Skip to Main Content
Yale Biomedical Informatics & Computing
YSM Home

INFORMATION FOR

Large Language Models

Models and Descriptions

The Kamino environment hosts a range of Llama models tailored to various NLP tasks:

  • Large Models (70B Parameters):
    • meta-llama/Llama-3.1-70B and Meta-Llama-3-70B are powerful general-purpose models for complex tasks.
    • Instruction-tuned versions (Llama-3.1-70B_Instruct and Meta-Llama-3-70B-Instruct) excel at following detailed user instructions for tasks like question answering and reasoning.
  • Mid-Sized Models (8B Parameters):
    • meta-llama/Llama-3.1-8B and Meta-Llama-3-8B are efficient models for everyday text processing tasks like classification and summarization.
    • Instruction-tuned variants (Llama-3.1-8B-Instruct and Meta-Llama-3-8B-Instruct) provide enhanced performance in tasks requiring precise instructions, such as named entity recognition (NER) and text generation.
  • Lightweight Model (3B Parameters):
    • meta-llama/Llama-3.2-3B-Instruct is a small, fast, instruction-tuned model suitable for lightweight tasks like FAQ generation and sentiment analysis.
    • These models cater to a variety of applications, from general-purpose text tasks to instruction-specific applications, balancing performance and efficiency based on their size.

Applications

  • Text Classification: Categorizing documents or text data into predefined classes, such as spam detection, sentiment analysis, or medical report classification.
  • Named Entity Recognition (NER): Identifying entities like names, dates, organizations, or medical terms in unstructured text.
  • Text Summarization: Generating concise summaries of long documents or articles.
  • Text Generation: Creating coherent and contextually accurate text for tasks like content creation, report generation, or dialogue simulation.
  • Question Answering: Extracting relevant answers from texts or databases for specific queries.
  • Machine Translation: Translating text from one language to another with contextual accuracy.
  • Sentiment Analysis: Identifying the emotional tone of text for applications like customer feedback analysis or social media monitoring.
  • Logical Reasoning and Step-by-Step Guidance: Solving multi-step problems or providing structured guidance in domains like education or troubleshooting.
  • Token-Level Tasks: Fine-grained tasks such as part-of-speech tagging, text highlighting, or syntactic analysis.
  • Lightweight NLP Applications: Quick and resource-efficient tasks like autocomplete, keyword extraction, or FAQ generation.