Was there a version update? (i don't think so but i feel i'm going crazy)
A week ago i was able to convert the Softmax layer:
softmax = nn.Softmax(dim=1)
x = softmax(x)
This is the conversion code:
model_file = open(ONNX, 'rb')
model_proto = onnx_pb.ModelProto()
model_proto.ParseFromString(model_file.read())
coreml_model = convert(model_proto,
image_input_names=['input'],
image_output_names=['output'],
)
but now it throws the error:
Error while converting op of type: Softmax. Error message: Unsupported axis 3 for softmax
Please try converting with higher minimum_ios_deployment_target.
You can also provide custom function/layer to convert the model.
if i add minimum_ios_deployment_target='13' it works, but it did work before that (am i going crazy?)
onnx_coreml = 1.3
onnx = 1.7.0
coremltools =3.4
torch = 1.6.0+cu101
Just realized the importance of pip freeze :( a little too late.
What am i missing?
Thanks in advance.