Adapt python to work with HIP

bitsandbytes/autograd/_functions.py

@@ -224,7 +224,7 @@ matmul_cublas = MatMul8bit.apply
 
 def supports_igemmlt(device: torch.device) -> bool:
     """check if this device supports the optimized int8 kernel"""
-    if torch.cuda.get_device_capability(device=device) < (7, 5):
+    if torch.cuda.get_device_capability(device=device) < (7, 5) or torch.version.hip:
         return False
     device_name = torch.cuda.get_device_name(device=device)
     nvidia16_models = ('GTX 1630', 'GTX 1650', 'GTX 1660')  # https://en.wikipedia.org/wiki/GeForce_16_series

bitsandbytes/cuda_setup/main.py

@@ -308,10 +308,14 @@ def determine_cuda_runtime_lib_path() -> Union[Path, None]:
 
 # https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART____VERSION.html#group__CUDART____VERSION
 def get_cuda_version():
-    major, minor = map(int, torch.version.cuda.split("."))
+    if torch.version.cuda:
+        major, minor = map(int, torch.version.cuda.split("."))
 
     if major < 11:
         CUDASetup.get_instance().add_log_entry('CUDA SETUP: CUDA version lower than 11 are currently not supported for LLM.int8(). You will be only to use 8-bit optimizers and quantization routines!!')
+    elif torch.version.hip:
+        major, minor = map(int, torch.version.hip.split("."))
+    
 
     return f'{major}{minor}'
 
@@ -332,7 +336,9 @@ def evaluate_cuda_setup():
         cuda_setup.add_log_entry(('Welcome to bitsandbytes. For bug reports, please run\n\npython -m bitsandbytes\n\n'),
               ('and submit this information together with your error trace to: https://github.com/TimDettmers/bitsandbytes/issues'))
         cuda_setup.add_log_entry('='*80)
+
     if not torch.cuda.is_available(): return 'libbitsandbytes_cpu.so', None, None, None
+    if torch.version.hip: return 'libbitsandbytes_hip_nohipblaslt.so', None, None, None
 
     cudart_path = determine_cuda_runtime_lib_path()
     ccs = get_compute_capabilities()