blackteay/GPT-SoVITS
1
0
Fork 0
mirror of https://github.com/RVC-Boss/GPT-SoVITS.git synced 2025-04-06 03:57:44 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #438 from NaruseMioShirakana/main

Browse Source 
Update Onnx Export
This commit is contained in:
RVC-Boss 2024-02-08 21:56:07 +08:00 committed by GitHub
commit f02aab9892
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 66 additions and 59 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
GPT_SoVITS/AR/models/t2s_model_onnx.py
View File

@ -57,7 +57,7 @@ def logits_to_probs(
logits = logits / max(temperature, 1e-5) logits = logits / max(temperature, 1e-5)
if top_k is not None: if top_k is not None:
v, _ = torch.topk(logits, min(top_k, logits.size(-1))) v, _ = torch.topk(logits, top_k)
pivot = v.select(-1, -1).unsqueeze(-1) pivot = v.select(-1, -1).unsqueeze(-1)
logits = torch.where(logits < pivot, inf_tensor_value, logits) logits = torch.where(logits < pivot, inf_tensor_value, logits)

47
GPT_SoVITS/module/attentions_onnx.py
View File

@ -188,38 +188,27 @@ class MultiHeadAttention(nn.Module):
query = query.view(b, self.n_heads, self.k_channels, -1).transpose(2, 3) query = query.view(b, self.n_heads, self.k_channels, -1).transpose(2, 3)
key = key.view(b, self.n_heads, self.k_channels, -1).transpose(2, 3) key = key.view(b, self.n_heads, self.k_channels, -1).transpose(2, 3)
value = value.view(b, self.n_heads, self.k_channels, -1).transpose(2, 3) value = value.view(b, self.n_heads, self.k_channels, -1).transpose(2, 3)
scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1)) scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1))
if self.window_size is not None: if self.window_size is not None:
key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s) key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s)
rel_logits = self._matmul_with_relative_keys( rel_logits = self._matmul_with_relative_keys(query / math.sqrt(self.k_channels), key_relative_embeddings)
query / math.sqrt(self.k_channels), key_relative_embeddings
)
scores_local = self._relative_position_to_absolute_position(rel_logits) scores_local = self._relative_position_to_absolute_position(rel_logits)
scores = scores + scores_local scores = scores + scores_local
if mask is not None: if mask is not None:
scores = scores.masked_fill(mask == 0, -1e4) scores = scores.masked_fill(mask == 0, -1e4)
if self.block_length is not None:
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) p_attn = F.softmax(scores, dim=-1)
p_attn = self.drop(p_attn) p_attn = self.drop(p_attn)
output = torch.matmul(p_attn, value) output = torch.matmul(p_attn, value)
if self.window_size is not None: if self.window_size is not None:
relative_weights = self._absolute_position_to_relative_position(p_attn) relative_weights = self._absolute_position_to_relative_position(p_attn)
value_relative_embeddings = self._get_relative_embeddings( value_relative_embeddings = self._get_relative_embeddings(self.emb_rel_v, t_s)
self.emb_rel_v, t_s output = output + self._matmul_with_relative_values(relative_weights, value_relative_embeddings)
)
output = output + self._matmul_with_relative_values( output = (output.transpose(2, 3).contiguous().view(b, d, -1))
relative_weights, value_relative_embeddings
)
output = (
output.transpose(2, 3).contiguous().view(b, d, -1)
)
return output, p_attn return output, p_attn
def _matmul_with_relative_values(self, x, y): def _matmul_with_relative_values(self, x, y):
@ -243,16 +232,16 @@ class MultiHeadAttention(nn.Module):
def _get_relative_embeddings(self, relative_embeddings, length): def _get_relative_embeddings(self, relative_embeddings, length):
max_relative_position = 2 * self.window_size + 1 max_relative_position = 2 * self.window_size + 1
# Pad first before slice to avoid using cond ops. # Pad first before slice to avoid using cond ops.
pad_length = max(length - (self.window_size + 1), 0) pad_l = torch.zeros((1), dtype = torch.int64) + length - (self.window_size + 1)
slice_start_position = max((self.window_size + 1) - length, 0) pad_s = torch.zeros((1), dtype = torch.int64) + (self.window_size + 1) - length
pad_length = torch.max(pad_l, other=torch.zeros((1), dtype = torch.int64))
slice_start_position = torch.max(pad_s, other=torch.zeros((1), dtype = torch.int64))
slice_end_position = slice_start_position + 2 * length - 1 slice_end_position = slice_start_position + 2 * length - 1
if pad_length > 0: padded_relative_embeddings = F.pad(
padded_relative_embeddings = F.pad( relative_embeddings,
relative_embeddings, commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]),
commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]), )
)
else:
padded_relative_embeddings = relative_embeddings
used_relative_embeddings = padded_relative_embeddings[ used_relative_embeddings = padded_relative_embeddings[
:, slice_start_position:slice_end_position :, slice_start_position:slice_end_position
] ]

