_path, providers=providers)
+        self.sampling_rate = sr
+        self.hop_size = hop_size
+
+    def forward(self, hubert, hubert_length, pitch, pitchf, ds, rnd):
+        onnx_input = {
+            self.model.get_inputs()[0].name: hubert,
+            self.model.get_inputs()[1].name: hubert_length,
+            self.model.get_inputs()[2].name: pitch,
+            self.model.get_inputs()[3].name: pitchf,
+            self.model.get_inputs()[4].name: ds,
+            self.model.get_inputs()[5].name: rnd,
+        }
+        return (self.model.run(None, onnx_input)[0] * 32767).astype(np.int16)
+
+    def inference(
+        self,
+        raw_path,
+        sid,
+        f0_method="dio",
+        f0_up_key=0,
+        pad_time=0.5,
+        cr_threshold=0.02,
+    ):
+        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_predictor = get_f0_predictor(
+            f0_method,
+            hop_length=self.hop_size,
+            sampling_rate=self.sampling_rate,
+            threshold=cr_threshold,
+        )
+        wav, sr = librosa.load(raw_path, sr=self.sampling_rate)
+        org_length = len(wav)
+        if org_length / sr > 50.0:
+            raise RuntimeError("Reached Max Length")
+
+        wav16k = librosa.resample(wav, orig_sr=self.sampling_rate, target_sr=16000)
+        wav16k = wav16k
+
+        hubert = self.vec_model(wav16k)
+        hubert = np.repeat(hubert, 2, axis=2).transpose(0, 2, 1).astype(np.float32)
+        hubert_length = hubert.shape[1]
+
+        pitchf = f0_predictor.compute_f0(wav, hubert_length)
+        pitchf = pitchf * 2 ** (f0_up_key / 12)
+        pitch = pitchf.copy()
+        f0_mel = 1127 * np.log(1 + pitch / 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
+        pitch = np.rint(f0_mel).astype(np.int64)
+
+        pitchf = pitchf.reshape(1, len(pitchf)).astype(np.float32)
+        pitch = pitch.reshape(1, len(pitch))
+        ds = np.array([sid]).astype(np.int64)
+
+        rnd = np.random.randn(1, 192, hubert_length).astype(np.float32)
+        hubert_length = np.array([hubert_length]).astype(np.int64)
+
+        out_wav = self.forward(hubert, hubert_length, pitch, pitchf, ds, rnd).squeeze()
+        out_wav = np.pad(out_wav, (0, 2 * self.hop_size), "constant")
+        return out_wav[0:org_length]
diff --git a/lib/infer_pack/transforms.py b/lib/infer_pack/transforms.py
new file mode 100644
index 0000000000000000000000000000000000000000..a11f799e023864ff7082c1f49c0cc18351a13b47
--- /dev/null
+++ b/lib/infer_pack/transforms.py
@@ -0,0 +1,209 @@
+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=1.0,
+    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=1.0,
+    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=0.0,
+    right=1.0,
+    bottom=0.0,
+    top=1.0,
+    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/weights/aether-jp/added_IVF865_Flat_nprobe_1_aether-jp_v2.index b/weights/aether-jp/added_IVF865_Flat_nprobe_1_aether-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..2e56fe4ed75072d46da8b3c899cc66088c555874
--- /dev/null
+++ b/weights/aether-jp/added_IVF865_Flat_nprobe_1_aether-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98f6891416fb85d2040e9e6fec7f08f591870eb04ef6e9f7cc57c81f3282b6b8
+size 106580459
diff --git a/weights/aether-jp/aether-jp.pth b/weights/aether-jp/aether-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..92bd5705dbcb4d1b716024b2028c4c26add3df35
--- /dev/null
+++ b/weights/aether-jp/aether-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:615a7105d7c5f307fdc7174856f0bd53a473a1ea03d21749a9903dacaec5d87c
+size 57589001
diff --git a/weights/aether-jp/cover.png b/weights/aether-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..29530da623d5d5fe7096880cea767104dcb21b5c
--- /dev/null
+++ b/weights/aether-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eec11b3470cd570df1127d8a67ff04a966b4aa370a81a9b671f201b1f07e13c6
+size 674333
diff --git a/weights/albedo-jp/added_IVF1336_Flat_nprobe_1_albedo-jp_v2.index b/weights/albedo-jp/added_IVF1336_Flat_nprobe_1_albedo-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..9c63e73ac4a794a2a19430bc677d220dc256da5a
--- /dev/null
+++ b/weights/albedo-jp/added_IVF1336_Flat_nprobe_1_albedo-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bc741ad0bcdb73e07bdbbc4ef19263bf3acb1119ba4b63de4ea2d607ec796dc
+size 164703139
diff --git a/weights/albedo-jp/albedo-jp.pth b/weights/albedo-jp/albedo-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..f309f33279dcf209367a51a22976d5dc436dd67b
--- /dev/null
+++ b/weights/albedo-jp/albedo-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04e6fbe4adcbf0655444055b2c4405a980663bb2b685d7e6850d4652878c81c3
+size 55224197
diff --git a/weights/albedo-jp/cover.png b/weights/albedo-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..5ae59b02f71bd5cd6c2b6ff66b222b64d9c12122
--- /dev/null
+++ b/weights/albedo-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:410eeadc05f9493faf1641ab11b3c487904014a6d9ff3790ffc9427f528a8675
+size 3148729
diff --git a/weights/alhaitham-jp/added_IVF519_Flat_nprobe_1.index b/weights/alhaitham-jp/added_IVF519_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..b9e3f0e0a84460830eb1040ce12cec430b213c58
--- /dev/null
+++ b/weights/alhaitham-jp/added_IVF519_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85f6bec29b85a4943cf13fa410f5e1ed33a2aff40d65ac4be130670a01bff95a
+size 21444067
diff --git a/weights/alhaitham-jp/alhaitham-jp.pth b/weights/alhaitham-jp/alhaitham-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..5f92f58c2f636cbb8d8c0e3abbd245c7ce2271fc
--- /dev/null
+++ b/weights/alhaitham-jp/alhaitham-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f69897b83edba6bfcb9a5dabd415b8065d2d5b831bc437351982d3d43e55d07
+size 55027589
diff --git a/weights/alhaitham-jp/cover.png b/weights/alhaitham-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..0b9d733ef0f8ac7dce98ea56480bf72ecc5a36b9
--- /dev/null
+++ b/weights/alhaitham-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98c5815f3675a7ea738e60e636603ac62c73a2a17c778c34e840192dc6dd486c
+size 2132294
diff --git a/weights/alhaitham-jp/total_fea.npy b/weights/alhaitham-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..7765f95b27ed770761a2616e91736472811145fe
--- /dev/null
+++ b/weights/alhaitham-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a5f8dd07f026a4b52bb473c48e2f528d39d4ceb75ca3bee4fbf3f3b6bd5b2f7
+size 20746368
diff --git a/weights/amber-jp/added_IVF837_Flat_nprobe_1_amber-v2_v2.index b/weights/amber-jp/added_IVF837_Flat_nprobe_1_amber-v2_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..4c8b2234180c07978a6ab957cbec29130b326b2a
--- /dev/null
+++ b/weights/amber-jp/added_IVF837_Flat_nprobe_1_amber-v2_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3c71b5a9843dfb300aed5813549e412df5ad6222e39e09fa37cf5057cf2e717
+size 103118539
diff --git a/weights/amber-jp/amber-jp.pth b/weights/amber-jp/amber-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..aaa31f81686809a449da278809939501c554a3dd
--- /dev/null
+++ b/weights/amber-jp/amber-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2404264ca151a4a506d506363c5b552b2a6ac92009d88789a509971f422459c7
+size 55223738
diff --git a/weights/amber-jp/cover.png b/weights/amber-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..1d7caa63fb6ac8da649bafc83d43974d00cc6c1b
--- /dev/null
+++ b/weights/amber-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fe3f1a827f51eb19be8f8d57d5926a5aec19efb2e4c0c50752a5ad18509d68b
+size 177376
diff --git a/weights/ayaka-jp/added_IVF415_Flat_nprobe_6.index b/weights/ayaka-jp/added_IVF415_Flat_nprobe_6.index
new file mode 100644
index 0000000000000000000000000000000000000000..d91aa1e4bedfbde4dba33cbface22865c53593be
--- /dev/null
+++ b/weights/ayaka-jp/added_IVF415_Flat_nprobe_6.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98a4411fb7fbe94d4878df2cc1351f15e3c0c67ac22a7671700653e82dd24373
+size 17155075
diff --git a/weights/ayaka-jp/ayaka-jp.pth b/weights/ayaka-jp/ayaka-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..bca6bc15df5238ac2a748391a01c975952844e7b
--- /dev/null
+++ b/weights/ayaka-jp/ayaka-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:633efdfcf421aeb3ae4df55dd6c17d1570a48b48bff6eaede078492c20d0e0f6
+size 55027130
diff --git a/weights/ayaka-jp/cover.png b/weights/ayaka-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..cb93f95ecbf18f6a786ff041d63521e119c2fd3f
--- /dev/null
+++ b/weights/ayaka-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fe85d2c9895d4d0010660eb42ffa154edb7e0decc97f4444ba6009c69d029c0
+size 837584
diff --git a/weights/ayaka-jp/total_fea.npy b/weights/ayaka-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..800f04c04696d823f635168eb287cc54fa13bd1f
--- /dev/null
+++ b/weights/ayaka-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36cda9528a9782c44637380dce3dad6e05ed27efcb94fc511dcbc66e9ab852c1
+size 16597120
diff --git a/weights/ayato-jp/added_IVF1304_Flat_nprobe_1.index b/weights/ayato-jp/added_IVF1304_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..dcf5f16795bba5f568f9e7dd12e9d7dc9acbc5a4
--- /dev/null
+++ b/weights/ayato-jp/added_IVF1304_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fd600f40f3d8ef0c490eb66d73f25f89056d982edd4f467bda6b699d13e1fca
+size 53868475
diff --git a/weights/ayato-jp/ayato-jp.pth b/weights/ayato-jp/ayato-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..1d15e9ca1dec7f1d50723e97718540604948ba31
--- /dev/null
+++ b/weights/ayato-jp/ayato-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6139fd4482942ab0075f08d5960d8b9b576945e6204ef2965dd3bb96ccccca4
+size 55027130
diff --git a/weights/ayato-jp/cover.png b/weights/ayato-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..218d119f8371168aa8a8edd18c2efe4910ae0bf3
--- /dev/null
