Qwen

/

Qwen2-72B-Instruct

like 659

I attempt to run the sample code given by HF:

It throws the following error:

KeyError                                  Traceback (most recent call last)
Cell In[5], line 7
      2 from transformers import pipeline
      4 messages = [
      5     {"role": "user", "content": "Who are you?"},
      6 ]
----> 7 pipe = pipeline("text-generation", model="Qwen/Qwen2-72B-Instruct")
      8 pipe(messages)

File ~/.local/lib/python3.10/site-packages/transformers/pipelines/__init__.py:751, in pipeline(task, model, config, tokenizer, feature_extractor, image_processor, framework, revision, use_fast, token, device, device_map, torch_dtype, trust_remote_code, model_kwargs, pipeline_class, **kwargs)
    748                 adapter_config = json.load(f)
    749                 model = adapter_config["base_model_name_or_path"]
--> 751     config = AutoConfig.from_pretrained(model, _from_pipeline=task, **hub_kwargs, **model_kwargs)
    752     hub_kwargs["_commit_hash"] = config._commit_hash
    754 custom_tasks = {}

File ~/.local/lib/python3.10/site-packages/transformers/models/auto/configuration_auto.py:1050, in AutoConfig.from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
   1048     return config_class.from_pretrained(pretrained_model_name_or_path, **kwargs)
   1049 elif "model_type" in config_dict:
-> 1050     config_class = CONFIG_MAPPING[config_dict["model_type"]]
   1051     return config_class.from_dict(config_dict, **unused_kwargs)
   1052 else:
   1053     # Fallback: use pattern matching on the string.
   1054     # We go from longer names to shorter names to catch roberta before bert (for instance)
...
--> 748     raise KeyError(key)
    749 value = self._mapping[key]
    750 module_name = model_type_to_module_name(key)

KeyError: 'qwen2'```