model with dynamic inputs

[IamExperimenting]

Hi Team,

I have built the model with dynamic inputs, so how do I convert my model to tensorrt?
here it takes input shape

# create example data
x = torch.ones((1, 3, 224, 224)).cuda()

# convert to TensorRT feeding sample data as input
model_trt = torch2trt(model, [x])

however, in my use case it take a dynamic input shape eg:

1. example = torch.rand(1, 3, 224, 224)
2. example = torch.rand(1, 3, 224, 550)
3. example = torch.rand(1, 3, 640, 480)

[Burhan-Q]

Maybe try using (for conversion)

x = torch.ones((1, 3, -1, -1)).cuda()