You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Note that it's possible you'll run into similar issues with other array operations, as we haven't been careful when using 32-bit integer indices, and huge arrays like the one you allocated here may trip up some of those kernels (#1968).
Describe the bug$5 \cdot 10^4$ ) with
Sampling large matrices (starting at dimension
Float16
entries results in the following error:To reproduce
The Minimal Working Example (MWE) for this bug:
Manifest.toml
Expected behavior
Matrix available on the GPU
Version info
Details on Julia:
Details on CUDA:
Additional context
Works correctly for
Float32
.Tested this on the instance mentioned in this report and others with RTX3090 and RTX4090. All have the same issue.
The text was updated successfully, but these errors were encountered: