Error when using fast_jl

[xianglinyang]

I am using
torch 2.4.0+cu124
fast_jl 0.1.3
traker 0.1.3

But keep getting this error

line 12, in
fast_jl.project_rademacher_8(torch.zeros(8, 1_000, device=device), 512, 0, num_sms)
RuntimeError: CUDA error: the provided PTX was compiled with an unsupported toolchain.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Could you pass the complete requirements (hopefully using the newest pytorch) for installing less? Thanks

[gavinwxy]

Hi, have you solve this problem?