Fixed issue where the CUDA SETUP was not printed.

bitsandbytes/cextension.py

 import ctypes as ct
+import os
 import torch
 
 from pathlib import Path
@@ -10,6 +11,8 @@ from bitsandbytes.cuda_setup.main import CUDASetup
 setup = CUDASetup.get_instance()
 if setup.initialized != True:
     setup.run_cuda_setup()
+    if 'BITSANDBYTES_NOWELCOME' not in os.environ or str(os.environ['BITSANDBYTES_NOWELCOME']) == '0':
+        setup.print_log_stack()
 
 lib = setup.lib
 try:

check_bnb_install.py

+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!')

setup.py

@@ -18,7 +18,7 @@ def read(fname):
 
 setup(
     name=f"bitsandbytes",
-    version=f"0.36.0",
+    version=f"0.36.0-2",
     author="Tim Dettmers",
     author_email="dettmers@cs.washington.edu",
     description="8-bit optimizers and matrix multiplication routines.",