In correct torch.baddbmm when beta=0

[roastduck]

Quoated from the PyTorch doc:

If beta is 0, then input will be ignored, and nan and inf in it will not be propagated.

But torch_musa propagates nan.

>>> import torch              
>>> import torch_musa         
>>> dev = torch.device('musa')
>>> inp = torch.full((8, 8, 8), float('nan'), dtype=torch.float16).to(dev)
>>> batch1 = torch.zeros((8, 8, 8), dtype=torch.float16).to(dev)          
>>> batch2 = torch.zeros((8, 8, 8), dtype=torch.float16).to(dev)          
>>> torch.baddbmm(inp, batch1, batch2, beta=0)                            
tensor([[[nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan],                        
         [nan, nan, nan, nan, nan, nan, nan, nan]],                       

(more outputs are omitted)

Tested with torch_musa 1.1.0+fb1871f.

[hanhaowen-mt]

thank you! we are fixing it!