Skip to content

Latest commit

 

History

History
19 lines (16 loc) · 1.17 KB

README.md

File metadata and controls

19 lines (16 loc) · 1.17 KB
Raw

To reproduce a fail of compilation with tune information

  1. Download bertsquad10.onnx from ONNX model zoo: https://github.com/onnx/models/tree/master/text/machine_comprehension/bert-squad/model
  2. Downnlaad TVM, build with llvm support, set up environment variables to find tvm python module
  3. In separate console run rpc tracker: python -m tvm.exec.rpc_tracker --host 127.0.0.1 --port 9190
  4. In separate console run rpc server: python -m tvm.exec.rpc_server --tracker 127.0.0.1:9190 --port 9090 --key localcpu --no-fork
  5. Execute search_bert_local.py: python search_bert_local.py
  6. After finishing of statistic collection in intem 5, execute compile_bert_tuned_local.py: python compile_bert_tuned_local.py

There will be a fail: Check failed: false == false: [22:31:18] src/relay/op/nn/nn.h:73:

An internal invariant was violated during the execution of TVM. Please read TVM's error reporting guidelines. More details can be found here: https://discuss.tvm.ai/t/error-reporting/7793.

Check failed: static_cast(weight->shape.size()) == 2 == false: