Lucien0 opened a new issue, #14134:
URL: https://github.com/apache/tvm/issues/14134
I have this Relay Program:
```
import numpy as np
import tvm
from tvm import relay, transform, IRModule
x_shape = (1, relay.Any(), 100, 17, 1)
x = relay.var('x', shape=x_shape)
y = relay.squeeze(x, axis=[-1])
y = relay.sqrt(y)
mod = IRModule.from_expr(y)
with transform.PassContext(opt_level=1):
vm_exec = relay.vm.compile(mod, target='llvm')
vm = tvm.runtime.vm.VirtualMachine(vm_exec, tvm.cpu())
x_input = np.zeros(shape=(1, 100, 100, 17, 1), dtype='float32')
data = vm.run(x=x_input)
```
### Expected behavior
run correctly.
### Actual behavior
raise error:
`Check failed: ret == 0 (-1 vs. 0) : Assert fail: 4 ==
T.tvm_struct_get(arg_T_sqrt, 0, 4, "int32"), arg.T_sqrt.ndim is expected to
equal 4`
### Environment
None
### Steps to reproduce
run this case
I find in squeeze shape function:
https://github.com/apache/tvm/blob/f21a17b67cfd1efb40dbb2271ebca28e7a7c450f/python/tvm/relay/op/_transform.py#L914-L927
Here if the axis is -1, then will get the wrong result.
--
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]
