lightning-ai

Quadra: Effortless and reproducible deep learning workflows with configuration files.

Actually Robust Training - Tool Inspired by Andrej Karpathy "Recipe for training neural networks". It allows you to decompose your Deep Learning pipeline into modular and insightful "Steps". Additionally it has many features for testing and debugging neural nets.

End-to-End Speech Recognition Training with Conformer CTC using PyTorch Lightning⚡

A computer vision pipeline built with PyTorch for advanced NBA analytics. The system uses YOLO for player/ball detection, multi-object tracking, and a zero-shot classifier for team affiliation, and court key point detection to create a tactical top-down view and calculate real-world metrics like speed, distance, and passes.

A ⚡️ Lightning.ai ⚡️ component for train and test data drift detection

Open-source Armenian video dubbing pipeline with ASR, translation, voice cloning, lip-sync, and emotion-aware TTS

Leverage ModernGLiNER's capabilities using LitServe.

Track video objects with YOLOv8 and StrongSORT, using appearance embeddings and MOT metrics for reliable multi-object tracking

Train MedGemma-1.5-4B on MedQA-SWE with LoRA SFT on LightningAI

Leverage GOT-OCR2's optical character recognition capabilities using LitServe.

Leverage TangoFlux's text-to-audio capabilities using LitServe.

Leverage ModernBERT's capabilities using LitServe.

核心網頁瀏覽器模組:提供 個具備安全控制與自動化機制 的輕量化網頁瀏覽器/擷取代 理 (StormCoreBrowser) 專門用於抓取公開氣象與電力 災情網頁。 技術規格:採用 Python 3.11+,結合 requests 與 BeautifulSoup (或可擴充 為 Playwright/ Selenium),內建請求頻率 限制、User-Agent 輪換與資 料防護機制。 開源與安全規範:符合專案規 範,嚴格遵守 Public Layer 邊 界,不碰觸敏感憑證l

Self-hosted text-to-image and image-to-image generation studio built with Stable Diffusion XL, FastAPI, and React -- deployed on a single T4 GPU.

Leverage Gemma 3's capabilities using LitServe.

Leverage VideoLLaMA 3's capabilities using LitServe.

Leverage Reader-LM's capabilities using LitServe.

⚡🗼Liquid AI × WAY × AMD hackathon — 🐈 NyanGuard AI is a local-first moderation demo for audio streams. 📅2026-06-06. It accepts an audio or video file, transcribes speech locally when a local ASR backend is available, replaces toxic language with Nyan Nyan, and returns a playable moderated audio file.

Leverage SigLIP 2's capabilities using LitServe.

Leverage PaliGemma 2's DOCCI fine-tuned variant capabilities using LitServe.