from_disk method throws ValueError: Cannot deserialize model: mismatched structure

[Tejas-Garhewal]

The following code is taken from https://spacy.io/usage/processing-pipelines#pipelines:

lang = "en"
pipeline = ["tok2vec", "tagger", "parser", "ner", "attribute_ruler", "lemmatizer"]
data_path = "path/to/en_core_web_sm/en_core_web_sm-3.0.0" ### I only changed this path

cls = spacy.util.get_lang_class(lang)  # 1. Get Language class, e.g. English
nlp = cls()                            # 2. Initialize it
for name in pipeline:
    nlp.add_pipe(name)                 # 3. Add the component to the pipeline
nlp.from_disk(data_path)               # 4. Load in the binary data

And I receive the above error.

My spacy version is 3.6.0 and model version is en_core_web_sm-3.2.0 (I have also changed the model name appropriately)

Can anyone please tell me where I'm going wrong? Thank you in advance!

Please note that spacy.load("en_core_web_sm) works fine

[adrianeboyd]

If you're just trying to load from a directory, please use spacy.load("/path/to/dir") instead.

If you're trying to understand how spacy.load works, please look at spacy.util.load_model_from_path for the exact details. (This code in the docs is intended more as a sketch/pseudocode than a 100% runnable example, but I agree that this isn't clear from the docs.)

[Tejas-Garhewal]

Thank you very much for your help!!

For others who would like to see how it works without digging in their files, the following code
works on my machine:

lang = "en"
pipeline = ["tok2vec", "tagger", "parser", "ner", "attribute_ruler", "lemmatizer"]
data_path = "C:/Users/scien/anaconda3/lib/site-packages/en_core_web_sm/en_core_web_sm-3.6.0" # I upgraded my model to match the spacy version as part of my previous efforts to get the original code to work

# cls = spacy.util.get_lang_class(lang)  # 1. Get Language class, e.g. English
# nlp = cls()                            # 2. Initialize it
# for name in pipeline:
#     nlp.add_pipe(name)                 # 3. Add the component to the pipeline
# nlp.from_disk(data_path)               # 4. Load in the binary data

meta = spacy.util.get_model_meta(data_path)
config_path = data_path + "/config.cfg"
overrides = spacy.util.dict_to_dot(spacy.util.SimpleFrozenDict(), #this is config
                                   for_overrides=True)
config = spacy.util.load_config(config_path, overrides=overrides)
nlp__ = spacy.util.load_model_from_config(config, 
                                          vocab=True,
                                          disable = spacy.util._DEFAULT_EMPTY_PIPES,
                                          enable = spacy.util._DEFAULT_EMPTY_PIPES,
                                          exclude = spacy.util._DEFAULT_EMPTY_PIPES,
                                          meta = meta
                                         )
nlp__.from_disk(data_path, 
              exclude = spacy.util._DEFAULT_EMPTY_PIPES,
              overrides = overrides
             )