Skip to main content
Ctrl+K
Sports Intelligence Playbooks - Home Sports Intelligence Playbooks - Home

Sports Intelligence Playbooks

Sports Intelligence Playbooks - Home Sports Intelligence Playbooks - Home

Sports Intelligence Playbooks

Table of Contents

Multimodal Language Models

  • Data Curation
  • Training
    • Setup
    • SFT
    • LoRA
  • Inference
  • Evaluation
  • Learnings
    • Data Size Impact
    • Frame Resolution Impact
    • Training Data Bias Impact
    • Video Chunking Impact
    • Vision Encoder Impact
  • Reference Sports
    • Tennis
    • Basketball
  • Published Checkpoints
  • Runtime Optimization
    • Quantization
  • Deployment
    • vLLM serving
  • Performance

Reference

  • Licenses
  • Published Checkpoints

Published Checkpoints#

Sport-specific SFT checkpoints built with the Sports Intelligence Playbooks.

Sport

Hugging Face model

Training Data

Tennis

NVIDIA-NemotronLabs-AI-for-Media-Sports-Tennis

In-house, manually labeled point-level clips (1.31M Q&A pairs, 43k point clips from 239 tennis matches).

Basketball

NVIDIA-NemotronLabs-AI-for-Media-Sports-Basketball

BARD basketball broadcast clips (175k Q&A rows from 14,676 clips).

previous

Basketball

next

Runtime Optimization

NVIDIA NVIDIA
Privacy Policy | Your Privacy Choices | Terms of Service | Accessibility | Corporate Policies | Product Security | Contact

Copyright © 2026, NVIDIA Corporation.

Last updated on Nov 03, 2026.