Load models with YAML or JSON¶
Camphr provides a top-level function, camphr.load
:
You can pass the model configuration in a format that omegaconf can parse, such as dict, yaml and json, for example:
>>> import camphr
>>> cfg = {
>>> "lang": {"name": "en"},
>>> "pipeline": {
>>> "sentencizer": {"punct_chars": [".", "!"]},
>>> "transformers_model": {"trf_name_or_path": "xlnet-base-cased"},
>>> },
>>> }
>>> nlp = camphr.load(cfg)
>>> print(nlp.pipe_names)
['sentencizer', 'transformers_tokenizer', 'transformers_model']
“transformers_tokenizer” is automatically added because it is required to use “transformers_model”.
You can also pass yaml string* to camphr.load
:
>>> import camphr
>>> cfg = """
lang:
name: en
pipeline:
sentencizer:
punct_chars: [".", "!"]
transformers_model:
trf_name_or_path: xlnet-base-cased
"""
>>> nlp = camphr.load(cfg)
>>> print(nlp.pipe_names)
['sentencizer', 'transformers_tokenizer', 'transformers_model']