%conv1 = "tfl.conv_2d"(%1, %2, %cst) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 2 : i32, stride_w = 2 : i32} : (tensor<1x224x224x3xf32>, tensor<32x3x3x3xf32>, tensor<32xf32>) -> tensor<1x112x112x32xf32>

// CHECK: }

// CHECK-LABEL: private @composite_conv_with_bias_dynamic_fn_1
// CHECK: %[[CONV:.*]] = stablehlo.convolution(%arg0, %arg1)
// CHECK: %[[SHAPE_OF:.*]] = shape.shape_of %[[CONV]]
// CHECK: %[[DYNAMIC_BROADCAST_IN_DIM:.*]] = stablehlo.dynamic_broadcast_in_dim %arg2, %[[SHAPE_OF]]
// CHECK: %[[ADD:.*]] = stablehlo.add %[[CONV]], %[[DYNAMIC_BROADCAST_IN_DIM]]
// CHECK: return %[[ADD]] : tensor<?x28x28x16xf32>
// CHECK: }

        # One pure conv
        out = nn_ops.conv2d(
            out,
            self.conv_filters,
            strides=(1, 1, 2, 1),
            dilations=(1, 1, 1, 1),
            padding='SAME',
            data_format='NHWC',
        )

        # One fakequant attached conv
        if is_qat_model:
          out = array_ops.fake_quant_with_min_max_args(

			conv.getTypeIDs[n.Go[:len(n.Go)-9]] = true
		}
	}
	for i, n := range names {
		if types[i] == nil {
			continue
		}
		pos := f.NamePos[n]
		f, fok := types[i].(*dwarf.FuncType)
		if n.Kind != "type" && fok {
			n.Kind = "func"
			n.FuncType = conv.FuncType(f, pos)
		} else {
			n.Type = conv.Type(types[i], pos)
			switch n.Kind {
			case "iconst":

                conf
                conf2
                conf3
                conf4
            }
            dependencies {
                conf 'org:a:1.0', 'org:b:1.0'
                conf2 'org:a:1.0', 'org:c:1.0'
                conf3 'org:b:1.0', 'org:c:1.0'
                conf4 'org:b:1.0', 'org:c:1.0', 'org:d:1.0'
            }
            task checkDeps {

      auto pad_output_type = UnrankedTensorType::get(elem_type);
      input = rewriter.create<TF::PadOp>(op->getLoc(), pad_output_type, input,
                                         padding_const);

      // Set Conv padding to `VALID` since padding has been handled by Pad op.
      state.padding = rewriter.getStringAttr("VALID");
    }
    auto conv_op = static_cast<const ConcreteType *>(this)->createTFLOp(

module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {

func.func @conv(tensor<256x32x32x3xf32>, tensor<3x3x3x16xf32>, tensor<256x3x32x32xf32>) -> (tensor<256x8x7x16xf32>, tensor<256x16x32x32xf32>, tensor<256x8x6x16xf32>, tensor<256x32x32x16xf32>, tensor<256x32x32x16xf32>) {

    ConstBoolAttrTrue, $asymmetric_quantize_inputs),
  [(HasRank<2> $input),
   (AreLastTwoDimsTransposed $perm_value),
   (IsBoolAttrEqual<"false"> $adj_y)]>;

// Replace conv-->transpose-->add with conv-->add-->transpose
// The bias needs only reshape (i.e. ReshapeNCHWBiasToNHWC) and not transpose
// because the bias's shape simply changes from NxCx1x1 to Nx1x1xC.
def ReorderNCHWTransposeAdd : Pat <

          {"BN",
           {"FusedBatchNorm", "FusedBatchNormV2", "FusedBatchNormV3",
            "_FusedBatchNormEx", "FusedBatchNormGrad", "FusedBatchNormGradV2",
            "FusedBatchNormGradV3"}},
          {"Conv", {"_FusedConv2D"}},
          {"SORT", {"TopKV2"}},  // XLA version much faster then TF version.
          {"MISC",
           // clang-format off
     {"ApproxTopK", "BroadcastTo", "ExpandDims", "Fill", "NoOp",

    // Replace filter uses with uniform quantized filter.
    rewriter.replaceAllUsesWith(filter_op->getResult(0),
                                quantized_filter_constant_op.getResult());

    // Replace conv op with a new convolution op that has quantized output type.
    // Quantize -> Dequantize following r3.
    auto output_uniform_quantize_call_op = cast<func::CallOp>(
        *combined_scale_multiply_op.getResult().user_begin());