Bert-as-service: What does non-empty mean?

[kirdin]

Prerequisites

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

• [x] Are you running the latest bert-as-service?
• [x] Did you follow the installation and the usage instructions in README.md?
• [x] Did you check the FAQ list in README.md?
• [x] 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 pip3:
• TensorFlow version 1.11.9:
• Python version: 3.6.3
• bert-as-service version: 1.5.0
• GPU model and memory None:
• 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 -model_dir PATH  -num_worker=4

and calling the server via:

bc = BertClient()
bc.encode()

Then this issue shows up:

I'm giving list (about 135k sentences, it's a list of strings) as an argument to bc.encode. However, the system answer is AttributeError: "texts" must be "List[str]" and non-empty!

I replace all the occurences of '', ' ' and ' ' and so on in my list for a 'empty' string. However, it turned to be unsufficient.
What is considered to be "empty"? And may be there is a possibility of printing this "empty" sentence in except-branch?

Thank you in advance!

[hanxiao]

The current version behaves as follows:

for k in ([], [''], ['', ''], ['', []], ['\n', '\r\n'], ['', 's', 'ss'], ['s', 'ss'], ['', 2, 'ss'], [2, 'ss']):
    print('%s\t%s' % (BertClient._is_valid_input(k), k))

False   []
False   ['']
False   ['', '']
False   ['', []]
False   ['\n', '\r\n']
False   ['', 's', 'ss']
True    ['s', 'ss']
False   ['', 2, 'ss']
False   [2, 'ss']

But you are right, would be better to raise an error on the empty case. Mark it as enhancement and will fix it

[hanxiao]

fyi, this issue is fixed and available in 1.5.1, please do

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

[praateekmahajan]

Why not return np.zeros whenever input is empty?