Here is the DAG in my case.
```python
== Task 1 (workload key: ["33a10bb123ba1d747c97e5296eee59ed", 256, 1,
256, 1, 16, 16, 256, 1, 1, 256]) ==
compile_engine_const() = 0
placeholder = PLACEHOLDER [256, 1]
T_repeat(ax0, ax1) = placeholder[ax0, floordiv(ax1, 2)]
T_take(ax0) = T_r
What's your compute_dag that causes this problem? If it is a bug, you are
welcome to contribute a patch.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/autoscheduler-cannot-break-sample-init-population-loop/12713/2)
to respond.
You are receiving this because you enabled mailing list
You can follow the scripts in this repo to correctly set the tuning parameters
of Ansor
https://github.com/tlc-pack/TLCBench
Note that Ansor works better for NHWC layout, so probably you can follow the
scripts above to convert your model into NHWC layout.
---
[Visit
Topic](https://discus
There are no existing examples avaiable. But you can try to build one by
translating existing python code to c++ code. The c++ entry point is file
https://github.com/apache/tvm/blob/main/src/auto_scheduler/auto_schedule.cc
---
[Visit
Topic](https://discuss.tvm.apache.org/t/c-apis-of-ansor
100 trials / 2.8 hours is not expected. In addition to Cody's suggestions, you
can also tweak the argument of RPCRunner/LocalRunner to accelerate the
measurement.
---
[Visit
Topic](https://discuss.tvm.apache.org/t/ansor-number-of-tuning-trials/12864/5)
to respond.
You are receiving this
