Init Config from Python: typer.Option default value not working

[janvdp]

Goal

Create the config.cfg from Python and not from the command line.

From the source:
https://github.com/explosion/spaCy/blob/master/spacy/cli/init_config.py#L28

How to reproduce the behaviour

Preparation

from pathlib import Path
from spacy.cli.init_config import init_config_cli, Optimizations

my_config = Path("./my_config.cfg")

This should work, ...

# This code 
init_config_cli(output_file=my_config, lang=lang, pipeline="ner",optimize=Optimizations.efficiency) 

..., however:

⚠ To generate a more effective transformer-based config (GPU-only),
install the spacy-transformers package and re-run this command. The config
generated now does not use transformers.
---------------------------------------------------------------------------
UndefinedError                            Traceback (most recent call last)
/tmp/ipykernel_664/2418944264.py in <module>
      3 
      4 #init_config_cli(output_file=my_config, lang=lang, pipeline="ner",optimize=Optimizations.efficiency, gpu=False, pretraining=False)
----> 5 init_config_cli(output_file=my_config, lang=lang, pipeline="ner",optimize=Optimizations.efficiency) #, gpu=False, pretraining=False)
      6 
      7 

/opt/conda/lib/python3.8/site-packages/spacy/cli/init_config.py in init_config_cli(output_file, lang, pipeline, optimize, gpu, pretraining, force_overwrite)
     60             exits=1,
     61         )
---> 62     config = init_config(
     63         lang=lang,
     64         pipeline=pipeline,

/opt/conda/lib/python3.8/site-packages/spacy/cli/init_config.py in init_config(lang, pipeline, optimize, gpu, pretraining, silent)
    171         )
    172         variables["transformer_data"] = None
--> 173     base_template = template.render(variables).strip()
    174     # Giving up on getting the newlines right in jinja for now
    175     base_template = re.sub(r"\n\n\n+", "\n\n", base_template)

/opt/conda/lib/python3.8/site-packages/jinja2/environment.py in render(self, *args, **kwargs)
   1289             return concat(self.root_render_func(ctx))  # type: ignore
   1290         except Exception:
-> 1291             self.environment.handle_exception()
   1292 
   1293     async def render_async(self, *args: t.Any, **kwargs: t.Any) -> str:

/opt/conda/lib/python3.8/site-packages/jinja2/environment.py in handle_exception(self, source)
    923         from .debug import rewrite_traceback_stack
    924 
--> 925         raise rewrite_traceback_stack(source=source)
    926 
    927     def join_path(self, template: str, parent: str) -> str:

<template> in top-level template code()

/opt/conda/lib/python3.8/site-packages/jinja2/environment.py in getitem(self, obj, argument)
    453         """Get an item or attribute of an object but prefer the item."""
    454         try:
--> 455             return obj[argument]
    456         except (AttributeError, TypeError, LookupError):
    457             if isinstance(argument, str):

UndefinedError: 'None' has no attribute 'efficiency'

The problem

The problem is related to typer.Option: fastapi/typer#106, which does not set the correct default value for gpu and pretraining:

gpu: bool = Opt(False, "--gpu", "-G", help="Whether the model can run on GPU. This will impact the choice of architecture, pretrained weights and related hyperparameters."),
pretraining: bool = Opt(False, "--pretraining", "-pt", help="Include config for pretraining (with 'spacy pretrain')"),
(I'm not sure why this does work from the command line.

E.g.: if you print the params gpu and pretraining, you get

<typer.models.OptionInfo object at 0x7f7eb4592430>
<typer.models.OptionInfo object at 0x7f7eb4592f10>

instead of their similar default value False.

Workaround

As a workaround, I force the value of gpu and pretraining:

init_config_cli(output_file=my_config, lang=lang, pipeline="ner",optimize=Optimizations.efficiency, gpu=False, pretraining=False)

Your Environment

• spaCy version: 3.2.4
• Platform: Linux-5.4.0-109-generic-x86_64-with-glibc2.17
• Python version: 3.8.12

[rmitsch]

Yeah, that should be fixed. Thanks for doing some digging already, @janvdp! We'll look into this.

[rmitsch]

#10788 tackles this by introducing default values for init_config() (which are in synch with those from init_config_cli()). So if want to initialize your config from Python code, please use init_config().

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.