无法使用 CLI 在 spacy 3.0 中评估自定义 ner

Question

我正在尝试从磁盘加载带有自定义管道的自定义预训练模型作为 spacy 3.0 中的管道：

工厂的代码是这样的：

@CustomEng.factory("ner-crf")
def create_my_component(nlp, name):
    crf_extractor = CRFExtractor().from_disk("path-to-model")
    return CRFEntityExtractor(nlp, crf_extractor=crf_extractor)

然后我将 'ner-crf' 添加到我的语言 class 中，如下所示：

    nlp = spacy.blank('custom-eng')
    nlp.add_pipe('ner-crf')
    nlp.to_disk('../model')

有一件事我认为可能是相关的：当我使用 to_disk 来保存 nlp 对象时，没有 ner-crf 包（我刚刚添加的管道）保存的对象。

然后我运行这个 CLI 命令来评估 NER 管道：

python -m spacy evaluate ../model/ ../corpus/dev.spacy --output ../model/metrics.json --gpu-id 0 --code ../../../spacy_utils/custom-eng/__init__.py

但是我得到这个错误：

Traceback (most recent call last):
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/__main__.py", line 4, in <module>
    setup_cli()
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/cli/_util.py", line 69, in setup_cli
    command(prog_name=COMMAND)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/click/core.py", line 829, in __call__
    return self.main(*args, **kwargs)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/click/core.py", line 782, in main
    rv = self.invoke(ctx)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/click/core.py", line 1259, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/click/core.py", line 1066, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/click/core.py", line 610, in invoke
    return callback(*args, **kwargs)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/typer/main.py", line 497, in wrapper
    return callback(**use_params)  # type: ignore
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/cli/evaluate.py", line 42, in evaluate_cli
    evaluate(
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/cli/evaluate.py", line 75, in evaluate
    nlp = util.load_model(model)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/util.py", line 326, in load_model
    return load_model_from_path(Path(name), **kwargs)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/util.py", line 392, in load_model_from_path
    return nlp.from_disk(model_path, exclude=exclude)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/language.py", line 1883, in from_disk
    util.from_disk(path, deserializers, exclude)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/util.py", line 1176, in from_disk
    reader(path / key)
  File "/home/marzi/anaconda3/envs/spacy-tutorial/lib/python3.8/site-packages/spacy/language.py", line 1877, in <lambda>
    deserializers[name] = lambda p, proc=proc: proc.from_disk(
TypeError: from_disk() got an unexpected keyword argument 'exclude'

我使用的自定义 NER classes 属于 spacy-crfsuite 库，它在 spacy 2 中运行良好，但它们还没有 Spacy 3 的示例代码，所以我正在尝试使其运行我自己在 spacy 3.0 中。

Answer 1

从 spaCy v3.0 开始，管道组件应在其 to_disk 方法中支持 exclude 关键字。你可以在你的函数中添加exclude关键字，给它一个默认值，并且干脆不在函数体中使用它的值，这个错误应该可以解决。

为了完整起见，这里是从 v2 到 v3 的迁移指南，其中可能包含一些其他有趣的提示：https://spacy.io/usage/v3#migrating

无法使用 CLI 在 spacy 3.0 中评估自定义 ner

Can't evaluate custom ner in spacy 3.0 using CLI

python

spacy

spacy-3