Browse Source
Update test
Signed-off-by: Jael Gu <mengjia.gu@zilliz.com>
main
1 changed files with
2 additions and
1 deletions
-
test.py
|
|
@ -7,6 +7,7 @@ import numpy |
|
|
import onnx |
|
|
import onnx |
|
|
import onnxruntime |
|
|
import onnxruntime |
|
|
|
|
|
|
|
|
|
|
|
device = 'cuda' if torch.cuda.is_available() else 'cpu' |
|
|
|
|
|
|
|
|
def to_numpy(tensor): |
|
|
def to_numpy(tensor): |
|
|
return tensor.detach().cpu().numpy() if tensor.requires_grad else tensor.cpu().numpy() |
|
|
return tensor.detach().cpu().numpy() if tensor.requires_grad else tensor.cpu().numpy() |
|
|
@ -14,7 +15,7 @@ def to_numpy(tensor): |
|
|
|
|
|
|
|
|
# decode = ops.audio_decode.ffmpeg() |
|
|
# decode = ops.audio_decode.ffmpeg() |
|
|
# audio = [x[0] for x in decode('path/to/audio.wav')] |
|
|
# audio = [x[0] for x in decode('path/to/audio.wav')] |
|
|
audio = torch.rand(10, 256, 32) |
|
|
|
|
|
|
|
|
audio = torch.rand(10, 256, 32).to(device) |
|
|
|
|
|
|
|
|
op = ops.audio_embedding.nnfp() |
|
|
op = ops.audio_embedding.nnfp() |
|
|
out0 = op.get_op().model(audio) |
|
|
out0 = op.get_op().model(audio) |
|
|
|
