[CI][Fix] Remove additional qnn.op.transpose_conv2d from docs

Fixes CI after #10077, and replaces misuse elsewhere.

Change-Id: I095fc8ea2b8d268b09538832cba1f5482a73a9d9

Author: lhutton1

docs/reference/langref/relay_op.rst

@@ -245,4 +245,3 @@ This level supports dialect operators.
    tvm.relay.qnn.op.simulated_dequantize
    tvm.relay.qnn.op.simulated_quantize
    tvm.relay.qnn.op.subtract
-   tvm.relay.qnn.op.transpose_conv2d

python/tvm/relay/op/strategy/generic.py

@@ -482,7 +482,7 @@ def conv2d_transpose_strategy(attrs, inputs, out_type, target):
             wrap_topi_schedule(topi.generic.schedule_conv2d_transpose_nchw),
             name="conv2d_transpose_nchw.generic",
         )
-    else:  # group_transpose_conv2d
+    else:  # group_conv2d_transpose
         strategy.add_implementation(
             wrap_compute_conv2d_transpose(topi.nn.group_conv2d_transpose_nchw, has_groups=True),
             wrap_topi_schedule(topi.generic.schedule_group_conv2d_transpose_nchw),

python/tvm/relay/qnn/op/qnn.py

@@ -527,8 +527,8 @@ def conv2d_transpose(
     kernel_scale: tvm.relay.Expr
            The scale for the weight tensor. The scale for the weight tensor is
            stored for access to this during relay. This information is not
-           needed in the pass pipeline after qnn.transpose_conv2d is lowered to the
-           sequence of steps as in nn.transpose_conv2d. See also input_scale in Requantize.
+           needed in the pass pipeline after qnn.conv2d_transpose is lowered to the
+           sequence of steps as in nn.conv2d_transpose. See also input_scale in Requantize.
 
     strides : Tuple[int], optional
         The strides of convolution.