+++ b/weights/ayato-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77a250dad75461af6cd593e2ce8a03d9bb12e664e36b7327fcaab6151ccbf93c
+size 2986782
diff --git a/weights/barbara-jp/added_IVF548_Flat_nprobe_1.index b/weights/barbara-jp/added_IVF548_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..a611f325a6bda7b747f44359b9569eebfd7fab99
--- /dev/null
+++ b/weights/barbara-jp/added_IVF548_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2be7b7d0ee5d9d447271f7be807f47ac9f614138e76c1655633f39c0bd2ab588
+size 22632931
diff --git a/weights/barbara-jp/barbara-jp.pth b/weights/barbara-jp/barbara-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..26a1a888a41b6296174c6f7b8852c24a71d6d2f0
--- /dev/null
+++ b/weights/barbara-jp/barbara-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19c63519ce1ecb070d128d5c894bff8c5b6ddd4bf8c216ceb0d07c2278409270
+size 55026095
diff --git a/weights/barbara-jp/cover.png b/weights/barbara-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..7b48c209a315c79818cbd83eecee6c4679cd2b66
--- /dev/null
+++ b/weights/barbara-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39eb3957ffaba5501491bda93f3fd48825460219455d2bca051e355f23c5368c
+size 2206352
diff --git a/weights/barbara-jp/total_fea.npy b/weights/barbara-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..0654a17cb7df8c1971731cfd4bd951c16a40f9ec
--- /dev/null
+++ b/weights/barbara-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78f62234c8be9ee720f6d81c8183800f74902ca6df987c2b195ad3824361bdef
+size 21896320
diff --git a/weights/bennett-jp/added_IVF900_Flat_nprobe_1_benett-v2_v2.index b/weights/bennett-jp/added_IVF900_Flat_nprobe_1_benett-v2_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..f50902bf16e510a9e72cbd954ba745d4b9b08856
--- /dev/null
+++ b/weights/bennett-jp/added_IVF900_Flat_nprobe_1_benett-v2_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26bddef4a5b5d77f022e902e0488da4f5d3841b6836d6cc99583598b347ff64a
+size 110926339
diff --git a/weights/bennett-jp/bennett-jp.pth b/weights/bennett-jp/bennett-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..053834f12896675059db399059172f389dad7d57
--- /dev/null
+++ b/weights/bennett-jp/bennett-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0ba3a39928644aa3b516a643b5eeda486bd5b32a3849828b9d24c4027631d6b
+size 55224197
diff --git a/weights/bennett-jp/cover.png b/weights/bennett-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..e9a6ec3cb39e6eb9908d718bb0bf190cd892d604
--- /dev/null
+++ b/weights/bennett-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59be3682cc99262ab54adcd13f66c5ccfdf0597d4150809e723b4bdd72368835
+size 310302
diff --git a/weights/candace-jp/added_IVF920_Flat_nprobe_1_candace_v2.index b/weights/candace-jp/added_IVF920_Flat_nprobe_1_candace_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..5c7e1f7d2be299b46b6e8d71da04d2461dbe108f
--- /dev/null
+++ b/weights/candace-jp/added_IVF920_Flat_nprobe_1_candace_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06bef32a5fcf926a92b05a35984dc40b29c4c367f91f38134a620f63293b266d
+size 113402659
diff --git a/weights/candace-jp/candace-jp.pth b/weights/candace-jp/candace-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..213874970782466ced702fbffb40030c22e6650a
--- /dev/null
+++ b/weights/candace-jp/candace-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa51dee508f1cd9ec83b53fd9aeded03edfa645733eaa7a981ab4dd51ae27f63
+size 55216420
diff --git a/weights/candace-jp/cover.png b/weights/candace-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..2fbf97f879974dc733025a978038d66bb7444b56
--- /dev/null
+++ b/weights/candace-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fee3786269ee818349474b08f13e7823b88ddc517011c99ea4a7aced2faa38f1
+size 271987
diff --git a/weights/charlotte-jp/added_IVF1318_Flat_nprobe_1_charlotte-jp_v2_400.index b/weights/charlotte-jp/added_IVF1318_Flat_nprobe_1_charlotte-jp_v2_400.index
new file mode 100644
index 0000000000000000000000000000000000000000..7c747c2f56cd494b3e4f49b26b0180d66cbef30f
--- /dev/null
+++ b/weights/charlotte-jp/added_IVF1318_Flat_nprobe_1_charlotte-jp_v2_400.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d18fb9a5a38bba0440a740bc3b9c997dffc0d4d0d97a24cd76293ef9d48ee855
+size 162414699
diff --git a/weights/charlotte-jp/charlotte-jp.pth b/weights/charlotte-jp/charlotte-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..e06a7a6e821b9a6499f68e3a8cfb4f53c0eba907
--- /dev/null
+++ b/weights/charlotte-jp/charlotte-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53fa283c1066d6cb0672ac753868941c3a0f1095412f0b65daf84c7bb7c90087
+size 57590442
diff --git a/weights/charlotte-jp/cover.png b/weights/charlotte-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..521517ba5a490cb21a89824e7a17d052680507ef
--- /dev/null
+++ b/weights/charlotte-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e971d7e7e3808448b6e2cd758d3b3e9eed35f73b1b5b24bf99939c9ddeab6b2
+size 380512
diff --git a/weights/childe-jp/added_IVF428_Flat_nprobe_6.index b/weights/childe-jp/added_IVF428_Flat_nprobe_6.index
new file mode 100644
index 0000000000000000000000000000000000000000..d9c6476f50cae7bce3da01999ad8fc40f58fcaac
--- /dev/null
+++ b/weights/childe-jp/added_IVF428_Flat_nprobe_6.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4801f410ada2c2447c783db9160b414f73ebd12d7fa456a351bae3c7727120d8
+size 17688619
diff --git a/weights/childe-jp/childe-jp.pth b/weights/childe-jp/childe-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..3c273d8af3e4f2ffd96e674e713c9967bd02d3aa
--- /dev/null
+++ b/weights/childe-jp/childe-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c50bbeb4094fb622f74fc3481e0807e049092709ae4b877d266ab0be45303f9
+size 55027589
diff --git a/weights/childe-jp/cover.png b/weights/childe-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..2d8a667a5f6a3b7b24a310be32378a2ff7ea1bc7
--- /dev/null
+++ b/weights/childe-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89f4a95e7a1fc11a5fcc066f97129871cfe5067ac9a9061181e73fbb475f188d
+size 3703090
diff --git a/weights/childe-jp/total_fea.npy b/weights/childe-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..3c5ce6ae905c6ae610a7eb48fd1acdcba733e436
--- /dev/null
+++ b/weights/childe-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ea1a5a8674a3122afa6f23c728c555eebfe6501700033754e47c25404167ee7
+size 17113216
diff --git a/weights/chongyun-jp/added_IVF545_Flat_nprobe_1.index b/weights/chongyun-jp/added_IVF545_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..9a3f02c33be0ffc6edf784f13f89553fa0fe1698
--- /dev/null
+++ b/weights/chongyun-jp/added_IVF545_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18f276e45825f506042ac0345a6e0450f57d2b83bd76bec502caf282d7a525c2
+size 22525603
diff --git a/weights/chongyun-jp/chongyun-jp.pth b/weights/chongyun-jp/chongyun-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..52b06ea8089d2c21eeedd36a653824f57a1f510a
--- /dev/null
+++ b/weights/chongyun-jp/chongyun-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56885119bd878ded72a7d37226e1ec2380f2b60fea09b89c4113463b10626002
+size 55028507
diff --git a/weights/chongyun-jp/cover.png b/weights/chongyun-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..6ef48a14b03b15ef898092755b15fe36998d37e2
--- /dev/null
+++ b/weights/chongyun-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6836655797760f30ca627de4792c5b1bac15036e7123fa3f5b4892b00d77ae48
+size 7293987
diff --git a/weights/chongyun-jp/total_fea.npy b/weights/chongyun-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..191c8d44611eb64d3745f58df69fe84468daee9e
--- /dev/null
+++ b/weights/chongyun-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:602ed8eee37f336d283cfd579833897162fcdbb3cecebacebd16323cd322320b
+size 21792896
diff --git a/weights/collei-jp/added_IVF1127_Flat_nprobe_1_collei_v2.index b/weights/collei-jp/added_IVF1127_Flat_nprobe_1_collei_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..4ffb4778a13e65a6562b4644ebcf6234d6f7bbea
--- /dev/null
+++ b/weights/collei-jp/added_IVF1127_Flat_nprobe_1_collei_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fed2da73c8abf97c72c4f45700baa31168b2cd7b786bd1a794fdb9ef85edb15
+size 138874259
diff --git a/weights/collei-jp/collei-jp.pth b/weights/collei-jp/collei-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..beb0ed5b14d63e852a946323032f8e85482e245b
--- /dev/null
+++ b/weights/collei-jp/collei-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:253515c4fbc4abd333b2e8e9f351dc0c6cd9b828905840e6f925b9892c50c305
+size 55216420
diff --git a/weights/collei-jp/cover.png b/weights/collei-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..83add9d6cb84be8862f0c6644f22016d3dade3c1
--- /dev/null
+++ b/weights/collei-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f74c0d5091d9e9d27971f2f38b0f04f2973166816dc61bb2fbf427e90277836
+size 228176
diff --git a/weights/cyno-jp/added_IVF380_Flat_nprobe_1_cyno-jp_v1.index b/weights/cyno-jp/added_IVF380_Flat_nprobe_1_cyno-jp_v1.index
new file mode 100644
index 0000000000000000000000000000000000000000..cb022361380ce3a8bfcd2573e5e47a6990447874
--- /dev/null
+++ b/weights/cyno-jp/added_IVF380_Flat_nprobe_1_cyno-jp_v1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fa013e7086ac42099fd213e5a89281dc4272a841a946039125d07be4308c6e5
+size 15699955
diff --git a/weights/cyno-jp/cover.png b/weights/cyno-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..867d1429c5baa15a2d3be927f060c40dc54671c0
--- /dev/null
+++ b/weights/cyno-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26fe49b06f770a3b6ff317ecc0892113b8a9812ee6390283dad73505589bebe1
+size 865801
diff --git a/weights/cyno-jp/cyno-jp.pth b/weights/cyno-jp/cyno-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..636e33dbd46df07c4263bf8fef88314c6cc0e32f
--- /dev/null
+++ b/weights/cyno-jp/cyno-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bfe9d88e0d89c6d26b3d406afa97915e12cf49c5c72af329343c2b935b07932
+size 55026095
diff --git a/weights/cyno-jp/total_fea.npy b/weights/cyno-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..d9606dd1d194b6c10c1b77427b0cbd224529f178
--- /dev/null
+++ b/weights/cyno-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6490bacafecf129a3d020c72aeb3dcfe875e5cf04272314eb4793db59ec2c077
+size 15189120
diff --git a/weights/diluc-jp/added_IVF1511_Flat_nprobe_1.index b/weights/diluc-jp/added_IVF1511_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..6ef4c99954f74d0b86064d66071c4af0ba02d9b5
--- /dev/null
+++ b/weights/diluc-jp/added_IVF1511_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e2a2500a55fb9c888f75af94f7fef54e2efbb453e8294f9bcda4f84ae4ca5dd
+size 62405179
diff --git a/weights/diluc-jp/cover.png b/weights/diluc-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..91f909d5b187d796edb9c4053af84af6c3fc5cd9
--- /dev/null
+++ b/weights/diluc-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bff8978711329395eedc7b4d3b8aeaea043d4dbea6c7c2f4668132b8abfe5adf
+size 1229695
diff --git a/weights/diluc-jp/diluc-jp.pth b/weights/diluc-jp/diluc-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..a4c9144d3a667d8ccb1354b5122e47acccd93d5b
--- /dev/null
+++ b/weights/diluc-jp/diluc-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8d2324a1c31d9e5516c807e0ab5c560fb5bf080be74f8a3849809b126d4b1dc
+size 55027130
diff --git a/weights/eula-jp/added_IVF2219_Flat_nprobe_1.index b/weights/eula-jp/added_IVF2219_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..c3485f75dc13800484f6fb28fd777dcc35f64827
--- /dev/null
+++ b/weights/eula-jp/added_IVF2219_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b315239e25bc0e71239d5b825618a2d7c16af9a87d36a1f84a2a5c31e48be861
+size 91614907
diff --git a/weights/eula-jp/cover.png b/weights/eula-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..9845fd5be1494c0fec008afdf50438075fe863f4
--- /dev/null
+++ b/weights/eula-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e865a7dfd1b35c6851b8707a7a7cfcdff10802c6ee645179e467b02fda83fb9
+size 2659755
diff --git a/weights/eula-jp/eula-jp.pth b/weights/eula-jp/eula-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..6ff9bd077625f0399675e3b0a4322bc45df64073
--- /dev/null
+++ b/weights/eula-jp/eula-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b806e046ee5485ea42522b7f1df79e64e7bec7a248cf8dda21a85a86cfd88054
+size 55026095
diff --git a/weights/faruzan-jp/added_IVF256_Flat_nprobe_1_faruzan-jp_v2.index b/weights/faruzan-jp/added_IVF256_Flat_nprobe_1_faruzan-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..8993dea9636dfcace30ee4b15260a8cd98b1b444
--- /dev/null
+++ b/weights/faruzan-jp/added_IVF256_Flat_nprobe_1_faruzan-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de4875ecdb7a22a8f54ea1fe897edf8b147eb0dfc9372f104b6706e0ab338c14
+size 31588619
diff --git a/weights/faruzan-jp/cover.png b/weights/faruzan-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..e57fd8a6e8498688b5a63a64679fff2921f1ff80
--- /dev/null
+++ b/weights/faruzan-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b654797aa533f9580a1f24bec4483bfbcc57a37289368001b98ed2b57a4d570c
+size 392510
diff --git a/weights/faruzan-jp/faruzan-jp.pth b/weights/faruzan-jp/faruzan-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..f8b0218da332e3d215bf65672fae49c3ac306a4c
--- /dev/null
+++ b/weights/faruzan-jp/faruzan-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57164f27f40b4566d5d2e801dc6b89fce48c76354ca8af15e9c600901442f795
+size 57589524
diff --git a/weights/fischl-jp/added_IVF1225_Flat_nprobe_1.index b/weights/fischl-jp/added_IVF1225_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..12914076f7d483bdae076b4cb28916d52bbae742
--- /dev/null
+++ b/weights/fischl-jp/added_IVF1225_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cb77f4db13ba45f4d48f711ad3c9e13159b9e8ff7b5e8ff0ac705ad0ed3bad4
+size 50599099
diff --git a/weights/fischl-jp/cover.png b/weights/fischl-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..fb68b848519d22658b5c7623d9052360286a1e72
--- /dev/null
+++ b/weights/fischl-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52ebc66caf1fb853652ecaee99be63c1383e7c8869e4c4d006c52805383db7d4
+size 214329
diff --git a/weights/fischl-jp/fischl-jp.pth b/weights/fischl-jp/fischl-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..0a07b61434dc0bf92ccfd5b51c2dad7bffb053b7
--- /dev/null
+++ b/weights/fischl-jp/fischl-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28c04e4cd0ddb1abc8a3d0ad0b7aea94fba37b9ef5903e2f4e8c2e62015844f5
+size 55027589
diff --git a/weights/ganyu-jp/added_IVF1636_Flat_nprobe_1.index b/weights/ganyu-jp/added_IVF1636_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..8a1e5da0adc6ef4a8a295a5f505e3a9db51af309
--- /dev/null
+++ b/weights/ganyu-jp/added_IVF1636_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bee4d6ad7b7f8f0739b3f3f27639b8da7f9ad6fa216e10e98cf933505fe745c5
+size 67551763
diff --git a/weights/ganyu-jp/cover.png b/weights/ganyu-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..7091081ea93806674376f643a44086f2ad567629
--- /dev/null
+++ b/weights/ganyu-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03e153e8941af764c7381d9ba6580b24b67a10a5cb2342a9f7f39895b159939b
+size 190367
diff --git a/weights/ganyu-jp/ganyu-jp.pth b/weights/ganyu-jp/ganyu-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..d7a0de68257a6fc8f8af5c48ed9638d530cb78e0
--- /dev/null
+++ b/weights/ganyu-jp/ganyu-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70e71be2ae8ce8929a3a1e95dfa7f6a5c15d7c8a0854ea2bdaf36cf5aaf266b5
+size 55027130
diff --git a/weights/heizou-jp/added_IVF466_Flat_nprobe_1_heizou-jp_v1.index b/weights/heizou-jp/added_IVF466_Flat_nprobe_1_heizou-jp_v1.index
new file mode 100644
index 0000000000000000000000000000000000000000..94f8c679c8ecc6782f999c1a27e83c6afd583aed
--- /dev/null
+++ b/weights/heizou-jp/added_IVF466_Flat_nprobe_1_heizou-jp_v1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:646274e8ca826094651422679b72f345a9c74e7a8da7e3cd1f92bb4b4b5e4812
+size 19270675
diff --git a/weights/heizou-jp/cover.png b/weights/heizou-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..1286c38020b758f2f17e84e1b9270291ee377e19
--- /dev/null
+++ b/weights/heizou-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1705f8de0802a3b4010f30376fbaf2aad566400a71cb194730de9cd150be99aa
+size 655834
diff --git a/weights/heizou-jp/heizou-jp.pth b/weights/heizou-jp/heizou-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..85f108f995df0e8e20ea08d36b5cbcec1dd772ac
--- /dev/null
+++ b/weights/heizou-jp/heizou-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:069502645d5790d7598f971450fbfffbcc9927bd55b262b7c54010d5507fd622
+size 55027589
diff --git a/weights/heizou-jp/total_fea.npy b/weights/heizou-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..57b2385801eb5b9c120e911d23007b9d8c55445a
--- /dev/null
+++ b/weights/heizou-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:196d069595f6c9471eb5cc5b97916597f3df7450ec4ecb6a58d1b2e1cbc875c0
+size 18644096
diff --git a/weights/hutao-jp/added_IVF265_Flat_nprobe_5.index b/weights/hutao-jp/added_IVF265_Flat_nprobe_5.index
new file mode 100644
index 0000000000000000000000000000000000000000..b11623d9924d8f802bb4dcb6aa4e12bc4d557720
--- /dev/null
+++ b/weights/hutao-jp/added_IVF265_Flat_nprobe_5.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1250005b613d5ffb368da9f93fbde733074757cdba9fb1fd04aa1f286b883e0a
+size 10962043
diff --git a/weights/hutao-jp/cover.png b/weights/hutao-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..2679b51ee930530f9b541a16c884d5985814eb8c
--- /dev/null
+++ b/weights/hutao-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d7f42be2e382feec5ed0345196473715ef5be0c9ac68628dc77e75fae0e7316
+size 6616541
diff --git a/weights/hutao-jp/hutao-jp.pth b/weights/hutao-jp/hutao-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..75c2dd4bc0448275dc059299dca3d9c31a5a679f
--- /dev/null
+++ b/weights/hutao-jp/hutao-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52ae0bd715c555b8e0072bc98bf4b462aab42b93a48307652e7a9164a0539990
+size 55027130
diff --git a/weights/hutao-jp/total_fea.npy b/weights/hutao-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..bf396316f4cb1eabab289f6749d7959ac5856c49
--- /dev/null
+++ b/weights/hutao-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e6c6687b5d0881113a54ca53b0486292538853c6a6488c702bf94d991b7aec5
+size 10605696
diff --git a/weights/itto-jp/added_IVF4454_Flat_nprobe_1_itto-jp_v2.index b/weights/itto-jp/added_IVF4454_Flat_nprobe_1_itto-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..918ce7b8510c9a8710440b9c3ce57e179f40932d
--- /dev/null
+++ b/weights/itto-jp/added_IVF4454_Flat_nprobe_1_itto-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:449be981798328db47a934432bafb22475f5d7cff4e4ce0688c15731cbe7cd49
+size 548843819
diff --git a/weights/itto-jp/cover.png b/weights/itto-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..ddd4bc8b74aa53b35805331d8a94647099150fe7
--- /dev/null
+++ b/weights/itto-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca9c27a5d1bf6d5a91b5c077d6ec533d727cb8650288c962e92d8d3cbeacebb9
+size 1964620
diff --git a/weights/itto-jp/itto-jp.pth b/weights/itto-jp/itto-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..56530f8d9e835a259bc57c6200cb6b8d873b5c9e
--- /dev/null
+++ b/weights/itto-jp/itto-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9307e39f5d55608f0f76107a9cff19bea8f9e456c9932b735aaf87234b632e1e
+size 55228787
diff --git a/weights/kaeya-jp/added_IVF1655_Flat_nprobe_1.index b/weights/kaeya-jp/added_IVF1655_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..bb6bd1cd3fe27464f5bd63edb5c27817d9602922
--- /dev/null
