When running Tiny BERT (from here: https://github.com/google-research/bert) in TVM following this tutorial (https://tvm.apache.org/docs/vta/tutorials/frontend/deploy_classification.html#sphx-glr-vta-tutorials-frontend-deploy-classification-py) I get an error on the quantization step (
relay.quantize.quantize). Specifically, I get the following error:
File "/usr/tvm/src/relay/quantize/realize.cc", line 422 TVMError: Check failed: new_args.size() == 1 (4 vs. 1) :
I debugged and found that this error occurs for the strided_slice operation in the IdentityRealize function, and that the new_args, new_args, new_args all contain meta relay constants. I am thus confused as to why the number of arguments should only be 1 (as is checked here: https://github.com/apache/incubator-tvm/blob/master/src/relay/quantize/realize.cc#L422) while strided_slice seems to take in more than one argument. Any help or insight is appreciated.