cxx122 opened a new issue, #12379:
URL: https://github.com/apache/tvm/issues/12379
The sample_initial_population will fall into an infinite loop when given a
specific te program
### Environment
Operating System: Ubuntu 18.04, TVM version: tag0.9.0 [d361585]
### Steps to reproduce
```
import tvm
from tvm import te, auto_scheduler, topi
from tvm.auto_scheduler.workload_registry import register_workload_tensors
from tvm.auto_scheduler.cost_model import XGBModel, RandomModel
def te_test():
reduce_0 = te.reduce_axis((5, 0), name="REDUCE_0")
tensor_0 = te.placeholder([5], dtype="float16", name="TENSOR_0")
tensor_1 = te.compute([5,5], lambda
wce,xcy:te.sum(expr=tensor_0[reduce_0], axis=[reduce_0]), name ="TENSOR_1")
return [tensor_0,tensor_1]
tensors = te_test()
dag = auto_scheduler.ComputeDAG(tensors)
key = register_workload_tensors(dag.workload_key(), tensors)
task = auto_scheduler.SearchTask(workload_key=key,
target=tvm.target.Target("llvm"))
policy = auto_scheduler.SketchPolicy(task, program_cost_model=RandomModel(),
verbose=0)
states = policy.sample_initial_population()[:]
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
