fix: load the local finetuning model from pipeline YAML (#4729) (#4760)

If using the local model in pipeline YAML. The PromptModel cannot select
the HFLocalInvocationLayer, because of the get_task cannot support the
offline model.

*Local model usage:
  add the task_name parameter in model_kwargs for local model. for
  example text-generation or text2text-generation.

- name: PModel
  type: PromptModel
  params:
    model_name_or_path: /local_model_path
    model_kwargs:
      task_name: text-generation
- name: Prompter
  params:
    model_name_or_path: PModel
    default_prompt_template: question-answering
  type: PromptNode

Signed-off-by: yuanwu <yuan.wu@intel.com>

haystack/nodes/prompt/invocation_layer/hugging_face.py

@@ -1,5 +1,6 @@
 from typing import Optional, Union, List, Dict
 import logging
+import os
 
 import torch
 
@@ -266,6 +267,9 @@ class HFLocalInvocationLayer(PromptModelInvocationLayer):
     @classmethod
     def supports(cls, model_name_or_path: str, **kwargs) -> bool:
         task_name: Optional[str] = None
+        if os.path.exists(model_name_or_path):
+            return True
+
         try:
             task_name = get_task(model_name_or_path, use_auth_token=kwargs.get("use_auth_token", None))
         except RuntimeError: