Unable to interact with md_gender Multi Task Classifier

[lnisenbaum-tdab]

Bug description
I am not able to interact with md_gender Multi Task Classsifier model from the zoo. I tried two different approaches:

Approach 1) Google Colab Notebook

Model downloaded as expected but the following error message appears: No module named 'parlai_internal'

Reproduction steps
Enter steps to reproduce the behavior.

# Import the Interactive script
from parlai.scripts.interactive import Interactive

# call it with particular args
Interactive.main(
    # the model_file is a filename path pointing to a particular model dump.
    # Model files that begin with "zoo:" are special files distributed by the ParlAI team.
    # They'll be automatically downloaded when you ask to use them.
    model_file='zoo:md_gender/model'
)

Expected behavior
I was expecting to be able to interact with Multi Task Classifier similarly to what is shown in Google Colab Parlai Tutorial for the tutorial_transformer_generator

Logs
Please paste the command line output:

15:04:53 | building data: /usr/local/lib/python3.7/dist-packages/data/models/md_gender/md_gender_classifier.tgz
15:04:53 | Downloading http://parl.ai/downloads/_models/md_gender/md_gender_classifier.tgz to /usr/local/lib/python3.7/dist-packages/data/models/md_gender/md_gender_classifier.tgz
Downloading md_gender_classifier.tgz: 100%|██████████| 2.29G/2.29G [01:37<00:00, 23.6MB/s]
---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-12-f7aa5f599605> in <module>()
      7     # Model files that begin with "zoo:" are special files distributed by the ParlAI team.
      8     # They'll be automatically downloaded when you ask to use them.
----> 9     model_file='zoo:md_gender/model'
     10 )

22 frames
/usr/local/lib/python3.7/dist-packages/parlai/core/script.py in main(cls, *args, **kwargs)
    106             return cls._run_args(args)
    107         elif kwargs:
--> 108             return cls._run_kwargs(kwargs)
    109         else:
    110             return cls._run_args(None)

/usr/local/lib/python3.7/dist-packages/parlai/core/script.py in _run_kwargs(cls, kwargs)
     71         """
     72         parser = cls.setup_args()
---> 73         opt = parser.parse_kwargs(**kwargs)
     74         return cls._run_from_parser_and_opt(opt, parser)
     75 

/usr/local/lib/python3.7/dist-packages/parlai/core/params.py in parse_kwargs(self, **kwargs)
   1265         self.error = _captured_error
   1266         try:
-> 1267             string_args = self._kwargs_to_str_args(**kwargs)
   1268             return self.parse_args(args=string_args)
   1269         finally:

/usr/local/lib/python3.7/dist-packages/parlai/core/params.py in _kwargs_to_str_args(self, **kwargs)
   1217         # become aware of any extra args that might be specified if the user
   1218         # provides something like model="transformer/generator".
-> 1219         self.add_extra_args(string_args)
   1220 
   1221         # do it again, this time knowing about ALL args.

/usr/local/lib/python3.7/dist-packages/parlai/core/params.py in add_extra_args(self, args)
    936         model = get_model_name(parsed)
    937         if model is not None:
--> 938             self.add_model_subargs(model, partial)
    939 
    940         # add world args, if we know a priori which world is being used

/usr/local/lib/python3.7/dist-packages/parlai/core/params.py in add_model_subargs(self, model, partial)
    810         Add arguments specific to a particular model.
    811         """
--> 812         agent = load_agent_module(model)
    813         try:
    814             if hasattr(agent, 'add_cmdline_args'):

/usr/local/lib/python3.7/dist-packages/parlai/core/loader.py in load_agent_module(agent_path)
    186             module_name = "%s.agents.%s.%s" % (repo, agent_path, agent_path)
    187 
--> 188     my_module = importlib.import_module(module_name)
    189     model_class = getattr(my_module, class_name)
    190 

/usr/lib/python3.7/importlib/__init__.py in import_module(name, package)
    125                 break
    126             level += 1
--> 127     return _bootstrap._gcd_import(name[level:], package, level)
    128 
    129 

/usr/lib/python3.7/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)

/usr/lib/python3.7/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)

/usr/lib/python3.7/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _call_with_frames_removed(f, *args, **kwds)

/usr/lib/python3.7/importlib/_bootstrap.py in _gcd_import(name, package, level)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load(name, import_)

/usr/lib/python3.7/importlib/_bootstrap.py in _find_and_load_unlocked(name, import_)

ModuleNotFoundError: No module named 'parlai_internal'

---------------------------------------------------------------------------
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the
"Open Examples" button below.
---------------------------------------------------------------------------

---

Approach 2) Run locally by previously defining an environment

I also tried running the model locally on my machine on a previously defined conda environment

Reproduction steps
Enter steps to reproduce the behavior.

