(tg_out, tg_data)
| 9 | |
| 10 | @TinyJit |
| 11 | def f(tg_out, tg_data): return tg_out.assign(tg_data[:, :, 0] * 0.2989 + tg_data[:, :, 1] * 0.5870 + tg_data[:, :, 2] * 0.1140).realize() |
| 12 | |
| 13 | def custom_kernel(data: torch.Tensor, device="CUDA") -> torch.Tensor: |
| 14 | assert data.dtype == torch.float32 |
no test coverage detected