When I build the onnx model, I performed the following optimizations: with tvm.transform.PassContext(opt_level=3): lib = relay.build_module.build(mod, target=target, params=params)
But after the program is executed to a certain position, there is no output change in the terminal.
[17:01:20] /home/guanyiliao/tvm/src/ir/transform.cc:412: Executing module pass : InferType with opt level: 0 [17:01:20] /home/guanyiliao/tvm/src/relay/transforms/type_infer.cc:818: tvm::relay::transform::InferType [17:01:20] /home/guanyiliao/tvm/src/ir/transform.cc:412: Executing module pass : RemoveUnusedFunctions with opt level: 1 [17:01:20] /home/guanyiliao/tvm/src/ir/transform.cc:412: Executing module pass : ToBasicBlockNormalForm with opt level: 1
But before that, I used some simple models (resnet50) and 224x224 input size is no problem。
Now I tried to build an other complex model. At this time, the input became a 1080x1920 RGB image, and then I couldn’t continue to execute it.So this problem is related to the model and the size of the input?
Is this thread-related? I see that the source code seems to use only single thread. Can someone help me？