[PYTHON] Added option to show PTX source code in Python

python/examples/tutorials/mat_mul.py

@@ -121,8 +121,8 @@ dot = _dot.apply
 torch.manual_seed(0)
 
 M, N, K = 2048, 2048, 2048
-a = torch.rand((M, K)).cuda()
-b = torch.rand((K, N)).cuda()
+a = torch.rand((M, K)).cuda().half()
+b = torch.rand((K, N)).cuda().half()
 
 #a[:] = 1
 #b[:] = 1

python/examples/tutorials/vec_add.py

@@ -23,12 +23,9 @@ __global__ void add(float* z, float* x, float* y, int N) {
     @staticmethod
     def forward(ctx, x, y):
        z = torch.empty_like(x).cuda()
-
        N = x.numel()
        grid = lambda opt: (triton.cdiv(N, opt.d('TILE')),)
-
        _add.kernel(z,x,y, N, grid=grid)
-
        return z
 
 add = _add.apply