+++ b/weights/kaeya-jp/added_IVF1655_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1afba8e77f937bf76bd10f54fcbb4ca119dc3b2c599dc4cbf0843d618c26ad15
+size 68325763
diff --git a/weights/kaeya-jp/cover.png b/weights/kaeya-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..db149827664a46b4552b93dac4df82d8d0c2c118
--- /dev/null
+++ b/weights/kaeya-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4b12f3c2ce55d28d8cadcd17464d0922e55aac37495654a7bcfbb03c8beffd5
+size 623087
diff --git a/weights/kaeya-jp/kaeya-jp.pth b/weights/kaeya-jp/kaeya-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..ede481f507e03d309edfa9afefbf538c01a72f59
--- /dev/null
+++ b/weights/kaeya-jp/kaeya-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6492cc522ecaef80436bae3ebdb6e6bd9e6242dc6be74c4e6283c7825285d480
+size 55027130
diff --git a/weights/kaveh-jp/added_IVF4967_Flat_nprobe_1_kaveh-jp_v2.index b/weights/kaveh-jp/added_IVF4967_Flat_nprobe_1_kaveh-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..29941bb80bb666823516c68f6db7483c80100739
--- /dev/null
+++ b/weights/kaveh-jp/added_IVF4967_Flat_nprobe_1_kaveh-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:964f5e602826c7d276468918bf9848d0adee099753e5f4fd12db49f176a61792
+size 611943779
diff --git a/weights/kaveh-jp/added_IVF613_Flat_nprobe_1_kaveh_v2_v2.index b/weights/kaveh-jp/added_IVF613_Flat_nprobe_1_kaveh_v2_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..b8c31afd3979e86fba9650c61ab181f8b3510007
--- /dev/null
+++ b/weights/kaveh-jp/added_IVF613_Flat_nprobe_1_kaveh_v2_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ae118c5f5e5cdd1b0e37e7b46c95428e7d5305cb5f2708dbb8954a53ddafc07
+size 75540219
diff --git a/weights/kaveh-jp/cover.png b/weights/kaveh-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..fb653c6de451001a776bde68b184bc74634e9673
--- /dev/null
+++ b/weights/kaveh-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b95bb4f18577150d55bf366b2c93e2475f5110c3ab460f09692e8636144649b9
+size 292685
diff --git a/weights/kaveh-jp/kaveh-jp.pth b/weights/kaveh-jp/kaveh-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..06de63235666319a4eb27cfc016019e6ad25052a
--- /dev/null
+++ b/weights/kaveh-jp/kaveh-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0731c70f58b771ce4edd2dc145e026ef3103f1e313bb3cdfac4fc67668fc1e7f
+size 57583034
diff --git a/weights/kazuha-jp/added_IVF677_Flat_nprobe_7.index b/weights/kazuha-jp/added_IVF677_Flat_nprobe_7.index
new file mode 100644
index 0000000000000000000000000000000000000000..932bd064bf70acaadc761da45a986ae7c1426552
--- /dev/null
+++ b/weights/kazuha-jp/added_IVF677_Flat_nprobe_7.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47d1f762e5d990d317ee7fb7d18190a37d6a7169ededd726b912b013c67fa20f
+size 27970435
diff --git a/weights/kazuha-jp/cover.png b/weights/kazuha-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..afe44007a48d63de51cbffc9b5e62fd073c14a2d
--- /dev/null
+++ b/weights/kazuha-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db361de17f6d33684be140dfedd74794dec79b85df97ba7595b3755680139f57
+size 5805839
diff --git a/weights/kazuha-jp/kazuha-jp.pth b/weights/kazuha-jp/kazuha-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..7507702fe3bf2b6c8ec207add11c65b5e19b6c22
--- /dev/null
+++ b/weights/kazuha-jp/kazuha-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d69c9c77250cff7c47a74de33ac11089ccc41122f7a5461b77f14f28738dd7e7
+size 55027589
diff --git a/weights/kazuha-jp/total_fea.npy b/weights/kazuha-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..968b678ead5fe3f2163a08541fe825eb3e2d8e60
--- /dev/null
+++ b/weights/kazuha-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f8bd3cc84c5e8230f4631dac4839b922f0b4e726b325813bc75773e19a2f19f
+size 27060352
diff --git a/weights/keqing-jp/added_IVF1634_Flat_nprobe_1.index b/weights/keqing-jp/added_IVF1634_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..0ea872cc5dafa0a10aa7a80127ff81ccc944d680
--- /dev/null
+++ b/weights/keqing-jp/added_IVF1634_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9937fc19ea1cfd7ce88478ebb746f7e41522ca9562bc33b81478a61843ae43ae
+size 67465075
diff --git a/weights/keqing-jp/cover.png b/weights/keqing-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..221015d8650b9967ba3ee3b51d7d811359e59a98
--- /dev/null
+++ b/weights/keqing-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af5af757312915725a637d3e26af82be378ed2af18cfc2ed498207d7c948b0c3
+size 2106388
diff --git a/weights/keqing-jp/keqing-jp.pth b/weights/keqing-jp/keqing-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..5a58ae9d1f0f740cdc3e556959d64ca33f9272ef
--- /dev/null
+++ b/weights/keqing-jp/keqing-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b0f2587bd933a2933364224b134fd6bd63cc9434c215f7f94adae46daa0ca90
+size 55027589
diff --git a/weights/kirara-jp/added_IVF672_Flat_nprobe_1.index b/weights/kirara-jp/added_IVF672_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..951570e9d8fb09751f0b2d2feffc766d7a788068
--- /dev/null
+++ b/weights/kirara-jp/added_IVF672_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d81665d33cf0b15b259a9142280a82869170b50fc99a70d90a4cb535b14d6115
+size 27741331
diff --git a/weights/kirara-jp/cover.png b/weights/kirara-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..257f6f330b843af640f411dcbdc9ce7b1f4af2fd
--- /dev/null
+++ b/weights/kirara-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:742a99d20fa0d2c3cae65907bae912c8d9edcf8015a3c1a084456127e8eb55a7
+size 330959
diff --git a/weights/kirara-jp/kirara-jp.pth b/weights/kirara-jp/kirara-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..6f48325ec973bc35364ec4be614ad6ddccc76855
--- /dev/null
+++ b/weights/kirara-jp/kirara-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c444de82a81661354f96fd203f0f6535157abcb46e93bbb6bc5e19098fbdba9a
+size 55019812
diff --git a/weights/kirara-jp/total_fea.npy b/weights/kirara-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..aaaabc66e369a850703accd52cffebd392e0cdf3
--- /dev/null
+++ b/weights/kirara-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7c09bf1ef996bc4ec9ba26138a2baf9b4355d38a0c77b78d3a85ad445caab1f
+size 26838144
diff --git a/weights/klee-jp/added_IVF282_Flat_nprobe_5.index b/weights/klee-jp/added_IVF282_Flat_nprobe_5.index
new file mode 100644
index 0000000000000000000000000000000000000000..f3de57e253545ac0fa97bde6e448dc24f71a8382
--- /dev/null
+++ b/weights/klee-jp/added_IVF282_Flat_nprobe_5.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bfb669f7b701d8bebb49660eaa50b6cfb1bae59d43e82ad054f1ffed73fcac2
+size 11671027
diff --git a/weights/klee-jp/cover.png b/weights/klee-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..0da6b6af0eb1d9e11a917ee007215dfdaa00222b
--- /dev/null
+++ b/weights/klee-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c30ca76b76807992b11f27ee49d0f1aa80ac81da6c9b1b60d2d2a3e75f609752
+size 3242773
diff --git a/weights/klee-jp/klee-jp.pth b/weights/klee-jp/klee-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..a3377722569ccc0f6364052f799d9240d4ae0e90
--- /dev/null
+++ b/weights/klee-jp/klee-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:054e796c9c1214d400847e51cfc25f0f39bfd50437d1e8f001051d426303bbea
+size 55026095
diff --git a/weights/klee-jp/total_fea.npy b/weights/klee-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..f330c0cb715f633160b567ad03927694a283329e
--- /dev/null
+++ b/weights/klee-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:070407ecc7c8b125c8f68723d579c5e002b37e21ec85cda096000f5c18b56bb0
+size 11291776
diff --git a/weights/kokomi-jp/added_IVF1723_Flat_nprobe_1.index b/weights/kokomi-jp/added_IVF1723_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..d67370301e7173cc830d0ba0977208928a358dbf
--- /dev/null
+++ b/weights/kokomi-jp/added_IVF1723_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5762d6c5e9d147844b1c6eda0ecd16595b7a33d77607e9d2b729e1dd0a0fc87f
+size 71141059
diff --git a/weights/kokomi-jp/cover.png b/weights/kokomi-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..841330eca26152b1d21fd1c0ff561463f54d003d
--- /dev/null
+++ b/weights/kokomi-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebb00e1c70bdacfb5660333ce1f8d19e81e0b3c4dde9bb44bb2eeac0e6cfe3aa
+size 292883
diff --git a/weights/kokomi-jp/kokomi-jp.pth b/weights/kokomi-jp/kokomi-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..5fb6249156c2fc3c70da8cc33726c71534edd2bb
--- /dev/null
+++ b/weights/kokomi-jp/kokomi-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bcd3d8988f86503cbec98170cd97da4a0187f9f603385a563ca0dc3d7758ba4
+size 55027589
diff --git a/weights/layla-jp/added_IVF1099_Flat_nprobe_1_layla-v2_v2.index b/weights/layla-jp/added_IVF1099_Flat_nprobe_1_layla-v2_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..530db8295de2a4f63b62d06531def25c0a542079
--- /dev/null
+++ b/weights/layla-jp/added_IVF1099_Flat_nprobe_1_layla-v2_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83681f693197d19cff135daa155ea4c4d0d302f461473573dbdf55072e108209
+size 135501659
diff --git a/weights/layla-jp/cover.png b/weights/layla-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..f31e744b552ab938628ded8d63dce667b88cda40
--- /dev/null
+++ b/weights/layla-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bbf868f3e8e6998416e8ed0f844d156c2c6566366636c03a1549fca8bab5cd6
+size 311684
diff --git a/weights/layla-jp/layla-jp.pth b/weights/layla-jp/layla-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..fef4b13476b9e5ebf931b43df2fbb8893fe4de1b
--- /dev/null
+++ b/weights/layla-jp/layla-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3921520217fdbd60dbeaa0dfacde610171a7417d27d656afafc33086d9d4b95
+size 55223738
diff --git a/weights/lisa-jp/added_IVF758_Flat_nprobe_1_lisa_v2.index b/weights/lisa-jp/added_IVF758_Flat_nprobe_1_lisa_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..3c6115f8e1d56b585176d255d1eb0b8eb7e204e0
