Cnly commented on issue #9955:
URL: https://github.com/apache/tvm/issues/9955#issuecomment-1015518259
I guess this only affects operators implemented with `@hybrid.script`? A
quick grep shows not much usage in tvm. To see if I'm using the scatter
operator wrong and to confirm my guess, I tried with `relay.unique` and saw the
same exception. Reproduction snippet:
```py
import numpy as np
import tvm
import tvm.relay as relay
idxes = np.array([[0, 1],
[2, 3],
[4, 5]])
x = relay.zeros((3, 10), 'int32')
y = relay.var("y", shape=idxes.shape, dtype='int32')
z = relay.unique(relay.sum(x, axis=0))[0]
f = relay.Function([], z)
m = tvm.IRModule.from_expr(f)
# Either line triggers the error:
relay.build(m, target='llvm')
relay.create_executor('graph', mod=m).evaluate()()
```
The line pointed to by traceback:
https://github.com/apache/tvm/blob/31de5bc40f22ba761361464921b9b73af51bb214/python/tvm/topi/unique.py#L273
--
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]
