Dictionary update length error

[SzaboGergo01]

I tried the quantization in our own project and got this error:

We have 3 separated config, one tagger, one parser and one ner. After that we use an assemble step to merge these together. And I tried to quantize this assembled model and got that error.
config files: https://gist.github.com/SzaboGergo01/5ca9abeaaf199a686d8b41335e9fe261

Running command: /home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/bin/python -m spacy quantize-transformer --max-mse-loss 0.000003 models/hu_core_news_trf_xl-3.5.0 models/hu_core_news_trf_xl-3.5.0-quantized
Traceback (most recent call last):
  File "/home/a100/anaconda3/lib/python3.8/runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/a100/anaconda3/lib/python3.8/runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/spacy/__main__.py", line 4, in <module>
    setup_cli()
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/spacy/cli/_util.py", line 74, in setup_cli
    command(prog_name=COMMAND)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/click/core.py", line 829, in __call__
    return self.main(*args, **kwargs)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/click/core.py", line 782, in main
    rv = self.invoke(ctx)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/click/core.py", line 1259, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/click/core.py", line 1066, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/click/core.py", line 610, in invoke
    return callback(*args, **kwargs)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/typer/main.py", line 497, in wrapper
    return callback(**use_params)  # type: ignore
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/curated_transformers/cli/quantize.py", line 43, in quantize_cli
    nlp_quantize_dynamic(
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/curated_transformers/cli/quantize.py", line 75, in nlp_quantize_dynamic
    quantize_dynamic(
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/curated_transformers/cli/quantize.py", line 108, in quantize_dynamic
    quantized_model = torch.quantization.quantize_dynamic(
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/torch/ao/quantization/quantize.py", line 447, in quantize_dynamic
    model = copy.deepcopy(model)
  File "/home/a100/anaconda3/lib/python3.8/copy.py", line 172, in deepcopy
    y = _reconstruct(x, memo, *rv)
  File "/home/a100/anaconda3/lib/python3.8/copy.py", line 272, in _reconstruct
    y.__setstate__(state)
  File "/home/a100/gszabo/roberta/hu_core_news_trf_xl/.venv/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1237, in __setstate__
    self.__dict__.update(state)
ValueError: dictionary update sequence element #0 has length 12; 2 is required

My spacy environment:

## Info about spaCy

- **spaCy version:** 3.5.1
- **Platform:** Linux-5.4.0-136-generic-x86_64-with-glibc2.10
- **Python version:** 3.8.5

In principle the torch.quantization.quantize_dynamic() function is the source of the error. Have you ever encountered something similar or do you know a solution to this?

[danieldk]

Thanks for reporting! I haven't encountered this error yet. We'll look into it.

[danieldk]

I can reproduce an issue and have made a PR with fixes (and tests, so that we don't have such a regression again):

#129