From 8b11d46bc8ccc6d753bb546036cd02fe81646f55 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: Tue, 23 May 2023 00:02:00 +0800
Subject: [PATCH] Add files via upload

---
 vencoder/ContentVec768L9_Onnx.py | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/vencoder/ContentVec768L9_Onnx.py b/vencoder/ContentVec768L9_Onnx.py
index 8b13789..6cc24e0 100644
--- a/vencoder/ContentVec768L9_Onnx.py
+++ b/vencoder/ContentVec768L9_Onnx.py
@@ -1 +1,28 @@
+from vencoder.encoder import SpeechEncoder
+import onnxruntime
+import torch
 
+class ContentVec768L9_Onnx(SpeechEncoder):
+    def __init__(self,vec_path = "pretrain/vec-768-layer-9.onnx",device=None):
+        print("load model(s) from {}".format(vec_path))
+        self.hidden_dim = 768
+        if device is None:
+            self.dev = torch.device("cpu")
+        else:
+            self.dev = torch.device(device)
+        if device == 'cpu' or device == torch.device("cpu") or device is None:
+            providers = ['CPUExecutionProvider']
+        elif device == 'cuda' or device == torch.device("cuda"):
+            providers = ['CUDAExecutionProvider', 'CPUExecutionProvider']
+        self.model = onnxruntime.InferenceSession(vec_path, providers=providers)
+
+    def encoder(self, wav):
+        feats = wav
+        if feats.dim() == 2:  # double channels
+          feats = feats.mean(-1)
+        assert feats.dim() == 1, feats.dim()
+        feats = feats.view(1, -1)
+        feats = feats.unsqueeze(0).detach().numpy()
+        onnx_input = {self.model.get_inputs()[0].name: feats}
+        logits = self.model.run(None, onnx_input)
+        return torch.tensor(logits[0]).transpose(1, 2)
\ No newline at end of file