import bitsandbytes as bnb | |
import torch | |
p = torch.nn.Parameter(torch.rand(10,10).cuda()) | |
a = torch.rand(10,10).cuda() | |
p1 = p.data.sum().item() | |
adam = bnb.optim.Adam([p]) | |
out = a*p | |
loss = out.sum() | |
loss.backward() | |
adam.step() | |
p2 = p.data.sum().item() | |
assert p1 != p2 | |
print('SUCCESS!') | |
print('Installation was successful!') |