--- /dev/null
+++ b/weights/lisa-jp/added_IVF758_Flat_nprobe_1_lisa_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1104ddc0bf73c98a8d69496535e0055385266315b54aacd016ea8d970e08b742
+size 93468899
diff --git a/weights/lisa-jp/cover.png b/weights/lisa-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..033942fb314b6c228e90d51f397c3320be5d2a92
--- /dev/null
+++ b/weights/lisa-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07cbc5378f35950da17667ec6d23fab426b30e5bd8eed70811534045b781dd9
+size 291648
diff --git a/weights/lisa-jp/lisa-jp.pth b/weights/lisa-jp/lisa-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..18f20d057572fa660992242e29ec440d91284f28
--- /dev/null
+++ b/weights/lisa-jp/lisa-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f70b41090008cffa4d8d4f6eebcb9319f79587aeb15b84a0e4e8c72a7c8336dc
+size 55192782
diff --git a/weights/lumine-jp/added_IVF1165_Flat_nprobe_1_lumine-v2-new_v2.index b/weights/lumine-jp/added_IVF1165_Flat_nprobe_1_lumine-v2-new_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..a4c39ed4161518e5cf6c8592f9612f981efd49c7
--- /dev/null
+++ b/weights/lumine-jp/added_IVF1165_Flat_nprobe_1_lumine-v2-new_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:671ed09251150cbdf54bd46983a22d399c32256e6a83b50fa3e1e207c316d986
+size 143555859
diff --git a/weights/lumine-jp/cover.png b/weights/lumine-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..4ae170ff5e91403574d09c708f04b555fb4d4508
--- /dev/null
+++ b/weights/lumine-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fb57ac1dbc9f3022581dc4776118239a6751734129c05ca30746e0bf5d9fb30
+size 285892
diff --git a/weights/lumine-jp/lumine-jp.pth b/weights/lumine-jp/lumine-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..a744dbc4ffbf5eabac7e52cbee5310dff470e22b
--- /dev/null
+++ b/weights/lumine-jp/lumine-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89811f0885b7dee838597a9ad1bfff306f1a0918a7902001cff4a6c6a0bd0b47
+size 55226033
diff --git a/weights/model_info.json b/weights/model_info.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a6320aa5d96f443c830c1b045ed86c42fa675c3
--- /dev/null
+++ b/weights/model_info.json
@@ -0,0 +1,437 @@
+{
+	"aether-jp": {
+		"enable": true,
+		"model_path": "aether-jp.pth",
+        "title": "Genshin Impact - Aether",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF865_Flat_nprobe_1_aether-jp_v2.index",
+		"author":"ArkanDash"
+    },
+	"albedo-jp": {
+		"enable": true,
+		"model_path": "albedo-jp.pth",
+        "title": "Genshin Impact - Albedo",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1336_Flat_nprobe_1_albedo-jp_v2.index",
+		"author":"KarmKarma"
+    },
+  	"alhaitham-jp": {
+		"enable": true,
+		"model_path": "alhaitham-jp.pth",
+		"title": "Genshin Impact - Alhaitham",
+		"cover": "cover.png",
+		"feature_retrieval_library": "added_IVF519_Flat_nprobe_1.index",
+		"author":"NMEX"
+    },
+	"amber-jp": {
+		"enable": true,
+		"model_path": "amber-jp.pth",
+        "title": "Genshin Impact - Amber",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF837_Flat_nprobe_1_amber-v2_v2.index",
+		"author": "HirumiM"
+    },
+	"ayaka-jp": {
+		"enable": true,
+		"model_path": "ayaka-jp.pth",
+        "title": "Genshin Impact - Kamisato Ayaka",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF415_Flat_nprobe_6.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"ayato-jp": {
+		"enable": true,
+		"model_path": "ayato-jp.pth",
+		"title": "Genshin Impact - Kamisato Ayato",
+		"cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1304_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+    "barbara-jp": {
+		"enable": true,
+		"model_path": "barbara-jp.pth",
+        "title": "Genshin Impact - Barbara",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF548_Flat_nprobe_1.index",
+		"feature_file": "total_fea.npy",
+		"author":"firzaelbuho"
+    },
+    "bennett-jp": {
+		"enable": true,
+		"model_path": "bennett-jp.pth",
+        "title": "Genshin Impact - Bennett",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF900_Flat_nprobe_1_benett-v2_v2.index",
+		"author": "HirumiM"
+    },
+    "candace-jp": {
+		"enable": true,
+		"model_path": "candace-jp.pth",
+        "title": "Genshin Impact - Candace",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF920_Flat_nprobe_1_candez_v2.index",
+		"author": "Mocci24"
+    },
+	"charlotte-jp": {
+		"enable": true,
+		"model_path": "charlotte-jp.pth",
+        "title": "Genshin Impact - Charlotte",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1318_Flat_nprobe_1_charlotte-jp_v2_400.index",
+		"author":"ArkanDash"
+    },
+	"childe-jp": {
+		"enable": true,
+		"model_path": "childe-jp.pth",
+        "title": "Genshin Impact - Ajax, aka Tartaglia or 'Childe'",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF428_Flat_nprobe_6.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+  	"chongyun-jp": {
+		"enable": true,
+		"model_path": "chongyun-jp.pth",
+        "title": "Genshin Impact - Chongyun",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF545_Flat_nprobe_1.index",
+		"feature_file": "total_fea.npy",
+		"author":"Eugenia"
+    },
+    "collei-jp": {
+		"enable": true,
+		"model_path": "collei-jp.pth",
+        "title": "Genshin Impact - Collei",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1127_Flat_nprobe_1_collei_v2.index",
+		"author": "HirumiM"
+    },
+	"cyno-jp": {
+		"enable": true,
+		"model_path": "cyno-jp.pth",
+        "title": "Genshin Impact - Cyno",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF380_Flat_nprobe_1_cyno-jp_v1.index",
+		"feature_file": "total_fea.npy",
+		"author":"KarmKarma"
+    },
+	"diluc-jp": {
+		"enable": true,
+		"model_path": "diluc-jp.pth",
+		"title": "Genshin Impact - Diluc",
+		"cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1511_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+	"eula-jp": {
+		"enable": true,
+		"model_path": "eula-jp.pth",
+        "title": "Genshin Impact - Eula",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF2219_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+	},
+	"faruzan-jp": {
+		"enable": true,
+		"model_path": "faruzan-jp.pth",
+        "title": "Genshin Impact - Faruzan",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF256_Flat_nprobe_1_faruzan-jp_v2.index",
+		"author":"ArkanDash"
+    },
+  	"fischl-jp": {
+		"enable": true,
+		"model_path": "fischl-jp.pth",
+        "title": "Genshin Impact - Fischl",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1225_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+  	"ganyu-jp": {
+		"enable": true,
+		"model_path": "ganyu-jp.pth",
+        "title": "Genshin Impact - Ganyu",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1636_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+  	"heizou-jp": {
+		"enable": true,
+		"model_path": "heizou-jp.pth",
+        "title": "Genshin Impact - Heizou",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF466_Flat_nprobe_1_heizou-jp_v1.index",
+		"feature_file": "total_fea.npy",
+		"author":"KarmKarma"
+    },
+	"hutao-jp": {
+		"enable": true,
+		"model_path": "hutao-jp.pth",
+        "title": "Genshin Impact - Hu Tao",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF265_Flat_nprobe_5.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+    "itto-jp": {
+		"enable": true,
+		"model_path": "itto-jp.pth",
+        "title": "Genshin Impact - Itto",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF4454_Flat_nprobe_1_itto-jp_v2.index",
+		"author":"ArkanDash"
+    },
+    "kaeya-jp": {
+		"enable": true,
+		"model_path": "kaeya-jp.pth",
+		"title": "Genshin Impact - Kaeya",
+		"cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1655_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+	"kaveh-jp": {
+		"enable": true,
+		"model_path": "kaveh-jp.pth",
+        "title": "Genshin Impact - Kaveh",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF4967_Flat_nprobe_1_kaveh-jp_v2.index",
+		"author":"ArkanDash"
+    },
+    "kazuha-jp": {
+		"enable": true,
+		"model_path": "kazuha-jp.pth",
+        "title": "Genshin Impact - Kaedehara Kazuha",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF677_Flat_nprobe_7.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"keqing-jp": {
+		"enable": true,
+		"model_path": "keqing-jp.pth",
+        "title": "Genshin Impact - Keqing",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1634_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+	},
+    "kirara-jp": {
+		"enable": true,
+		"model_path": "kirara-jp.pth",
+        "title": "Genshin Impact - Kirara",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF672_Flat_nprobe_1.index",
+		"feature_file": "total_fea.npy",
+		"author":"Mocci24"
+    },
+	"klee-jp": {
+		"enable": true,
+		"model_path": "klee-jp.pth",
+        "title": "Genshin Impact - Klee",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF282_Flat_nprobe_5.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"kokomi-jp": {
+		"enable": true,
+		"model_path": "kokomi-jp.pth",
+        "title": "Genshin Impact - Sangonomiya Kokomi",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1723_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+    "layla-jp": {
+		"enable": true,
+		"model_path": "layla-jp.pth",
+        "title": "Genshin Impact - Layla",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1099_Flat_nprobe_1_layla-v2_v2.index",
+		"author": "HirumiM"
+    },
+    "lisa-jp": {
+		"enable": true,
+		"model_path": "lisa-jp.pth",
+        "title": "Genshin Impact - Lisa",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF758_Flat_nprobe_1_lisa_v2.index",
+		"author": "Mocci24"
+    },
+	"lumine-jp": {
+		"enable": true,
+		"model_path": "lumine-jp.pth",
