change typecast behavior

bitsandbytes/autograd/_functions.py

@@ -321,7 +321,6 @@ class MatMul8bitLt(torch.autograd.Function):
 
         # 4. Mixed-precision decomposition matmul
         if coo_tensorA is not None and subA is not None:
-            assert subA.dtype == state.subB.dtype == output.dtype, (subA.dtype, state.subB.dtype, output.dtype)
             output.addmm_(subA, state.subB)
 
         # 5. Save state
@@ -330,6 +329,7 @@ class MatMul8bitLt(torch.autograd.Function):
         ctx.formatB = formatB
         ctx.grad_shape = input_shape
         ctx.req_grads = [requires_gradA, requires_gradB, requires_gradBias]
+        ctx.dtype_A, ctx.dtype_B, ctx.dtype_bias = A.dtype, B.dtype, None if bias is None else bias.dtype
 
         if requires_gradA or requires_gradB:
             ctx.tensors = (CAt, subA)
@@ -348,7 +348,7 @@ class MatMul8bitLt(torch.autograd.Function):
         if ctx.is_empty:
             bias_grad = (None if ctx.bias is None else torch.zeros_like(ctx.bias))
             return torch.zeros_like(ctx.A), torch.zeros_like(ctx.B), None, bias_grad, None
-        req_gradA, req_gradB, req_gradBias = ctx.req_grads
+        req_gradA, req_gradB, _, req_gradBias, _ = ctx.needs_input_grad
         CAt, subA = ctx.tensors
         SCAt, idx = ctx.tensor_states
         formatB = ctx.formatB
@@ -357,25 +357,22 @@ class MatMul8bitLt(torch.autograd.Function):
 
         if req_gradBias:
             # compute grad_bias first before changing grad_output dtype
-            grad_bias = grad_output.sum(0)
+            grad_bias = grad_output.sum(0).to(ctx.bias_dtype)
 
         # Cast grad_output to fp16
-        grad_output_dtype = grad_output.dtype
-        grad_output = grad_output.to(torch.float16)
-
         if len(grad_output.shape) == 3:
             grad_output = grad_output.reshape(
                 -1, grad_output.shape[-1]
             ).contiguous()
 
-        Cgrad, Cgradt, SCgrad, SCgradt, coo_tensor = F.double_quant(grad_output)
+        Cgrad, Cgradt, SCgrad, SCgradt, coo_tensor = F.double_quant(grad_output.to(torch.float16))
         if req_gradB:
             CxAt, SAt = F.transform(CAt, formatB, transpose=True)
             C32grad, Sgrad = F.transform(Cgradt, "col32", transpose=True)
             gradB32, SgradB32 = F.igemmlt(C32grad, CxAt, Sgrad, SAt)
-            grad_B = F.mm_dequant(gradB32, SgradB32, SCgradt, SCAt)
+            grad_B = F.mm_dequant(gradB32, SgradB32, SCgradt, SCAt).to(ctx.B_dtype)
             if state.threshold > 0.0 and subA is not None:
-                grad_B[:, idx] += torch.matmul(grad_output.t(), subA)
+                grad_B[:, idx].addmm_(grad_output.t(), subA)
 
         if req_gradA:
             if state.CBt is not None:
@@ -385,18 +382,16 @@ class MatMul8bitLt(torch.autograd.Function):
                         state.CBt, to_order=formatB, transpose=True
                     )
                 gradA32, SgradA32 = F.igemmlt(C32grad, state.CxBt, Sgrad, state.SBt)
-                grad_A = F.mm_dequant(gradA32, SgradA32, SCgrad, state.SCBt).view(ctx.grad_shape)
+                grad_A = F.mm_dequant(gradA32, SgradA32, SCgrad, state.SCBt).view(ctx.grad_shape).to(ctx.A_dtype)
+
             elif state.CB is not None:
-                CB = state.CB.half()
+                CB = state.CB.to(ctx.B_dtype)
                 SCB = (state.SCB.unsqueeze(1) / 127.0).half()
                 CB *= SCB
-                grad_A = torch.mm(grad_output, CB).view(ctx.grad_shape)
+                grad_A = torch.mm(grad_output, CB).view(ctx.grad_shape).to(ctx.A_dtype)
             else:
                 raise Exception('State must contain either CBt or CB matrix for backward')
 
-        # Cast grad_A back to grad_output_dtype
-        grad_output = grad_output.to(grad_output_dtype)
-
         return grad_A, grad_B, None, grad_bias, None