Improve torch.compile support for int8 with torch>=2.8 nightly

Author: matthewdouglas

bitsandbytes/autograd/_functions.py

@@ -236,7 +236,8 @@ def forward(
         ctx.state = state
 
         ctx.grad_shape = input_shape
-        ctx.dtype_A, ctx.dtype_B, ctx.dtype_bias = A.dtype, B.dtype, None if bias is None else bias.dtype
+        ctx.dtype_A = A.dtype
+        ctx.dtype_bias = None if bias is None else bias.dtype
 
         if any(ctx.needs_input_grad[:2]):
             ctx.tensors = (CAt, subA, A)