NVIDIA NeMo

Building and customising models with NeMo, then serving them with NIM.

Lab outcome

Serve the fine-tuned model through NIM on OpenShift, behind guardrails, and demo better answers than the base model.

  1. 01

    Overview

    How NeMo (build and customise models) fits with NIM (serve models) on OpenShift.

    Coming soon
  2. 02

    Curator

    Clean, filter and deduplicate a sample company dataset for training.

    Coming soon
  3. 03

    Customizer

    LoRA fine-tune a Llama or Mistral model on that dataset.

    Coming soon
  4. 04

    Evaluator

    Benchmark the base model against the fine-tuned model and record the results.

    Coming soon
  5. 05

    Guardrails

    Add topic, safety and jailbreak rails to the end-to-end chatbot.

    Coming soon
  6. 06

    Retriever

    Use the embedding and reranking NIMs in the RAG pipeline and compare retrieval quality.

    Coming soon