4
GPT_SoVITS/module/models_onnx.py
View File

@ -896,9 +896,6 @@ class SynthesizerTrn(nn.Module):
refer_mask = torch.ones_like(refer[:1,:1,:]) refer_mask = torch.ones_like(refer[:1,:1,:])
ge = self.ref_enc(refer * refer_mask, refer_mask) ge = self.ref_enc(refer * refer_mask, refer_mask)
y_lengths = torch.LongTensor([codes.size(2) * 2]).to(codes.device)
text_lengths = torch.LongTensor([text.size(-1)]).to(text.device)
quantized = self.quantizer.decode(codes) quantized = self.quantizer.decode(codes)
if self.semantic_frame_rate == "25hz": if self.semantic_frame_rate == "25hz":
dquantized = torch.cat([quantized, quantized]).permute(1, 2, 0) dquantized = torch.cat([quantized, quantized]).permute(1, 2, 0)
@ -907,6 +904,7 @@ class SynthesizerTrn(nn.Module):
x, m_p, logs_p, y_mask = self.enc_p( x, m_p, logs_p, y_mask = self.enc_p(
quantized, text, ge quantized, text, ge
) )
z_p = m_p + torch.randn_like(m_p) * torch.exp(logs_p) z_p = m_p + torch.randn_like(m_p) * torch.exp(logs_p)
z = self.flow(z_p, y_mask, g=ge, reverse=True) z = self.flow(z_p, y_mask, g=ge, reverse=True)

