首页 > 解决方案 > Swift for TensorFlow `错误:无法查找符号:TensorFlow.Tensor 的类型元数据访问器`

问题描述

以下几行 -

import TensorFlow
let matrix = Tensor<Float>([[1, 2], [3, 4]])

在 macOS 操场上执行会产生错误 -

error: Couldn't lookup symbols:
  type metadata accessor for TensorFlow.Tensor
  protocol witness table for TensorFlow.Tensor<A> : TensorFlow.TensorArrayProtocol in TensorFlow
  type metadata accessor for TensorFlow.TensorElementLiteral
  protocol witness table for TensorFlow.TensorElementLiteral<A> : TensorFlow.TensorArrayProtocol in TensorFlow
  protocol conformance descriptor for <A where A: TensorFlow.TensorArrayProtocol> Swift.Array<A> : TensorFlow.TensorArrayProtocol in TensorFlow
  protocol witness table for Swift.Float : TensorFlow.TensorFlowScalar in TensorFlow
  _TFE_DeleteOp
  _TFE_NewOp
  _TFE_OpSetAttrInt
  _TFE_OpSetAttrType
  _TF_DeleteStatus
  _TF_NewStatus
  ___tf_tensor_from_scalar
  __swift_tfc_CheckOk
  __swift_tfc_CreateTensorHandleFromC
  __swift_tfc_EagerExecute
  __swift_tfc_GetGlobalEagerContext
  __swift_tfc_OpAddInputFromTensorGroup
  __swift_tfc_OpSetDeviceFromScope

Xcode 10.2,Swift for TensorFlow 版本 0.2,2019-03-02

关于如何解决这个问题的任何想法?

PS 0.3、2019-04-23 工具链也存在该问题。

标签: swiftxcodetensorflowswift-playgroundswift-for-tensorflow

解决方案


好的,这似乎是一个已知问题 - https://github.com/tensorflow/swift/issues/173


推荐阅读