# AlibabaResearch/SparkDiffusion

> A video-generation acceleration framework for Diffusion Transformer models, combining sparse low-rank attention (RoLa), few-step distillation (CrossDistill) and custom operators to deliver 200×+ inference speedups for Wan 2.1/2.2.

- Magnitude: 1.3 out of 10 — Early signal
- Stars: 89 total · +0 stars today, ≈ 4 by evening
- Star trust: star growth looks organic
- Category: Generative media · Language: Python · License: Apache-2.0 · Created: 2026-09-18 · Last push: 2026-09-23
- GitHub: https://github.com/AlibabaResearch/SparkDiffusion · Page: https://gitnova.dev/en/r/AlibabaResearch/SparkDiffusion

## Useful for

- Speed up Wan 2.1/2.2 inference via sparse finetuning and distillation
- Finetune a DiT model with RoLa sparse attention on a custom dataset
- Run single-case T2V or I2V generation through the provided scripts

## Why it’s here

- 0 stars so far today, about 4 expected by the end of the day.
- The repository is 6 days old and already has 89 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Top new repositories this week: #185.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 2
- Issues and pull requests: 1
- Watchers: 1
- Average over the last week: 13 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 1

## Stars per day, last 12 days (oldest → newest, today is partial)

2026-09-13 … 2026-09-24: 0, 0, 0, 0, 0, 0, 0, 0, 1, 83, 5, 0

## Spotted in now

- Top new repositories this week: #185

## Similar by description

1. **hao-ai-lab/FastVideo** — 0.9 · Steady · Generative media · Python · +0 stars today, ≈ 2 by evening
   A framework for accelerated video generation: post-training (distillation, LoRA, sparse attention) and fast inference of diffusion video models on GPU and Apple Silicon.
   Full card: https://gitnova.dev/en/r/hao-ai-lab/FastVideo.md
2. **ml-explore/mlx** — 1.5 · Steady · Language models · C++ · +0 stars today, ≈ 6 by evening
   MLX is an array and machine learning framework from Apple, optimized for Apple silicon with unified memory and lazy computation. Its Python, C++, C, and Swift APIs mirror NumPy and PyTorch, simplifying model training and deployment.
   Full card: https://gitnova.dev/en/r/ml-explore/mlx.md
3. **microsoft/onnxruntime** — 1.6 · Steady · Language models · C++ · +0 stars today, ≈ 8 by evening
   Cross-platform accelerator for ML inference and training. Runs models from PyTorch, TensorFlow, scikit-learn and others via the ONNX format with graph optimizations and hardware acceleration.
   Full card: https://gitnova.dev/en/r/microsoft/onnxruntime.md
4. **Tencent-Hunyuan/Simple-Attention-Sparsification** — 0.0 · Steady · Language models · Python · +0 stars today, ≈ 0 by evening
   SAS is a gated sparse-attention mechanism that learns per-query which context blocks to attend to, optimizing context ranking end-to-end with the language modeling loss. The repo provides the paper's training and evaluation code.
   Full card: https://gitnova.dev/en/r/Tencent-Hunyuan/Simple-Attention-Sparsification.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-09-24 02:50 UTC, updated every 30 minutes.