Quick Start Error: (container log with python closure container) TypeError: 'str' object does not support item assignment

[FrankY-Sa]

python3.5 clipper-develop quick_start: sum-model

python_deployer.deploy_python_closure(clipper_conn, name="sum-model", version=1, input_type="doubles", func=feature_sum, pkgs_to_install=['numpy','scipy'])

Hi,
I tried deploying sum-model gets stuck on

18-09-18:10:13:51 INFO [docker_container_manager.py:257] Found 0 replicas for sum-model:1. Adding 1

I print the docker's healthy status (unhealthy)

18-09-18:10:13:52 INFO [docker_container_manager.py:272] Container status: running, Docker healthy status: starting
18-09-18:10:13:55 INFO [docker_container_manager.py:272] Container status: running, Docker healthy status: unhealthy

The container log shows that

Starting Python Closure container
Connecting to Clipper with default port: 7000
Traceback (most recent call last):
File "/container/python_closure_container.py", line 56, in
rpc_service.get_input_type())
File "/container/python_closure_container.py", line 28, in init
self.predict_func = load_predict_func(predict_path)
File "/container/python_closure_container.py", line 17, in load_predict_func
return cloudpickle.load(serialized_func_file)
File "/usr/local/lib/python3.5/site-packages/cloudpickle/cloudpickle.py", line 1044, in _make_skel_func
base_globals['builtins'] = builtins
TypeError: 'str' object does not support item assignment
Encountered error not related to missing packages. Please refer to the container log to diagnose.

Any help would be really appreciated!

[withsmilo]

@FrankY-Sa : Please share your feature_sum function with me.

[FrankY-Sa]

@withsmilo

clipper_conn.register_application(name="hello-world", input_type="doubles", default_output="-1.0", slo_micros=100000)

def feature_sum(xs):
return [str(sum(x)) for x in xs]

python_deployer.deploy_python_closure(clipper_conn, name="sum-model", version=1, input_type="doubles", func=feature_sum, pkgs_to_install=['numpy','scipy'])

[withsmilo]

@FrankY-Sa :
I think that the latest release(v0.5.6) of cloudpickle causes this problem. (link) Please downgrade it from v0.5.6 to v0.5.3 by applying this patch.

diff --git a/dockerfiles/Py35RPCDockerfile b/dockerfiles/Py35RPCDockerfile
index 79612a2..883d448 100644
--- a/dockerfiles/Py35RPCDockerfile
+++ b/dockerfiles/Py35RPCDockerfile
@@ -9,7 +9,7 @@ RUN mkdir -p /model \
       && apt-get update \
       && apt-get install -y libzmq3 libzmq3-dev redis-server libsodium13 build-essential

-RUN pip install cloudpickle==0.5.* pyzmq==17.0.* prometheus_client==0.1.* \
+RUN pip install cloudpickle==0.5.3 pyzmq==17.0.* prometheus_client==0.1.* \
     pyyaml==3.12.* jsonschema==2.6.* redis==2.10.* psutil==5.4.* flask==0.12.2 \
     numpy==1.14.*

[FrankY-Sa]

@withsmilo Thanks! others encountered the issue. The reason is that my local cloudpickle version(0.5.5) is higher than the one used in docker container(0.5.2), making them incompatible.