in the pass tensorrt support Implicit batch so although it does not support 5d input but can still init the input in to 4d than pass in the batch size like this builder.max_batch_size = maxBatchSize. so it can do 2d conv in 5d array. but now i have to pass the batch size at the begining which make it a 5d input but i only want 2d conv. IConvolutionLayer only support 4 dimension input so is that mean that theere is no way to switch to tansorrt10.3?
this is the repo i use:
tensorrtx/tsm/tsm_r50.py at master · wang-xinyu/tensorrtx (github.com)