使用 contrib 操作的 Tensorflow 服务

Question

如果有 tf.contrib 个操作，我如何使用 tensorflow-serving 服务模型。我通过 Docker（最新）（tf 1.11 版本）使用 Tensorflow Serving，当我为模型提供服务时，会出现下一条消息：

“Failed to start server. Error: Unknown: 1 servable(s) did not become available: {{{name: slider_universal version: 1} due to error: Not found: Op type not registered ‘ImageProjectiveTransformV2’ in binary running on 254345a5d9f1. Make sure the Op and Kernel are registered in the binary running in this process. Note that if you are loading a saved graph which used ops from tf.contrib, accessing (e.g.) tf.contrib.resampler should be done before importing the graph, as contrib ops are lazily registered when the module is first accessed.}, }”

我也是用bazel构建的，但是出现了同样的错误

我用tf.contrib.image.transform

如果我在导出模型时删除这个操作，它可以由 tensorflow serving 提供服务

Answer 1

我在使用相同的操作时遇到了问题，看来唯一的方法就是构建它。

tensorflow_serving/model_servers/BUILD 定义了哪些 Tensorflow ops 将包含在变量 SUPPORTED_TENSORFLOW_OPS 中，我对此有点困惑，因为它指定应包含 contrib ops。但是，由于 tensorflow contrib 构建规则似乎不包含 contrib.image 下的操作，因此我通过将此变量更新为以下

来明确添加这些操作

SUPPORTED_TENSORFLOW_OPS = [
    "@org_tensorflow//tensorflow/contrib:contrib_kernels",
    "@org_tensorflow//tensorflow/contrib:contrib_ops_op_lib",
    "@org_tensorflow//tensorflow/contrib/image:image_ops_kernels",
    "@org_tensorflow//tensorflow/contrib/image:image_ops_op_lib",
]

使用 contrib 操作的 Tensorflow 服务

Tensorflow serving with contrib operations

python

tensorflow

tensorflow-serving