Closed Johnson-yue closed 6 years ago
your docker image is everything OK.
It seems to be the problem of input json. Make sure you use the "gen_client" and the clients in this repo.
Did you run your server with Python3?
@tobegit3hub yes,my python version is 3.5, by the way ,do you know how can I convert checkpoint files(such as: .data-0000-of-0001, .meta, .index) to SavedModel for tf serving (like : saved_model.pb, variables )?
It is the problem of Python version compatibility. Issue refers to https://stackoverflow.com/questions/41314572/invalid-json-data-from-requests-get .
I will test with the :latest-py34
docker image and fix it soon.
For another issue, you can write the Python script to load the checkpoint and export SavedModel.
Another issue for Python 3 in https://stackoverflow.com/questions/18686903/dict-keys0-on-python-3 .
And we use has_key
in tensorflow_inference_service.py
.
Thank your reply! I check your update and still questions:
I run
simple_tensorflow_serving --model_base_path="./models/tensorflow_template_application_model"
and then runcurl http://localhost:8500/v1/models/__tensorflow_template_application_model__/gen_client?language=python > client.py
but the shell output:zsh: no matches found: http://localhost:8500/v1/models/tensorflow_template_application_model/gen_client?language=python
I replace
tensorflow_template_application_model
withdefault
in model_config_file.json like this: and the I runsimple_tensorflow_serving --model_config_file="./examples/model_config_file.json"
simple_tensorflow_serving is runing:In other terminal, I run
python client.py
: the output is like this :So. I have test your web _Test function: result is the same. I can not get right result..
My input_data of client.py is wrong??