tensorflow服务REST API错误：找不到可服务模型的基本路径/ models / model

Question

我是tensorflow服务和docker的新手，我经历了多个教程，并希望在经过重新训练的模型上进行尝试。

当我运行以下代码时，我得到以下错误：找不到基本路径，据我了解，目标路径不正确，泊坞窗找不到它。如何定义目标路径？非常感谢您的指导！

!docker run -p 8503:8501 \
    --name=ea4 \
    --mount type=bind,source=/home/jupyter/../saved_models/,target=/models/ea4 \
    -e MODEL_NAME=ea4 \
    -t tensorflow/serving

错误：

2020-05-28 01:55:31.719025: I tensorflow_serving/model_servers/server.cc:86] Building single TensorFlow model file config:  model_name: ea4 model_base_path: /models/ea4
2020-05-28 01:55:31.719326: I tensorflow_serving/model_servers/server_core.cc:462] Adding/updating models.
2020-05-28 01:55:31.719363: I tensorflow_serving/model_servers/server_core.cc:573]  (Re-)adding model: ea4
2020-05-28 01:55:31.719784: E tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:362] FileSystemStoragePathSource encountered a filesystem access error: Could not find base path /models/ea4 for servable ea4
2020-05-28 01:55:32.719680: E tensorflow_serving/sources/storage_path/file_system_storage_path_source.cc:362] FileSystemStoragePathSource encountered a filesystem access error: Could not find base path /models/ea4 for servable ea4

Answer 1

在按如下所示更改代码后，该端口现在可在Docker映像中使用。

!docker run -p 8501:8501 \
    --name=ea \
    -v "/home/../saved_models/:/models/ea/1" \
    -e MODEL_NAME=ea \
    -t tensorflow/serving