Automatic num_params in LLM + update GRetriever default llm

[Kh4L]

This PR:

• Introduces llm_model_name argument in g_retriever.py to allow specifying the LLM model
• Make num_params optional in the LLM constructor, automatically determining it using huggingface_hub metadata if not provided
• Change the default LLM model to the more recent meta-llama/Meta-Llama-3.1-8B-Instruct

Previous model meta-llama/Llama-2-7b-chat-hf metrics:

Hit: 0.6966
Precision: 0.6250
Recall: 0.5344
F1: 0.5405
Total Training Time: 556.111935s

Newer model meta-llama/Llama-3.1-8B-Instruct metrics:

Hit: 0.7629
Precision: 0.7145
Recall: 0.6027
F1: 0.6190
Total Training Time: 572.248117s

[puririshi98]

This is great improvements, thanks.

[puririshi98]

please get the CI green so its mergeable

[akihironitta]

@Kh4L The annotated type is no longer correct. Mind sending a quick PR?