DensePassageRetriever for a french model

[SinaArdehali]

I am getting an issue with loading the following french model, which requires a camembert tokenizer and it doesn't work:

dpr_retriever_etalab_french  = DensePassageRetriever(document_store=document_store, 
                                  query_embedding_model= "etalab-ia/dpr-question_encoder-fr_qa-camembert",
                                  passage_embedding_model= "etalab-ia/dpr-ctx_encoder-fr_qa-camembert"

Here is the error i get:

Traceback (most recent call last):
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/multiprocessing/process.py", line 315, in _bootstrap
    self.run()
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/site-packages/uvicorn/_subprocess.py", line 76, in subprocess_started
    target(sockets=sockets)
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/site-packages/uvicorn/server.py", line 60, in run
    return asyncio.run(self.serve(sockets=sockets))
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/asyncio/runners.py", line 44, in run
    return loop.run_until_complete(main)
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/asyncio/base_events.py", line 647, in run_until_complete
    return future.result()
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/site-packages/uvicorn/server.py", line 67, in serve
    config.load()
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/site-packages/uvicorn/config.py", line 477, in load
    self.loaded_app = import_from_string(self.app)
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/site-packages/uvicorn/importer.py", line 21, in import_from_string
    module = importlib.import_module(module_str)
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/importlib/__init__.py", line 127, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
  File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
  File "<frozen importlib._bootstrap>", line 986, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 680, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 850, in exec_module
  File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
  File "/Users/sinaardehali/Documents/personal_project_code/qa_projects/ima_poc/QA_handler/./main.py", line 133, in <module>
    prediction = querying_pipeline.run(
  File "/opt/homebrew/Caskroom/miniforge/base/envs/haystack_ima_CondaEnv/lib/python3.9/site-packages/haystack/pipelines/base.py", line 535, in run
    raise Exception(
Exception: Exception while running node 'DPRRetrieverEtalab': index out of range in self
Enable debug logging to see the data that was passed when the pipeline failed.

Requirements:

• Haystack version (commit or version number): farm-haystack==1.14.0

[bilgeyucel]

Hey @SinaArdehali, I could reproduce this error. This is expected as 'camembert' tokenizer doesn't seem compatible with Haystack.

This is the error that I get for context: Using a model of type 'camembert' which might be incompatible with DPR encoders. Only Bert-based encoders are supported. They need input_ids, token_type_ids, attention_mask as input tensors.

[SinaArdehali]

Hey @SinaArdehali, I could reproduce this error. This is expected as 'camembert' tokenizer doesn't seem compatible with Haystack.

This is the error that I get for context: Using a model of type 'camembert' which might be incompatible with DPR encoders. Only Bert-based encoders are supported. They need input_ids, token_type_ids, attention_mask as input tensors.

Thank you @bilgeyucel for your quick answer

[bilgeyucel]

Hey @SinaArdehali, I checked with the team, and apparently, Haystack was supporting this model in previous versions but not anymore. Feel free to open a feature request for it or even contribute! We appreciate contributions a lot :)

[SinaArdehali]

Hey @SinaArdehali, I checked with the team, and apparently, Haystack was supporting this model in previous versions but not anymore. Feel free to open a feature request for it or even contribute! We appreciate contributions a lot :)

ok great. Where can I open a feature request ? on the "issue" part on github ?

[bilgeyucel]

Yes. Before you open a new issue, you'll see options such as "Bug report" and "Feature request". You need to select "Feature request" and fill in the information required.

[sinaayyy]

I have the same error when I use a camembert-type model in my retriever. However I read in a discussion that the parameter infer_tokenizer_classes=True solved the problem however on the current version of haystack (1.14.0) this parameter is no longer present in the retriever.