mlx
mlx copied to clipboard
[BUG] cumsum fails with int64 array
Describe the bug When applying cumsum to an int64 array, there's a kernel error.
To Reproduce
import mlx.core as mx
a = mx.array([0, 1, 2], dtype=mx.int64)
mx.cumsum(a)
>>> libc++abi: terminating due to uncaught exception of type std::runtime_error: [metal::Device] Unable to load kernel contiguous_scan_inclusive_sum_int64_int64
[1] 48047 abort python
Desktop (please complete the following information):
- macOS 14.2
- branch: main
This is expected in as much as we don't have implementations of these for int64. Are you able to use a 32 bit type as a workaround?
Ok, just wanted to open an issue in case. Yes I can work with int32 here.