News Summarization:
Load Base Model

Introduction

The playground is designed for experimenting with different base models, allowing you to combine them with your datasets and tailored prompts to create a snapshot. This snapshot, comprising a large language model (LLM), a specific prompt, and possibly a dataset, can be saved for future uses like fine-tuning.

Part 3: Load Base Model

Open the Playground

  1. From the project’s Snapshots tab, select New Playground.

Configure Model Properties

  1. Open the Select model drawer and then pick a base model from the list, such as mpt-7b-instruct main or Llama-2-7b-chat-hf. Both models are already fine-tuned for taking instructions.
  2. Select a Resource pool. Opt for the a100 resource pool of NVIDIA A100s. NVIDIA A100s are generally recommended for inference and fine-tuning However, if your resources are limited, you can use NVIDIA V100s or NVIDIA T4s.
  3. Select Load.
Adjusting Model Generation Properties
Adjusting model generation properties is crucial as it significantly impacts output quality. Experiment with different settings to optimize the model’s performance for your specific needs.

Recap

  • You’ve successfully configured the playground with a base model suitable for instruction-based tasks.
  • An appropriate resource pool has been selected to support efficient model operation.