72
GPT_SoVITS/onnx_export.py
View File

@ -140,6 +140,7 @@ class T2SModel(nn.Module):
) )
onnx_encoder_export_output.save(f"onnx/{project_name}/{project_name}_t2s_encoder.onnx") onnx_encoder_export_output.save(f"onnx/{project_name}/{project_name}_t2s_encoder.onnx")
return return
torch.onnx.export( torch.onnx.export(
self.onnx_encoder, self.onnx_encoder,
(ref_seq, text_seq, ref_bert, text_bert, ssl_content), (ref_seq, text_seq, ref_bert, text_bert, ssl_content),
@ -147,16 +148,16 @@ class T2SModel(nn.Module):
input_names=["ref_seq", "text_seq", "ref_bert", "text_bert", "ssl_content"], input_names=["ref_seq", "text_seq", "ref_bert", "text_bert", "ssl_content"],
output_names=["x", "prompts"], output_names=["x", "prompts"],
dynamic_axes={ dynamic_axes={
"ref_seq": [1], "ref_seq": {1 : "ref_length"},
"text_seq": [1], "text_seq": {1 : "text_length"},
"ref_bert": [0], "ref_bert": {0 : "ref_length"},
"text_bert": [0], "text_bert": {0 : "text_length"},
"ssl_content": [2], "ssl_content": {2 : "ssl_length"},
}, },
opset_version=16 opset_version=16
) )
x, prompts = self.onnx_encoder(ref_seq, text_seq, ref_bert, text_bert, ssl_content) x, prompts = self.onnx_encoder(ref_seq, text_seq, ref_bert, text_bert, ssl_content)
torch.exp
torch.onnx.export( torch.onnx.export(
self.first_stage_decoder, self.first_stage_decoder,
(x, prompts), (x, prompts),
@ -164,10 +165,10 @@ class T2SModel(nn.Module):
input_names=["x", "prompts"], input_names=["x", "prompts"],
output_names=["y", "k", "v", "y_emb", "x_example"], output_names=["y", "k", "v", "y_emb", "x_example"],
dynamic_axes={ dynamic_axes={
"x": [1], "x": {1 : "x_length"},
"prompts": [1], "prompts": {1 : "prompts_length"},
}, },
verbose=True, verbose=False,
opset_version=16 opset_version=16
) )
y, k, v, y_emb, x_example = self.first_stage_decoder(x, prompts) y, k, v, y_emb, x_example = self.first_stage_decoder(x, prompts)
@ -179,13 +180,13 @@ class T2SModel(nn.Module):
input_names=["iy", "ik", "iv", "iy_emb", "ix_example"], input_names=["iy", "ik", "iv", "iy_emb", "ix_example"],
output_names=["y", "k", "v", "y_emb", "logits", "samples"], output_names=["y", "k", "v", "y_emb", "logits", "samples"],
dynamic_axes={ dynamic_axes={
"iy": [1], "iy": {1 : "iy_length"},
"ik": [1], "ik": {1 : "ik_length"},
"iv": [1], "iv": {1 : "iv_length"},
"iy_emb": [1], "iy_emb": {1 : "iy_emb_length"},
"ix_example": [1], "ix_example": {1 : "ix_example_length"},
}, },
verbose=True, verbose=False,
opset_version=16 opset_version=16
) )
@ -224,9 +225,19 @@ class GptSoVits(nn.Module):
self.vits = vits self.vits = vits
self.t2s = t2s self.t2s = t2s
def forward(self, ref_seq, text_seq, ref_bert, text_bert, ref_audio, ssl_content): def forward(self, ref_seq, text_seq, ref_bert, text_bert, ref_audio, ssl_content, debug=False):
pred_semantic = self.t2s(ref_seq, text_seq, ref_bert, text_bert, ssl_content) pred_semantic = self.t2s(ref_seq, text_seq, ref_bert, text_bert, ssl_content)
return self.vits(text_seq, pred_semantic, ref_audio) audio = self.vits(text_seq, pred_semantic, ref_audio)
if debug:
import onnxruntime
sess = onnxruntime.InferenceSession("onnx/koharu/koharu_vits.onnx", providers=["CPU"])
audio1 = sess.run(None, {
"text_seq" : text_seq.detach().cpu().numpy(),
"pred_semantic" : pred_semantic.detach().cpu().numpy(),
"ref_audio" : ref_audio.detach().cpu().numpy()
})
return audio, audio1
return audio
def export(self, ref_seq, text_seq, ref_bert, text_bert, ref_audio, ssl_content, project_name): def export(self, ref_seq, text_seq, ref_bert, text_bert, ref_audio, ssl_content, project_name):
self.t2s.export(ref_seq, text_seq, ref_bert, text_bert, ssl_content, project_name) self.t2s.export(ref_seq, text_seq, ref_bert, text_bert, ssl_content, project_name)
@ -238,11 +249,12 @@ class GptSoVits(nn.Module):
input_names=["text_seq", "pred_semantic", "ref_audio"], input_names=["text_seq", "pred_semantic", "ref_audio"],
output_names=["audio"], output_names=["audio"],
dynamic_axes={ dynamic_axes={
"text_seq": [1], "text_seq": {1 : "text_length"},
"pred_semantic": [2], "pred_semantic": {2 : "pred_length"},
"ref_audio": [1], "ref_audio": {1 : "audio_length"},
}, },
opset_version=17 opset_version=17,
verbose=False
) )
@ -261,7 +273,7 @@ def export(vits_path, gpt_path, project_name):
gpt_sovits = GptSoVits(vits, gpt) gpt_sovits = GptSoVits(vits, gpt)
ssl = SSLModel() ssl = SSLModel()
ref_seq = torch.LongTensor([cleaned_text_to_sequence(["n", "i2", "h", "ao3", ",", "w", "o3", "sh", "i4", "b", "ai2", "y", "e4"])]) ref_seq = torch.LongTensor([cleaned_text_to_sequence(["n", "i2", "h", "ao3", ",", "w", "o3", "sh", "i4", "b", "ai2", "y", "e4"])])
text_seq = torch.LongTensor([cleaned_text_to_sequence(["w", "o3", "sh", "i4", "b", "ai2", "y", "e4"])]) text_seq = torch.LongTensor([cleaned_text_to_sequence(["w", "o3", "sh", "i4", "b", "ai2", "y", "e4", "w", "o3", "sh", "i4", "b", "ai2", "y", "e4", "w", "o3", "sh", "i4", "b", "ai2", "y", "e4"])])
ref_bert = torch.randn((ref_seq.shape[1], 1024)).float() ref_bert = torch.randn((ref_seq.shape[1], 1024)).float()
text_bert = torch.randn((text_seq.shape[1], 1024)).float() text_bert = torch.randn((text_seq.shape[1], 1024)).float()
ref_audio = torch.randn((1, 48000 * 5)).float() ref_audio = torch.randn((1, 48000 * 5)).float()
@ -275,10 +287,18 @@ def export(vits_path, gpt_path, project_name):
pass pass
ssl_content = ssl(ref_audio_16k).float() ssl_content = ssl(ref_audio_16k).float()
debug = False
if debug:
a, b = gpt_sovits(ref_seq, text_seq, ref_bert, text_bert, ref_audio_sr, ssl_content, debug=debug)
soundfile.write("out1.wav", a.cpu().detach().numpy(), vits.hps.data.sampling_rate)
soundfile.write("out2.wav", b[0], vits.hps.data.sampling_rate)
return
a = gpt_sovits(ref_seq, text_seq, ref_bert, text_bert, ref_audio_sr, ssl_content).detach().cpu().numpy() a = gpt_sovits(ref_seq, text_seq, ref_bert, text_bert, ref_audio_sr, ssl_content).detach().cpu().numpy()
# soundfile.write("out.wav", a, vits.hps.data.sampling_rate) soundfile.write("out.wav", a, vits.hps.data.sampling_rate)
gpt_sovits.export(ref_seq, text_seq, ref_bert, text_bert, ref_audio_sr, ssl_content, project_name) gpt_sovits.export(ref_seq, text_seq, ref_bert, text_bert, ref_audio_sr, ssl_content, project_name)
@ -306,9 +326,9 @@ if __name__ == "__main__":
except: except:
pass pass
gpt_path = "pt_model/koharu-e20.ckpt" gpt_path = "GPT_weights/nahida-e25.ckpt"
vits_path = "pt_model/koharu_e20_s4960.pth" vits_path = "SoVITS_weights/nahida_e30_s3930.pth"
exp_path = "koharu" exp_path = "nahida"
export(vits_path, gpt_path, exp_path) export(vits_path, gpt_path, exp_path)
# soundfile.write("out.wav", a, vits.hps.data.sampling_rate) # soundfile.write("out.wav", a, vits.hps.data.sampling_rate)