+        "title": "Genshin Impact - Lumine",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1165_Flat_nprobe_1_lumine-v2-new_v2.index",
+		"author":"HirumiM"
+    },
+    "mona-jp": {
+		"enable": true,
+		"model_path": "mona-jp.pth",
+        "title": "Genshin Impact - Mona",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF2165_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+    },
+	"nahida-jp": {
+		"enable": true,
+		"model_path": "nahida-jp.pth",
+        "title": "Genshin Impact - Nahida",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF265_Flat_nprobe_5.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"nilou-jp": {
+		"enable": true,
+		"model_path": "nilou-jp.pth",
+        "title": "Genshin Impact - Nilou",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF218_Flat_nprobe_1.index",
+		"feature_file": "total_fea.npy",
+		"author":"ArkanDash"
+    },
+    "noelle-jp": {
+		"enable": true,
+		"model_path": "noelle-jp.pth",
+        "title": "Genshin Impact - Noelle",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF3498_Flat_nprobe_1_noelle-jp_v2.index",
+		"author":"ArkanDash"
+    },
+	"paimon-jp": {
+		"enable": true,
+		"model_path": "paimon-jp.pth",
+        "title": "Genshin Impact - Paimon",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF3904_Flat_nprobe_1_paimon-jp_v2.index",
+		"author":"ArkanDash"
+    },
+	"raiden-jp": {
+		"enable": true,
+		"model_path": "raiden-jp.pth",
+        "title": "Genshin Impact - Raiden Shogun",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF783_Flat_nprobe_7.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"shinobu-jp": {
+		"enable": true,
+		"model_path": "shinobu-jp.pth",
+		"title": "Genshin Impact - Kuki Shinobu",
+		"cover": "cover.png",
+		"feature_retrieval_library": "added_IVF4218_Flat_nprobe_1_shinobu-jp_v2.index",
+		"author": "ArkanDash"
+	},
+	"signora-jp": {
+		"enable": true,
+		"model_path": "signora-jp.pth",
+        "title": "Genshin Impact - Rosalyne-Kruzchka Lohefalter, aka La Signora or 'The Fair Lady'",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF478_Flat_nprobe_1_signora-jp_v2.index",
+		"author":"ArkanDash"
+    },
+    "sucrose-jp": {
+		"enable": true,
+		"model_path": "sucrose-jp.pth",
+        "title": "Genshin Impact - Sucrose",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF884_Flat_nprobe_1.index",
+        "feature_file": "total_fea.npy",
+		"author":"Mocci24"
+    },
+	"thoma-jp": {
+		"enable": true,
+		"model_path": "thoma-jp.pth",
+        "title": "Genshin Impact - Thoma",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF366_Flat_nprobe_1.index",
+		"feature_file": "total_fea.npy",
+		"author":"Eugenia"
+    },
+	"tighnari-jp": {
+		"enable": true,
+		"model_path": "tighnari-jp.pth",
+        "title": "Genshin Impact - Tighnari",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF446_Flat_nprobe_1_tignari-jp_v1.index",
+		"feature_file": "total_fea.npy",
+		"author":"KarmKarma"
+    },
+	"venti-jp": {
+		"enable": true,
+		"model_path": "venti-jp.pth",
+        "title": "Genshin Impact - Venti",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF3591_Flat_nprobe_1_venti-jp_v2.index",
+		"author":"ArkanDash"
+    },
+	"wanderer-jp": {
+		"enable": true,
+		"model_path": "wanderer-jp.pth",
+        "title": "Genshin Impact - Wanderer",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF128_Flat_nprobe_4.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"xiao-jp": {
+		"enable": true,
+		"model_path": "xiao-jp.pth",
+        "title": "Genshin Impact - Xiao",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF233_Flat_nprobe_5.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+    "xinyan-jp": {
+		"enable": true,
+		"model_path": "xinyan-jp.pth",
+        "title": "Genshin Impact - Xinyan",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF971_Flat_nprobe_1_xinyan_v2.index",
+		"author": "HirumiM"
+    },
+	"yae-jp": {
+		"enable": true,
+		"model_path": "yae-jp.pth",
+        "title": "Genshin Impact - Yae Miko",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF935_Flat_nprobe_1.index",
+		"feature_file": "total_fea.npy",
+		"author":"Ardha"
+    },
+	"yanfei-jp": {
+		"enable": true,
+		"model_path": "yanfei-jp.pth",
+        "title": "Genshin Impact - Yanfei",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1271_Flat_nprobe_1_yanfei-v2_v2.index",
+		"author": "HirumiM"
+    },
+    "yelan-jp": {
+		"enable": true,
+		"model_path": "yelan-jp.pth",
+        "title": "Genshin Impact - Yelan",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF2051_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+	},
+    "yoimiya-jp": {
+		"enable": true,
+		"model_path": "yoimiya-jp.pth",
+        "title": "Genshin Impact - Yoimiya",
+        "cover": "cover.png",
+		"feature_retrieval_library": "added_IVF2034_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+	},
+    "zhongli-jp": {
+		"enable": true,
+		"model_path": "zhongli-jp.pth",
+		"title": "Genshin Impact - Zhongli",
+		"cover": "cover.png",
+		"feature_retrieval_library": "added_IVF1672_Flat_nprobe_1.index",
+		"author":"ArkanDash"
+	}
+}
\ No newline at end of file
diff --git a/weights/mona-jp/added_IVF2165_Flat_nprobe_1.index b/weights/mona-jp/added_IVF2165_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..3c0be5910c81908a2e996ef20df2ee30c58ef6d2
--- /dev/null
+++ b/weights/mona-jp/added_IVF2165_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f1b5f79673815bfe59bd02176e36a736890728db6e7a4d08a78ae164ffc1bcf
+size 89383723
diff --git a/weights/mona-jp/cover.png b/weights/mona-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..367715bc261c5954f6b150bcf9748804393cd875
--- /dev/null
+++ b/weights/mona-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:850d09a002fc645e45b097cedf3f05ee4db1b403e7f522525bb90be58c3e9a9f
+size 2756373
diff --git a/weights/mona-jp/mona-jp.pth b/weights/mona-jp/mona-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..52eceb875676d3550e9a8135af956598e41544ad
--- /dev/null
+++ b/weights/mona-jp/mona-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d334231959d3eb9183f41d9d013136bc7163029f9fe2a44dc57487ecbb862188
+size 55124473
diff --git a/weights/nahida-jp/added_IVF265_Flat_nprobe_5.index b/weights/nahida-jp/added_IVF265_Flat_nprobe_5.index
new file mode 100644
index 0000000000000000000000000000000000000000..6f8a20c89e5d9c77ed2d9c0e9290d919717eea29
--- /dev/null
+++ b/weights/nahida-jp/added_IVF265_Flat_nprobe_5.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:223ac68eca791ce8ce7aaa4f56fcf3d7ddd9e5145f709a017191b0c9720daead
+size 10942435
diff --git a/weights/nahida-jp/cover.png b/weights/nahida-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..162a0f6a2eb70db75f6deff67fac3a3c99c26f14
--- /dev/null
+++ b/weights/nahida-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a40241729e26adf2d3500fd74e1a9fb082ef52578bf5dd7af4534aeac6089108
+size 5963743
diff --git a/weights/nahida-jp/nahida-jp.pth b/weights/nahida-jp/nahida-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..9186a0873944daebe3c4153fbfd0ae39477be3c3
--- /dev/null
+++ b/weights/nahida-jp/nahida-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c598ebc6b4a2f2f4dfb5a9e46851ffd2e20b9f8adf2743fc88db11fc3836748c
+size 55027589
diff --git a/weights/nahida-jp/total_fea.npy b/weights/nahida-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..9c60e76c88f8e2a575a35a4e49dbe5a0788af918
--- /dev/null
+++ b/weights/nahida-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8552383d1feae96bdea2008261709de4ffd9ae2beebafcf519d3fe8f1e18b598
+size 10586240
diff --git a/weights/nilou-jp/added_IVF2384_Flat_nprobe_1.index b/weights/nilou-jp/added_IVF2384_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..2c9dd6c9ed103bfdca3b59fbb23d8af92fb7d21a
--- /dev/null
+++ b/weights/nilou-jp/added_IVF2384_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96437f096c1afb1c1de18d5beeedd7d95a688b801687013a5e7e5f5269f49c38
+size 98433331
diff --git a/weights/nilou-jp/cover.png b/weights/nilou-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..16c97a930a9b94fa55f10cd6f7573804fde36dc4
--- /dev/null
+++ b/weights/nilou-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8baa3e813b4025804901d7b9028d3a08ff8b2814e0cc52ea9ad83326f929a521
+size 151336
diff --git a/weights/nilou-jp/nilou-jp.pth b/weights/nilou-jp/nilou-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..e5e78bc228ad97ddccefb7386dbe8566145aee5b
--- /dev/null
+++ b/weights/nilou-jp/nilou-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e64fe080b45e1a2b539090f88eb1813fc7310239bdc5e3abe3250d28e383d4af
+size 55027130
diff --git a/weights/noelle-jp/added_IVF3498_Flat_nprobe_1_noelle-jp_v2.index b/weights/noelle-jp/added_IVF3498_Flat_nprobe_1_noelle-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..3d4edb144ccc7dbd08c7b793662fececd18c926b
--- /dev/null
+++ b/weights/noelle-jp/added_IVF3498_Flat_nprobe_1_noelle-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b1826ace2f2725c676a32256f4443663b6422676b7c244565c7326efa08f866
+size 430969139
diff --git a/weights/noelle-jp/cover.png b/weights/noelle-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..7bd4c366e7816ed5e8f7d3c3e6f508a693033081
--- /dev/null
+++ b/weights/noelle-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3678daa9cdc1179951a08d00f1125bd997dea0a4c2fcc6c33bd91985fdfd72df
+size 1292298
diff --git a/weights/noelle-jp/noelle-jp.pth b/weights/noelle-jp/noelle-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..62875a2864c608815821f60946b46cffe9afa90c
--- /dev/null
+++ b/weights/noelle-jp/noelle-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6b77c63ea21631ae11edde6c7c987cad5efdfdf3eb1a31687a858497b9beb07
+size 57589001
