Llama2 FT - #2234

I have added two versions: llama2.py is in the old format and llama2_new.py is in the new version. Also, I added a new version of huggingface_interface_new.py. which I added the quantization config and AUtomodelCasualLM. Let me know what you think of the changes @mhn226 .. When we finalize the best for this interface, I will update the recipe accordingly

Thank you @poonehmousavi for this PR. It's a valuable addition to SpeechBrain. Here are my comments from a first review of the code:

1. Failing Tests:

   • It appears there are some failing tests, mainly related to style issues.

2. Permission Issue in Recipe Tests:

   • The recipe tests seem to be failing due to a permission issue.

     python -c 'from tests.utils.recipe_tests import run_recipe_tests; print("TEST FAILED!") if not(run_recipe_tests(filters_fields=["Dataset"], filters=[["MultiWOZ"]], do_checks=True, run_opts="--device=cuda")) else print("TEST PASSED")'

I got the same issue when running the llama2 recipe. Are you able to replicate it?

3. README Correction:
   • In the README file, there's a small typo in the directory path. Please correct:

     cd recipes/MultiWOZ/response_generation[LLM_model] => cd recipes/MultiWOZ/response_generation/[LLM_model]

@mhn226 do you have additional comments?

Thank you @poonehmousavi for this PR. It's a valuable addition to SpeechBrain. Here are my comments from a first review of the code:

1. **Failing Tests:**
   
   * It appears there are some failing tests, mainly related to style issues.

2. **Permission Issue in Recipe Tests:**
   
   * The recipe tests seem to be failing due to a permission issue.
     ```python
     python -c 'from tests.utils.recipe_tests import run_recipe_tests; print("TEST FAILED!") if not(run_recipe_tests(filters_fields=["Dataset"], filters=[["MultiWOZ"]], do_checks=True, run_opts="--device=cuda")) else print("TEST PASSED")'
     ```

I got the same issue when running the llama2 recipe. Are you able to replicate it?

3. **README Correction:**
   
   * In the README file, there's a small typo in the directory path. Please correct:
     ```shell
     cd recipes/MultiWOZ/response_generation[LLM_model] => cd recipes/MultiWOZ/response_generation/[LLM_model]
     ```

@mhn226 do you have additional comments?

Hi @mravanelli , maybe come back here next week because currently we're trying to figure out what is the best option for doing this. The PR is rather a draft version for now ^^
And yes for llama2 we need to ask meta for the access. So for testing we might need to skip or maybe use this.

I closed this PR and transfer all the changes to this PR