}
    }
  }
}
node {
  name: "MobilenetV1/Conv2d_0/weights/read"
  op: "Identity"
  input: "MobilenetV1/Conv2d_0/weights"
  attr {
    key: "T"
    value {
      type: DT_FLOAT
    }
  }
  attr {
    key: "_class"
    value {
      list {
        s: "loc:@MobilenetV1/Conv2d_0/weights"
      }
    }
  }
}
node {
  name: "MobilenetV1/MobilenetV1/Conv2d_0/Conv2D"

  const bool disable_per_channel_;

  // We should distinguish weights and bias constants. Biases are specified by
  // the quantization spec or are the operands of ops with same scale spec. The
  // rest are weights.
  DenseSet<Operation*> weights_;

  // The weights require narrow_range quantization. This map collects all the
  // weight operands defined by the op quant spec. The value of each entry is

namespace {

using ::stablehlo::quantization::Method;
using ::stablehlo::quantization::QuantizedType;
using ::stablehlo::quantization::WeightOnlyPtq;

// Inserts quantization parameters of weights for weight-only quantization and
// dynamic range quantization of `stablehlo.convolution` and
// `stablehlo.dot_general`.
class InsertWeightParamPass
    : public impl::InsertWeightParamPassBase<InsertWeightParamPass> {
 public:

      }
    }
  }
}
node {
  name: "MobilenetV1/Conv2d_0/weights/read"
  op: "Identity"
  input: "MobilenetV1/Conv2d_0/weights"
  attr {
    key: "T"
    value {
      type: DT_FLOAT
    }
  }
  attr {
    key: "_class"
    value {
      list {
        s: "loc:@MobilenetV1/Conv2d_0/weights"
      }
    }
  }
}
node {
  name: "MobilenetV1/MobilenetV1/Conv2d_0/Conv2D"

    // This is the argument used to refer to the pass in
    // the textual format (on the commandline for example).
    return "quant-quantize-weights";
  }

  StringRef getDescription() const final {
    // This is a brief description of the pass.
    return "Quantize weights used by quantizable ops.";
  }

  void getDependentDialects(DialectRegistry& registry) const override {

  // For now, restrict scale adjustment to ops with affine quantized weights,
  // and having weights and biases as constants. This currently only applies to
  // FC and Conv* ops. Restriction for the weight can be relaxed if there are
  // needs for adjusting scale of variable weights.
  auto affine_op = dyn_cast<AffineQuantizedOpInterface>(op);

				weights := make([]int, 0)
				for _, localityEndpoint := range cluster.LoadAssignment.Endpoints {
					weights = append(weights, int(localityEndpoint.LoadBalancingWeight.GetValue()))
				}
				if !reflect.DeepEqual(weights, tt.expected) {
					t.Errorf("Got weights %v expected %v", weights, tt.expected)
				}
			})
		}
	})

  // Whether to allow weight-only quantization. This scheme quantizes
  // weights but will dequantize them back at runtime which is useful for
  // memory bound case without kernel support available in lower precisions.
  // Used in MLIR dynamic range quantizer.
  bool weight_only_quantization = false;

  // The minimum number of elements in a weights array required to apply

  input_ = fused_func_op_.getArgument(0);
  bias_ = fused_func_op_.getArgument(2);

  weight_ = fused_func_op_.getArgument(1);
  weight_type_ = mlir::cast<RankedTensorType>(weight_.getType());

  if (weight_type_.getRank() != 2) {
    return fused_func_op_.emitError() << "The weight tensor was not of rank 2";
  }

  if (weight_type_.getDimSize(1) % num_gates_ != 0) {

                         METHOD_STATIC_RANGE_WEIGHT_ONLY_INT8,
                     "weight_only", "Post-training weight-only quantizaiton"))};

  Option<bool> enable_per_channel_quantization_{
      *this, "enable-per-channel-quantization", llvm::cl::init(false),
      llvm::cl::desc("Whether enable per-channel quantized weights.")};
};

// Apply constant transformations for the op_set.