input_shape=(1, 3, 4, 3), filter_shape=(2, 3, 3, 2)
    )

    signatures = {
        'serving_default': model.conv.get_concrete_function(),
    }
    save_opts = save_options.SaveOptions(
        function_aliases={'conv_func': model.conv}
    )

    saved_model_save.save(
        model, self._input_saved_model_path, signatures, save_opts
    )

// CHECK: %[[CONV:.+]] = stablehlo.convolution(%[[ARG0]], %[[DQ]])
// CHECK{LITERAL}: dim_numbers = [b, 0, 1, f]x[o, 0, 1, i]->[b, 0, 1, f], window = {pad = [[1, 1], [1, 1]]} {batch_group_count = 1 : i64, feature_group_count = 1 : i64}
// CHECK-SAME: (tensor<1x3x3x4xf32>, tensor<2x3x3x4xf32>) -> tensor<1x3x3x2xf32>
// CHECK: return %[[CONV]]

// -----

  if (elements_depth == 1) {
    return true;
  }

  // In TFLite Conv2D uses OHWI format for filter, and 1HWO for Depthwise Conv.
  // For conv:
  // Check if last dimension in filter equals the first dimension
  // For depthwise conv:
  // Check if the first in filter dimension equals the first dimension.
  if (filter_shape.empty() ||
      (is_depthwise ? filter_shape.back() != elements_depth

            (i != out_batch_dim && out_type.isDynamicDim(i))) {
          return false;
        }
      }
    }

    // All ones in "lhs_dilation" means this "mhlo.conv" op should be
    // converted to "tf.Conv2D" or "tf.DepthwiseConv2dNativeOp".
    auto lhs_dilation = conv_op.getLhsDilation().value();
    if (!lhs_dilation.isSplat() || lhs_dilation.getSplatValue<int64_t>() != 1)

","statusText","finalText","crossDomain","host","hasContent","ifModified","headers","beforeSend","success","send","nativeStatusText","responses","isSuccess","response","modified","ct","finalDataType","firstDataType","ajaxHandleResponses","conv2","current","conv","dataFilter","throws","ajaxConvert","getJSON","getScript","text script","wrapAll","firstElementChild","wrapInner","htmlIsFunction","unwrap","visible","xhr","XMLHttpRequest","xhrSuccessStatus","0","1223","xhrSupported","cors","errorCallba...

","statusText","finalText","crossDomain","host","hasContent","ifModified","headers","beforeSend","success","send","nativeStatusText","responses","isSuccess","response","modified","ct","finalDataType","firstDataType","ajaxHandleResponses","conv2","current","conv","dataFilter","throws","ajaxConvert","getJSON","getScript","text script","wrapAll","firstElementChild","wrapInner","htmlIsFunction","unwrap","visible","xhr","XMLHttpRequest","xhrSuccessStatus","0","1223","xhrSupported","cors","errorCallba...

    // CHECK: %[[SHAPE:.*]] = "tf.Shape"
    // CHECK: %[[CONV:.*]] = "tf.Conv2DBackpropInput"(%[[SHAPE]]
    // CHECK-SAME: (tensor<4xi32>, tensor<1x1x1x1xf32>, tensor<1x1x1x1xf32>) -> tensor<1x1x1x1xf32>
    // CHECK: return %[[CONV]] : tensor<1x1x1x1xf32>
    %0 = "tf.Shape"(%arg0) : (tensor<1x1x1x1xi32>) -> tensor<4xi32>
    %1 = "tf.Conv2DBackpropInput"(%0, %arg1, %arg1) {

      }
      func @_func(%input: tensor<2x112x112x12xf32>, %filter: tensor<7x7x3x64xf32>) {
        %filter_transform = "tf.Pad/tf.Transpose/tf.Reshape"(%filter): tensor<7x7x3x64xf32>) -> tensor<4x4x12x64xf32>
        %conv = "tf.Conv2D"(%input, %filter_transfrom) {strides = [1, 1, 1, 1]}: (tensor<2x112x112x12xf32>, tensor<4x4x12x64xf32>) -> tensor<2x112x112x64xf32>
      }
    }
    ```

897,"touches":3,"min_t":1228948796,"max_t":1236205616,"mean_t":1231544399},{"name":"animate-elem-46-t-expected.png","kids":[],"cl_weight":0.0005553709994722575,"touches":2,"min_t":1229478785,"max_t":1236205616,"mean_t":1232842200},{"name":"filters-conv-01-f-expected.png","kids":[],"cl_weight":0.0015634355156012897,"touches":3,"min_t":1228948796,"max_t":1236205616,"mean_t":1231544399},{"name":"pservers-grad-03-b-expected.png","kids":[],"cl_weight":0.0015634355156012897,"touches":3,"min_t":1228948...