diff --git a/weights/paimon-jp/added_IVF3904_Flat_nprobe_1_paimon-jp_v2.index b/weights/paimon-jp/added_IVF3904_Flat_nprobe_1_paimon-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..e2cce762b7b822c715f72726d2fe76a21270e3de
--- /dev/null
+++ b/weights/paimon-jp/added_IVF3904_Flat_nprobe_1_paimon-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41f253fd335a70447d3a2095e1eedf5909dc6e1e8b3f8a9fb9b75eedb3b57f7d
+size 481003739
diff --git a/weights/paimon-jp/cover.png b/weights/paimon-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..e919e9c75ba7c6e4418b1df78635e94f3caa336d
--- /dev/null
+++ b/weights/paimon-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c37f81e3eb60c88f23f18907798cb77bf62f2fe3d8aa6a49be0424a13f885aa6
+size 365552
diff --git a/weights/paimon-jp/paimon-jp.pth b/weights/paimon-jp/paimon-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..b1e00df9b4686b79eb160331dab20f812a8f0d10
--- /dev/null
+++ b/weights/paimon-jp/paimon-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a313f38098e2dd3a84e637607ecb3eb92948bfee533746014a5589162e4ef23d
+size 57589001
diff --git a/weights/raiden-jp/added_IVF783_Flat_nprobe_7.index b/weights/raiden-jp/added_IVF783_Flat_nprobe_7.index
new file mode 100644
index 0000000000000000000000000000000000000000..cdb04b02df5f6207adf94b394daca67773280427
--- /dev/null
+++ b/weights/raiden-jp/added_IVF783_Flat_nprobe_7.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71e9568b307152547c45653c6e464a991223071933858047cb81cfbd9fbb9679
+size 32349211
diff --git a/weights/raiden-jp/cover.png b/weights/raiden-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..9715577da3f3267dff2762a50ef2de39e18a4c4e
--- /dev/null
+++ b/weights/raiden-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae435b5488b7ba6f911616e9fa8d3b4d4d8f61ea687af1f23f6083b7d9067ee7
+size 8036189
diff --git a/weights/raiden-jp/raiden-jp.pth b/weights/raiden-jp/raiden-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..1b119e9d8aa2a1fb863921524aad52701d394486
--- /dev/null
+++ b/weights/raiden-jp/raiden-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7483e661b095a272b0c0fb7aa77075c7359d087c8d3e3ee8c362985a3d96fdd1
+size 55027589
diff --git a/weights/raiden-jp/total_fea.npy b/weights/raiden-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..07ce56cea533116472165f886944ac6096cd926f
--- /dev/null
+++ b/weights/raiden-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1374d1575f9ed5145bb7ea1d093c48ebe6142b9d0ea10c7ad0fc8ac3c8d1993a
+size 31296640
diff --git a/weights/shinobu-jp/added_IVF4218_Flat_nprobe_1_shinobu-jp_v2.index b/weights/shinobu-jp/added_IVF4218_Flat_nprobe_1_shinobu-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..9276baca01384b4028b5fa1e052e6c471000a40a
--- /dev/null
+++ b/weights/shinobu-jp/added_IVF4218_Flat_nprobe_1_shinobu-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b19ea6ca6e55cb5ed1500da4b13553a21a6a815a266857e06259748edd8b460
+size 519676219
diff --git a/weights/shinobu-jp/cover.png b/weights/shinobu-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..998a00257f80e6d7258fe4255fe3f0ab3873ef0f
--- /dev/null
+++ b/weights/shinobu-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bde21a8c7324e0c77448d0021e1b9a58dcc89539765b7cbd830e430a679f832
+size 583602
diff --git a/weights/shinobu-jp/shinobu-jp.pth b/weights/shinobu-jp/shinobu-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..37549fae342f1effeca7f947bb77a75bb9ee6a27
--- /dev/null
+++ b/weights/shinobu-jp/shinobu-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3253fa46c13a42a4e6de3a0b79b69d58bed62dd17991f7cafda9640e57ffb33
+size 57588083
diff --git a/weights/signora-jp/added_IVF478_Flat_nprobe_1_signora-jp_v2.index b/weights/signora-jp/added_IVF478_Flat_nprobe_1_signora-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..6ca4fef36ab6e5b9fd169c3a59ac40d877b32d5a
--- /dev/null
+++ b/weights/signora-jp/added_IVF478_Flat_nprobe_1_signora-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d79f3865e6d416916b4c42e7cf798b6a21da768968b1ef8a895bb77cf2806065
+size 58948259
diff --git a/weights/signora-jp/cover.png b/weights/signora-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..b14547726a9a7b7417df9f711148a6df1d76386e
--- /dev/null
+++ b/weights/signora-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d85c88748da0ffcf4d0515a5f91c99104b792789bd2e31ce021fec43fde32fc6
+size 460720
diff --git a/weights/signora-jp/signora-jp.pth b/weights/signora-jp/signora-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..1090c495ee1d331355720f51c952dba3a8a5f124
--- /dev/null
+++ b/weights/signora-jp/signora-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f0562d316357bf3c1af86bf228a69fd6a1e31e2f917528cf3908076d86bd05b
+size 57589983
diff --git a/weights/sucrose-jp/added_IVF884_Flat_nprobe_1.index b/weights/sucrose-jp/added_IVF884_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..9f15329db441300b4d408031f1e6ad1ce42acbc9
--- /dev/null
+++ b/weights/sucrose-jp/added_IVF884_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cc1bd428bcc9a7677c6c9799ad7a90aa02e33fc8116ab443fd04b964014cecd
+size 36516427
diff --git a/weights/sucrose-jp/cover.png b/weights/sucrose-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..b87a20c176ee3c1a97746dd09c346558a18713bb
--- /dev/null
+++ b/weights/sucrose-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65a648e980eb0d1126a85829e2d78b070922e8c2fac808151b913b4a34674c9c
+size 224600
diff --git a/weights/sucrose-jp/sucrose-jp.pth b/weights/sucrose-jp/sucrose-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..2901ee990881e27c7547a24b073bf9259011ec8e
--- /dev/null
+++ b/weights/sucrose-jp/sucrose-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de06f5c9b36cf276c48cdc9dc98756cf2357a5bef8375c945d77ea079c208655
+size 55026095
diff --git a/weights/sucrose-jp/total_fea.npy b/weights/sucrose-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..0f3d0ec93b76657facd26b8700bfd578813aa33b
--- /dev/null
+++ b/weights/sucrose-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69ced72b5675af5602fbe62f63d1a2b812d7f1c306de75c2cc4b89516743c9e1
+size 35328128
diff --git a/weights/thoma-jp/added_IVF366_Flat_nprobe_1.index b/weights/thoma-jp/added_IVF366_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..51cb68372c315c711943028a32315bd5110c6c69
--- /dev/null
+++ b/weights/thoma-jp/added_IVF366_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7434c35d2b8462cea1e93d2aa215b3995ce9ee5660cae9fc5c62072197618097
+size 15144739
diff --git a/weights/thoma-jp/cover.png b/weights/thoma-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..ba36555b4bbb2cc930c860edada3d69b769540e7
--- /dev/null
+++ b/weights/thoma-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8c4918b71ce15cbdf839ec37b2ab1c699892aa8b864f992fd697b5cb12d32a7
+size 2981457
diff --git a/weights/thoma-jp/thoma-jp.pth b/weights/thoma-jp/thoma-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..4c5aa8ca74c0812a44b133db617534aeb6e79892
--- /dev/null
+++ b/weights/thoma-jp/thoma-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79308af33fe821940d8df3d48f9419a2d9d70df81e29a62c7a3cf1f7b3f5368d
+size 55027130
diff --git a/weights/thoma-jp/total_fea.npy b/weights/thoma-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..d4f7e0e205b0f45c7a9f15fd3dab9bbf1d56ef82
--- /dev/null
+++ b/weights/thoma-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4533c1becfc69ff1746c5ca1bb042352a81fc2901d40efa23bc58b528c87f0fc
+size 14652544
diff --git a/weights/tighnari-jp/added_IVF446_Flat_nprobe_1_tignari-jp_v1.index b/weights/tighnari-jp/added_IVF446_Flat_nprobe_1_tignari-jp_v1.index
new file mode 100644
index 0000000000000000000000000000000000000000..62d13a0f90b9c498f8b89278fb9d5069ec6c4461
--- /dev/null
+++ b/weights/tighnari-jp/added_IVF446_Flat_nprobe_1_tignari-jp_v1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88827f9468860a8a37151cde4f593ce3ece1719d9fe5d4dff4b5c301ede53a3f
+size 18418243
diff --git a/weights/tighnari-jp/cover.png b/weights/tighnari-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..f4c565224c3f5935bb7ca095012f2a8fe84c0de4
--- /dev/null
+++ b/weights/tighnari-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b64bb3e8ebe4fe64d147dcb46c5fe07f672f4f5d47662017a038c993d3e09b3b
+size 1859544
diff --git a/weights/tighnari-jp/tighnari-jp.pth b/weights/tighnari-jp/tighnari-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..34b7f9863cc7d065a9d37b4bcba671bb49705345
--- /dev/null
+++ b/weights/tighnari-jp/tighnari-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93662bed0e7f208a16c1ac36d008b086f1eb1c28b3a7a0c3a7612a784158d26e
+size 55028048
diff --git a/weights/tighnari-jp/total_fea.npy b/weights/tighnari-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..0d57e0eda88d9e22eee20edf0fc3ae88bf26281d
--- /dev/null
+++ b/weights/tighnari-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cfa85f717c10e354c9414db146f7239a7391cbab7a5e18b770496f65cc84a64
+size 17818752
diff --git a/weights/venti-jp/added_IVF3591_Flat_nprobe_1_venti-jp_v2.index b/weights/venti-jp/added_IVF3591_Flat_nprobe_1_venti-jp_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..055d9ea9d4d865fac31cc673c008575cd05b41f8
--- /dev/null
+++ b/weights/venti-jp/added_IVF3591_Flat_nprobe_1_venti-jp_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b4561f58fcff1371e803bb910da5c182f2faa92a66737612f6c12fef578f19c
+size 442503739
diff --git a/weights/venti-jp/cover.png b/weights/venti-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..533db20d4ff140f34d9727b4b744aeef61a1d5d4
--- /dev/null
+++ b/weights/venti-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a6d04b831c913989af37aa317dc6d57c2d492386d5a2d39df037aeb7079d8f0
+size 8028474
diff --git a/weights/venti-jp/venti-jp.pth b/weights/venti-jp/venti-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..f274ae9e185fcec7f26b6aa18565e4cd186b9c83
--- /dev/null
+++ b/weights/venti-jp/venti-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd1c1885d3340b24213dd4bfe1bb0449eb5f32eab3a7a9ca5e0bf389f4a7f2e0
+size 57588542
diff --git a/weights/wanderer-jp/added_IVF128_Flat_nprobe_4.index b/weights/wanderer-jp/added_IVF128_Flat_nprobe_4.index
new file mode 100644
index 0000000000000000000000000000000000000000..ce54db39fe85ac939452aadd69cc1741f5bcd9fc
--- /dev/null
+++ b/weights/wanderer-jp/added_IVF128_Flat_nprobe_4.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6642c9a48863247025fd9fb09e8de4ffa220ef659dff1a4a297c57bd08cf1ec4
+size 5314939
diff --git a/weights/wanderer-jp/cover.png b/weights/wanderer-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..3a85da4b70e37855f15f4bb546e220c58971c6d4
--- /dev/null
+++ b/weights/wanderer-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af68ce227760b354d3c204b38589d328623ed869973d16db05daef784026526a
+size 12522761
diff --git a/weights/wanderer-jp/total_fea.npy b/weights/wanderer-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..aca280acfcac1c9f65b4c22637d24c32e8ec2127
--- /dev/null
+++ b/weights/wanderer-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edccbafbaa8512f9970686d305eb10b4ff5f9d79ca7118d3158b8699aac1975e
+size 5142656
diff --git a/weights/wanderer-jp/wanderer-jp.pth b/weights/wanderer-jp/wanderer-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..7f03d9fd91a830477f37ecd7cbab2abf20bd454b
--- /dev/null
+++ b/weights/wanderer-jp/wanderer-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d2d89c3cac427a5ae0439b2cd635015f20e8d4e1d2d6c0af39daa44f6c02890
+size 55028507
diff --git a/weights/xiao-jp/added_IVF233_Flat_nprobe_5.index b/weights/xiao-jp/added_IVF233_Flat_nprobe_5.index
new file mode 100644
index 0000000000000000000000000000000000000000..a5ae5974faffe91e6980205ad1d36e26972b22c6
--- /dev/null
+++ b/weights/xiao-jp/added_IVF233_Flat_nprobe_5.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65c124d9021bf14703a4b4446f86dc0356f8ab43c46b99efdaab183ccb42ccf8
+size 9641083
diff --git a/weights/xiao-jp/cover.png b/weights/xiao-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..02234b440e411f22104dc245100ea2f96e2e9154
--- /dev/null
+++ b/weights/xiao-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de0e9f3dfa8c1e5266fd519fc23e7365d06c1375caa47e4cc7cfcda4d1e26888
+size 10456216
diff --git a/weights/xiao-jp/total_fea.npy b/weights/xiao-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..42737c93b5ebc486dc64ecf2f80933acd9cf8cdf
--- /dev/null
+++ b/weights/xiao-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3650e806c3e9098c4a836d138c141b2d387351765317721d2331ff3428361fbe
+size 9327744
diff --git a/weights/xiao-jp/xiao-jp.pth b/weights/xiao-jp/xiao-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..8c3e534232a678288a9dd7e073027e83609e89bc
--- /dev/null
+++ b/weights/xiao-jp/xiao-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f85bf34a96dffe7dcaec4fb434f0ca1ed08b56784ac9ab88545ab2df8b66e1a
+size 55026095
diff --git a/weights/xinyan-jp/added_IVF971_Flat_nprobe_1_xinyan_v2.index b/weights/xinyan-jp/added_IVF971_Flat_nprobe_1_xinyan_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..a9256c3aac824afeb5f18f32589438be42bd42b8
--- /dev/null
+++ b/weights/xinyan-jp/added_IVF971_Flat_nprobe_1_xinyan_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94955edd259bd10becaed2c4ccf67bd4c5a88af75c949bcd3d14cee6156c155d
+size 119735139
diff --git a/weights/xinyan-jp/cover.png b/weights/xinyan-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..5009f6d6e6cf730cb2c5064030be18b34aa9f7da
--- /dev/null
+++ b/weights/xinyan-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46814695112bc091f81f4e208481245297d625459757dca4d76af707ca87009e
+size 294265
diff --git a/weights/xinyan-jp/xinyan-jp.pth b/weights/xinyan-jp/xinyan-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..903a42baa106979f80bdf58a1ff2416bd40bfacd
--- /dev/null
+++ b/weights/xinyan-jp/xinyan-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9dec1ae0b4492343a07d66e1ba4b0216679dd5fed0bae39720805da113802c35
+size 55216420
diff --git a/weights/yae-jp/added_IVF935_Flat_nprobe_1.index b/weights/yae-jp/added_IVF935_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..4a1e7575b0220604e14317af55c7f1eb2ba43567
--- /dev/null
+++ b/weights/yae-jp/added_IVF935_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f82e237ac49c261015c5159e4335f001653a5d3622ce7d7be95533060052314c
+size 38622739
diff --git a/weights/yae-jp/cover.png b/weights/yae-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..651702b26b79bce9931189d481291405fb80afcc
--- /dev/null
+++ b/weights/yae-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b1f7e1b6fa0dde40ed9cc4b9a0ad6296f8a890c23c73dc897ea8f6734d74be6
+size 5027456
diff --git a/weights/yae-jp/total_fea.npy b/weights/yae-jp/total_fea.npy
new file mode 100644
index 0000000000000000000000000000000000000000..b04f7ca3c44b29c737fd93f69694069d4461a216
--- /dev/null
+++ b/weights/yae-jp/total_fea.npy
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e89f702186c72b99d4821ef641c4a057a335b5c0a273cc04e06a48ab9981151
+size 37365888
diff --git a/weights/yae-jp/yae-jp.pth b/weights/yae-jp/yae-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..7b4d5c92816307cea3b9bdea6a03dcf49f76e41f
--- /dev/null
+++ b/weights/yae-jp/yae-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdab6dbf1c3d045c872822b21e780fa6c8037cf88288d216abe8d69c3b02b8b1
+size 54995587
diff --git a/weights/yanfei-jp/added_IVF1271_Flat_nprobe_1_yanfei-v2_v2.index b/weights/yanfei-jp/added_IVF1271_Flat_nprobe_1_yanfei-v2_v2.index
new file mode 100644
index 0000000000000000000000000000000000000000..df7ae2b4382692cfee6d68d9d83fd733c32004ef
--- /dev/null
+++ b/weights/yanfei-jp/added_IVF1271_Flat_nprobe_1_yanfei-v2_v2.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cf71cbde75213b451e5850ce6f6bde5647de1a1642a56df15367c0ff76783ba
+size 156608899
diff --git a/weights/yanfei-jp/cover.png b/weights/yanfei-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..cf542f8d862f71b1a8ddf0fc7aa7e7d3691cc4b9
--- /dev/null
+++ b/weights/yanfei-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d0687136bf908a0b05969cfd39dc396d1411a6f7e71b4a4b6d521bd5fa0c840
+size 310936
diff --git a/weights/yanfei-jp/yanfei-jp.pth b/weights/yanfei-jp/yanfei-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..28e57b006a425cdfe93d91d973761afd0688c77a
--- /dev/null
+++ b/weights/yanfei-jp/yanfei-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb0b7c6abad6e6f854d6a9c7a504147557481939fcfce837884d59784076e868
+size 55224197
diff --git a/weights/yelan-jp/added_IVF2051_Flat_nprobe_1.index b/weights/yelan-jp/added_IVF2051_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..d15cb9258a7bf4ea824d5df70ecf31a2bca856f5
--- /dev/null
+++ b/weights/yelan-jp/added_IVF2051_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba725109af57b0c32075582e17571e59ee3a3c33b34b00e13481bd15c0bf0920
+size 84694315
diff --git a/weights/yelan-jp/cover.png b/weights/yelan-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..84c5e0d29602dc83c5e85300bae6e1bd9d66775c
--- /dev/null
+++ b/weights/yelan-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:415727226e6fd9e15042c69c925848ed80b39701ffe954f35b91d59437d9176f
+size 583602
diff --git a/weights/yelan-jp/yelan-jp.pth b/weights/yelan-jp/yelan-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..6689d025fe95cfdfe02e41015794d4e3cd2f3cb8
--- /dev/null
+++ b/weights/yelan-jp/yelan-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d62ff825db07022e7d380d13c827f27730c99f6d029e8fb8b1d7cd90018d3c49
+size 55027130
diff --git a/weights/yoimiya-jp/added_IVF2034_Flat_nprobe_1.index b/weights/yoimiya-jp/added_IVF2034_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..8eb29f4a40f427f9b105738ad7bd71224d51ac0a
--- /dev/null
+++ b/weights/yoimiya-jp/added_IVF2034_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af434a9142b070f7091dcdbbf957b7a01bbc96294add99d186ef1e0d4b226eac
+size 83987395
diff --git a/weights/yoimiya-jp/cover.png b/weights/yoimiya-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..be502b43c4c802982ecbfeed4ea3385dd4ef65f2
--- /dev/null
+++ b/weights/yoimiya-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c044b81dd758a4efac819bf8e86c6a3dc565517e6c77525c36245ca26b73b42f
+size 199412
diff --git a/weights/yoimiya-jp/yoimiya-jp.pth b/weights/yoimiya-jp/yoimiya-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..84d7a9eb2db180ec681ac8b87c37ef8b352deff3
--- /dev/null
+++ b/weights/yoimiya-jp/yoimiya-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:896fcee182ecdcea6645a691366ac50153bc63015f43c981da135a8cabe2f088
+size 55028048
diff --git a/weights/zhongli-jp/added_IVF1672_Flat_nprobe_1.index b/weights/zhongli-jp/added_IVF1672_Flat_nprobe_1.index
new file mode 100644
index 0000000000000000000000000000000000000000..a59a5bed1cbedad0f946e73e56667658f59ececb
--- /dev/null
+++ b/weights/zhongli-jp/added_IVF1672_Flat_nprobe_1.index
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f112d545b6b6ea3063b91d072a745e38541bad2141273b84896102f727cdbe14
+size 69024427
diff --git a/weights/zhongli-jp/cover.png b/weights/zhongli-jp/cover.png
new file mode 100644
index 0000000000000000000000000000000000000000..a3eb27010d8baee2279b079a55fffcaa32ba137f
--- /dev/null
+++ b/weights/zhongli-jp/cover.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b2ba0d51052e547052fd14affd7bf74c4dc3c3ee114ca94b45b058b137a2dea
+size 2070002
diff --git a/weights/zhongli-jp/zhongli-jp.pth b/weights/zhongli-jp/zhongli-jp.pth
new file mode 100644
index 0000000000000000000000000000000000000000..c3d5b9c9eff9c9540c7a7b6895eb9cfa21e49ff9
--- /dev/null
+++ b/weights/zhongli-jp/zhongli-jp.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f31eb1a9446b26cf0893898b75c0a93ed83746cad9fc1f73f0990e7418afc23c
+size 55028048