for cuda backend, we have a get_workspace_size function, but for some model we got a big workspace size and out of the range of int, this will cause a issue when use nnfusion jit.
int get_workspace_size()
{
return 2655404160;
}
n_byte = get_workspace_size() # got -1639563136
if not n_byte:
return None
for cuda backend, we have a
get_workspace_size
function, but for some model we got a big workspace size and out of the range ofint
, this will cause a issue when use nnfusion jit.