Allennlp: Error when loading the coreference SpanBERT model

[izaskr]

Describe the bug
Loading the model for coreference resolution seems to not work: the model is not loaded/recognized.

Log:
Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/predictors/predictor.py", line 256, in from_path return Predictor.from_archive(load_archive(archive_path, cuda_device=cuda_device), predictor_name, File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/models/archival.py", line 230, in load_archive cuda_device=cuda_device) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/models/model.py", line 327, in load return cls.by_name(model_type)._load(config, serialization_dir, weights_file, cuda_device) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/models/model.py", line 265, in _load model = Model.from_params(vocab=vocab, params=model_params) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/from_params.py", line 365, in from_params return subclass.from_params(params=params, **extras) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/from_params.py", line 386, in from_params kwargs = create_kwargs(cls, params, **extras) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/from_params.py", line 133, in create_kwargs kwargs[name] = construct_arg(cls, name, annotation, param.default, params, **extras) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/from_params.py", line 229, in construct_arg return annotation.from_params(params=subparams, **subextras) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/from_params.py", line 365, in from_params return subclass.from_params(params=params, **extras) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/modules/text_field_embedders/basic_text_field_embedder.py", line 160, in from_params for name, subparams in token_embedder_params.items() File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/modules/text_field_embedders/basic_text_field_embedder.py", line 160, in <dictcomp> for name, subparams in token_embedder_params.items() File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/from_params.py", line 359, in from_params default_to_first_choice=default_to_first_choice) File "/home/CE/skr/anaconda3/envs/env_allennlp2/lib/python3.7/site-packages/allennlp/common/params.py", line 363, in pop_choice raise ConfigurationError(message) allennlp.common.checks.ConfigurationError: 'pretrained_transformer_mismatched not in acceptable choices for model.text_field_embedder.token_embedders.tokens.type: [\'embedding\', \'character_encoding\', \'elmo_token_embedder\', \'elmo_token_embedder_multilang\', \'openai_transformer_embedder\', \'bert-pretrained\', \'language_model_token_embedder\', \'bidirectional_lm_token_embedder\', \'bag_of_word_counts\', \'pass_through\', \'pretrained_transformer\']. You should either use the --include-package flag to make sure the correct module is loaded, or use a fully qualified class name in your config file like {"model": "my_module.models.MyModel"} to have it imported automatically.'

To Reproduce
The code:
from allennlp.predictors.predictor import Predictor
predictor = Predictor.from_path("https://storage.googleapis.com/allennlp-public-models/coref-spanbert-large-2020.02.27.tar.gz")
predictor.predict(document="The woman reading a newspaper sat on the bench with her dog.")
The error is raised after defining the predictor variable.

Expected behavior
Load the model with no errors.

System (please complete the following information):

• OS: Linux
• Python version: 3.7.0
• AllenNLP version: 0.9.0 (installed with pip into a clean conda env)

Additional context
I was following the demo here. I also looked at this example - the part of loading the coreference model, but the same error was raised. Using the predictor_name argument found here didn't solve the issue either.

[datix-vishu]

I don't think you can run the "coref-spanbert-large-2020.02.27.tar.gz" model with AllenNLP v0.9.0

You need a newer version to run this model. Try these packages if you want the same results as the demo:

AllenNLP package : https://github.com/allenai/allennlp/tree/adeb1b1278619ff2d74d4fd82825e50a36f95ff4

Model package : https://github.com/allenai/allennlp-models/tree/baf3a1ec3b74273a4ffa2112d37fb88e8b3dd39c

[izaskr]

Isn't 0.9.0 the newest official release of AllenNLP?

I have installed the packages you suggested into a clean conda env as follows:

• pip install git+git://github.com/allenai/allennlp.git@adeb1b1278619ff2d74d4fd82825e50a36f95ff4
• pip install git+https://github.com/allenai/allennlp-models.git@baf3a1ec3b74273a4ffa2112d37fb88e8b3dd39c

This installs the versions:

• allennlp===0.9.1-unreleased
• allennlp-models===0.0.1-unreleased

When using the code from the demo, I still get the error:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/CE/skr/anaconda3/envs/env_allennlp3/lib/python3.7/site-packages/allennlp/predictors/predictor.py", line 268, in from_path
    load_archive(archive_path, cuda_device=cuda_device),
  File "/home/CE/skr/anaconda3/envs/env_allennlp3/lib/python3.7/site-packages/allennlp/models/archival.py", line 186, in load_archive
    cuda_device=cuda_device,
  File "/home/CE/skr/anaconda3/envs/env_allennlp3/lib/python3.7/site-packages/allennlp/models/model.py", line 340, in load
    model_class: Type[Model] = cls.by_name(model_type)  # type: ignore
  File "/home/CE/skr/anaconda3/envs/env_allennlp3/lib/python3.7/site-packages/allennlp/common/registrable.py", line 137, in by_name
    subclass, constructor = cls.resolve_class_name(name)
  File "/home/CE/skr/anaconda3/envs/env_allennlp3/lib/python3.7/site-packages/allennlp/common/registrable.py", line 185, in resolve_class_name
    f"{name} is not a registered name for {cls.__name__}. "
allennlp.common.checks.ConfigurationError: coref is not a registered name for Model. You probably need to use the --include-package flag to load your custom code. Alternatively, you can specify your choices using fully-qualified paths, e.g. {"model": "my_module.models.MyModel"} in which case they will be automatically imported correctly.

Any ideas?

[datix-vishu]

use this code

from allennlp.predictors import Predictor
import allennlp_models.coref
coref_model = Predictor.from_path('..../coref-spanbert-large-2020.02.27.tar.gz')
coref_model.coref_resolved(text)

or

coref_model.predict(text)

[izaskr]

Thank you, your suggestions worked.

Alternatively, what one can do to load the model is:

• install the dev version 0.9.1 with pip install allennlp==0.9.1.dev20200228
• use the code snippet from the demo, so no need to install the allennlp_models or import them

[mshen2]

Thank you, your suggestions worked.

Alternatively, what one can do to load the model is:

• install the dev version 0.9.1 with pip install allennlp==0.9.1.dev20200228
• use the code snippet from the demo, so no need to install the allennlp_models or import them

Hi! I am currently using 0.9.1 without using allennlp_models, but I still could not have anything with the code in demo. It seems like the program freezes without giving any output. Do you have any idea about this issue? Thank you!

[izaskr]

I didn't have a similar problem. What are you system specifications? Did you install the packages into a clean environment?

[mshen2]

They released 1.0.0rc1 yesterday and after installing on a clean environment, I am able to use spanbert now!