not enough values to unpack in server.py

[thetuxedo]

Hello, I tried to run example5.py on a single-GPU ec2 instance. I set num of worker to 1.

After a few hours running, it will popup following errors:

on client (same instance, another session)

I tried twice, always the case. Do you have any suggestion? Thanks.

Also, usually how long it takes to finish the training of example5.py on a single-gpu machine?

[hanxiao]

I believe it's a bug. Will do a check.

[wanghm92]

I seem to run into a similar problem:

Any updates on the bug fix? Thanks!

[hanxiao]

pip install the latest version and try again. https://github.com/hanxiao/bert-as-service#install

[thetuxedo]

i tried again, there is a new error.

[mooninrain]

latest pip version, same error again

[hanxiao]

sorry but i can't reproduce the error on example5.py.

some general suggestions:

• Update to latest version 1.4.9 pip install -U bert-serving-server bert-serving-client
• Remove .prefetch(20) if you are running the server side on CPU, not necessary also prone to the race condition
• Set num_parallel_calls=1
• Or please provide concrete environment and steps for reproducing the error

[hanxiao]

close due to inactivity

[felixhao28]

We ran into a similar issue on version 1.6.4. The case seemingly appears at random and is not reliably reproducible.

[felixhao28]

Also the client hangs forever when server crashes even when timeout=5000 is set.

[hanxiao]

@felixhao28 please do pip install -U bert-serving-server bert-serving-client and use the latest version.

Better start a new issue and fill in the issue form, helping me to locate the problem faster.

Note that BertClient.encode() is not thread-safe, whereas BertClient itself is thread-safe. Therefore if you are using BertClient in a MT environment, please be careful and refer to #29 (comment)

[hanxiao]

timeout is another issue, will be addressed separately. Please always create a new issue if you are reporting something new. thanks

[wanghm92]

Like felixhao28 said, this issue is kind of random, which is probably why you could not reproduce it, at least not easily.

However, it does happen every time for sure, at least for me. It's just a matter of how long it holds until it crashes. I use one server hosted on one GPU and let multiple programs query it at random frequency. You may want to reopen this issue and hope for more inputs from more users.

I've not changed the default timeout, so I guess that's a separate issue.

Thanks!

[hanxiao]

@wanghm92 could you open a new issue and fill in the issue form, knowing the version and sys info can help me to understand problem much much faster.

[hanxiao]

are you all encountering this problem on example5.py?

[hanxiao]

fyi, @felixhao28 timeout issue is solved in #181 and is available since 1.6.7. Please do

pip install -U bert-serving-server bert-serving-client

for the update.

[felixhao28]

Thanks, we will try version 1.6.7 and report back to you later if the issue persists.

[yanzp0214]

my bert running server throw this error every once in a while whether i have used the server

[hanxiao]

could you please be more specific?

Prerequisites

Please fill in by replacing [ ] with [x].

• Are you running the latest bert-as-service?
• Did you follow the installation and the usage instructions in README.md?
• Did you check the FAQ list in README.md?
• Did you perform a cursory search on existing issues?

System information

Some of this information can be collected via this script.

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
• TensorFlow installed from (source or binary):
• TensorFlow version:
• Python version:
• bert-as-service version:
• GPU model and memory:
• CPU model and memory:

---

Description

Please replace YOUR_SERVER_ARGS and YOUR_CLIENT_ARGS accordingly. You can also write your own description for reproducing the issue.

I'm using this command to start the server:

bert-serving-start YOUR_SERVER_ARGS

and calling the server via:

bc = BertClient(YOUR_CLIENT_ARGS)
bc.encode()

Then this issue shows up:

...

[yanzp0214]

my bert service version is 1.6.9 and the latest throwing bug is after i starting bert server several hours .i never send a request to the server

[hanxiao]

huh? @yanzp0214 no request to the server and then it's down?

I'm reopening this issue.

[yanzp0214]

Yeah~
My test result is that no matter whether the bert server receives the request or not, the server will throw this exception after a period of time.