Spaces:
Running
Running
File size: 1,176 Bytes
f82071f 537486f f82071f 537486f f82071f 537486f f82071f 537486f f82071f 537486f f82071f 537486f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 |
import onnxruntime
import torch
from vencoder.encoder import SpeechEncoder
class HubertSoft_Onnx(SpeechEncoder):
def __init__(self, vec_path="pretrain/hubert-soft.onnx", device=None):
super().__init__()
print("load model(s) from {}".format(vec_path))
self.hidden_dim = 256
if device is None:
self.dev = torch.device("cpu")
else:
self.dev = torch.device(device)
if device == 'cuda' or device == torch.device("cuda"):
providers = ['CUDAExecutionProvider', 'CPUExecutionProvider']
else:
providers = ['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).cpu().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).to(self.dev)
|