1. Clonned parlai GH repo

2. Created a conda environment and executed pip install parlai

3. Made sure that every requirement listed in requirements.txt was present

4. Tried interacting with the model in the command line:
   parlai interactive -t md_gender -m projects.md_gender.bert_ranker_classifier.agents:BertRankerClassifierAgent -mf zoo:md_gender/model -ecands inline -cands inline --interactive_mode False --data-parallel False

5. At this point I had an error message stating that spacy and en_core_web_sm-2.2.0 packages were missing
   I then Installed spacy and also the package en_core_web_sm-2.2.0 running
   pip3 install https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-2.2.0/en_core_web_sm-2.2.0.tar.gz

6. Tried interacting with the model one more time but another error messsage popped up

Expected behavior
I was expecting to be able to interact with Multi Task Classifier similarly to what is shown in Google Colab Parlai Tutorial for the tutorial_transformer_generator

Logs
Please paste the command line output:

/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/util.py:717: UserWarning: [W094] Model 'en_core_web_sm' (2.2.0) specifies an under-constrained spaCy version requirement: >=2.2.0. This can lead to compatibility problems with older versions, or as new spaCy versions are released, because the model may say it's compatible when it's not. Consider changing the "spacy_version" in your meta.json to a version range, with a lower and upper pin. For example: >=3.0.6,<3.1.0
  warnings.warn(warn_msg)
Traceback (most recent call last):
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/bin/parlai", line 8, in <module>
    sys.exit(main())
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/__main__.py", line 14, in main
    superscript_main()
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/script.py", line 297, in superscript_main
    opt = parser.parse_args(args)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/params.py", line 1153, in parse_args
    self.add_extra_args(args)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/script.py", line 156, in add_extra_args
    sa.choices[args[0]].add_extra_args(args)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/params.py", line 930, in add_extra_args
    self.add_task_args(task, partial)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/params.py", line 839, in add_task_args
    agent = load_teacher_module(t)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/loader.py", line 281, in load_teacher_module
    task_module = load_task_module(taskname)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/core/loader.py", line 247, in load_task_module
    task_module = importlib.import_module(module_name)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1006, in _gcd_import
  File "<frozen importlib._bootstrap>", line 983, in _find_and_load
  File "<frozen importlib._bootstrap>", line 967, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 677, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 728, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/tasks/md_gender/agents.py", line 18, in <module>
    from .wikipedia import WikipediaTeacher  # noqa: F401
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/parlai/tasks/md_gender/wikipedia.py", line 25, in <module>
    NLP = spacy.load('en_core_web_sm')
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/__init__.py", line 51, in load
    name, vocab=vocab, disable=disable, exclude=exclude, config=config
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/util.py", line 324, in load_model
    return load_model_from_package(name, **kwargs)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/util.py", line 357, in load_model_from_package
    return cls.load(vocab=vocab, disable=disable, exclude=exclude, config=config)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/en_core_web_sm/__init__.py", line 12, in load
    return load_model_from_init_py(__file__, **overrides)
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/util.py", line 522, in load_model_from_init_py
    config=config,
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/util.py", line 390, in load_model_from_path
    config = load_config(config_path, overrides=dict_to_dot(config))
  File "/Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/spacy/util.py", line 547, in load_config
    raise IOError(Errors.E053.format(path=config_path, name="config.cfg"))
OSError: [E053] Could not read config.cfg from /Users/laureanonisenbaum/opt/anaconda3/envs/parlai/lib/python3.7/site-packages/en_core_web_sm/en_core_web_sm-2.2.0/config.cfg

Additional context
None

[stephenroller]

cc @emilydinan

[ivanscattergood]

To run the md_gender model so that it returns the stats about the self, partner and gender neutral you must set the interactive_mode='false', this will force it to use the worlds.py file and display sorted scores.

I have attached a text file, that if converted to a python file you can call this code interactively:
md_gender.txt

And you will get the following output:

[emilydinan]

Hi @lnisenbaum-tdab, just as @ivanscattergood suggest, you will have to set a couple of additional parameters with your call to the interactive script. See the README here to find the command:

parlai interactive -t md_gender -m projects.md_gender.bert_ranker_classifier.agents:BertRankerClassifierAgent -mf zoo:md_gender/model -ecands inline -cands inline --interactive_mode False --data-parallel False

You will have to specify the model path (projects.md_gender.bert_ranker_classifier.agents:BertRankerClassifierAgent) as well as the cands/ecands/interactive_mode/data_parallel parameters.

Feel free to re-open if you still have issues with this.

[lnisenbaum-tdab]

Yes, thanks! We got to interact with it by adding those parameters.

Also, we set up a Python 3.8.5 environment and run:

python setup.py develop

Then installed some missing dependencies:

pip install spacy>3.0.6

python -m spacy download en_core_web_sm

pip install pytorch-pretrained-bert

I will try to test it on a batch of samples now but in terms of setup we are good to close this

Cheers!