From 0eb6bb67be84fefbad6907010eb0b86ebc8167e3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=CE=9D=CE=B1=CF=81=CE=BF=CF=85=CF=83=CE=AD=C2=B7=CE=BC?=
 =?UTF-8?q?=C2=B7=CE=B3=CE=B9=CE=BF=CF=85=CE=BC=CE=B5=CE=BC=CE=AF=C2=B7?=
 =?UTF-8?q?=CE=A7=CE=B9=CE=BD=CE=B1=CE=BA=CE=AC=CE=BD=CE=BD=CE=B1?=
 <40709280+NaruseMioShirakana@users.noreply.github.com>
Date: Sat, 17 Jun 2023 22:49:16 +0800
Subject: [PATCH] =?UTF-8?q?Onnx=E6=8E=A8=E7=90=86dml=E6=94=AF=E6=8C=81=20(?=
 =?UTF-8?q?#556)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

* Add files via upload

* Add files via upload
---
 infer-web.py                 | 13 ++++++-------
 infer_pack/onnx_inference.py |  5 ++++-
 2 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/infer-web.py b/infer-web.py
index a3bac6d..e96fbb0 100644
--- a/infer-web.py
+++ b/infer-web.py
@@ -1119,12 +1119,12 @@ def change_info_(ckpt_path):
 from infer_pack.models_onnx import SynthesizerTrnMsNSFsidM
 
 
-def export_onnx(ModelPath, ExportedPath, MoeVS=True):
+def export_onnx(ModelPath, ExportedPath):
     cpt = torch.load(ModelPath, map_location="cpu")
-    cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]  # n_spk
-    hidden_channels = 256 if cpt.get("version","v1")=="v1"else 768#cpt["config"][-2]  # hidden_channels，为768Vec做准备
+    cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
+    vec_channels = 256 if cpt.get("version","v1")=="v1"else 768
 
-    test_phone = torch.rand(1, 200, hidden_channels)  # hidden unit
+    test_phone = torch.rand(1, 200, vec_channels)  # hidden unit
     test_phone_lengths = torch.tensor([200]).long()  # hidden unit 长度（貌似没啥用）
     test_pitch = torch.randint(size=(1, 200), low=5, high=255)  # 基频（单位赫兹）
     test_pitchf = torch.rand(1, 200)  # nsf基频
@@ -1160,7 +1160,7 @@ def export_onnx(ModelPath, ExportedPath, MoeVS=True):
             "rnd": [2],
         },
         do_constant_folding=False,
-        opset_version=16,
+        opset_version=13,
         verbose=False,
         input_names=input_names,
         output_names=output_names,
@@ -1835,11 +1835,10 @@ with gr.Blocks() as app:
                     label=i18n("Onnx输出路径"), value="", interactive=True
                 )
             with gr.Row():
-                moevs = gr.Checkbox(label=i18n("MoeVS模型"), value=False,visible=False)
                 infoOnnx = gr.Label(label="info")
             with gr.Row():
                 butOnnx = gr.Button(i18n("导出Onnx模型"), variant="primary")
-            butOnnx.click(export_onnx, [ckpt_dir, onnx_dir, moevs], infoOnnx)
+            butOnnx.click(export_onnx, [ckpt_dir, onnx_dir], infoOnnx)
 
         tab_faq = i18n("常见问题解答")
         with gr.TabItem(tab_faq):
diff --git a/infer_pack/onnx_inference.py b/infer_pack/onnx_inference.py
index 09a4ed2..7502543 100644
--- a/infer_pack/onnx_inference.py
+++ b/infer_pack/onnx_inference.py
@@ -3,7 +3,6 @@ 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))
@@ -11,6 +10,8 @@ class ContentVec:
             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)
@@ -68,6 +69,8 @@ class OnnxRVC:
             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)