lazycal opened a new issue #8505:
URL: https://github.com/apache/tvm/issues/8505
Minimum code to reproduce:
```python
import tvm
from tvm import relay
x = relay.var('x', shape=(32, 128), dtype='float16')
y = relay.var('y', shape=(20, 128), dtype='float16')
x1 = relay.nn.dense(x, y, units=20)
func = relay.Function([x,y], x1)
mod = tvm.IRModule.from_expr(func)
with tvm.transform.PassContext(opt_level=3):
with tvm.target.Target('cuda'):
lib = relay.optimize(mod)
```
A quick fix would be adding one line of code `new_attrs['units'] = None`
before
[here](https://github.com/apache/tvm/blob/main/python/tvm/topi/cuda/tensorcore_alter_op.py#L170)
(but that may not be desired since it makes a originally non-None `units`
None).
--
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]
