Question

嘿我已经在tensorflow中编写了一些用于预测声音（声音分类）的代码，但是当我构建时，我将此作为没有标签和分数的响应 *标签被假定为数字1-10

{
  {
    string_val: "0"
    string_val: "1"
    string_val: "2"
    string_val: "3"
    string_val: "4"
    string_val: "5"
    string_val: "6"
    string_val: "7"
    string_val: "8"
    string_val: "9"
  }
}
outputs {
  key: "scores"
  value {
    dtype: DT_FLOAT
    tensor_shape {
      dim {
        size: 1
      }
      dim {
        size: 10
      }
    }
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
    float_val: nan
  }
}

这是预测和分类代码

classification_signature = signature_def_utils.build_signature_def(
    inputs={signature_constants.CLASSIFY_INPUTS: classification_inputs},
    outputs={
        signature_constants.CLASSIFY_OUTPUT_CLASSES:
            classification_outputs_classes,
        signature_constants.CLASSIFY_OUTPUT_SCORES:
            classification_outputs_scores
    },
    method_name=signature_constants.CLASSIFY_METHOD_NAME)

tensor_info_x = utils.build_tensor_info(x)
tensor_info_y = utils.build_tensor_info(output)

prediction_signature = signature_def_utils.build_signature_def(
    inputs={'sounds': tensor_info_x},
    outputs={
        'classes' : classification_outputs_classes,
        'scores': classification_outputs_scores,
    },
    method_name=signature_constants.PREDICT_METHOD_NAME)


print('Exporting...')
legacy_init_op = tf.group(tf.initialize_all_tables(), name='legacy_init_op')
builder.add_meta_graph_and_variables(
    sess, [tag_constants.SERVING],
    signature_def_map={
        'predict_sound':
            prediction_signature,
        signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY:
            classification_signature,
    },
    legacy_init_op=legacy_init_op)

builder.save()

Tensorflow不计算预测分数

0 个答案: