# jy0205/Pyramid-Flow

[ICLR 2025] Pyramidal Flow Matching for Efficient Video Generative Modeling

Repository: https://github.com/jy0205/Pyramid-Flow
Canonical: https://ross.abutalabs.com/products/pyramid-flow
Homepage: https://pyramid-flow.github.io/
Language: Python
License: MIT
License Family: permissive
Topics: diffusion-models, flow-matching, video-generation
Last push: 2024-12-21T19:11:41+00:00

## Health v2 (maintenance only)
Score: 22/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 49
- inputs: {"age_days": 696, "days_push": 620, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3208, forks 304 (observed 2026-08-28T04:07:48.674357+00:00)

## What it is
Pyramid Flow is the official PyTorch implementation of a training-efficient autoregressive video generation model based on pyramidal flow matching, capable of generating 10-second 768p 24fps videos from text or images. It includes inference code with multi-GPU and CPU-offloading support, plus training code for the VAE and DiT finetuning.

## Use cases
- generate videos from text prompts
- image-to-video generation
- generate 10 second 768p videos on limited GPU memory
- finetune a video diffusion model on custom data
- train a video VAE
- run video generation across multiple GPUs

## When to choose
- you need open-source text-to-video or image-to-video generation with permissive MIT licensing
- you have limited GPU memory thanks to CPU offloading support
- you want to finetune or retrain video generation components yourself

## When to avoid
- you need production-grade video generation with commercial support
- you have no GPU available
- you need real-time video generation

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, video-processing, image-processing, llm-inference
- domain: deep-learning, image-processing, artificial-intelligence
- platform: python
- tags: video-generation, flow-matching, diffusion-models, text-to-video, image-to-video, autoregressive, research-code, iclr-2025, video, gpu, linux, docker

## Member repositories
- jy0205/Pyramid-Flow (main) score 22

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:48.674357+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:24:45.336778+00:00, confidence not recorded.
  - readme: https://github.com/jy0205/Pyramid-Flow (fetched 2026-08-28T04:07:48.674357+00:00, sha 1ee3439be823)
  - homepage: https://pyramid-flow.github.io/ (fetched 2026-08-29T09:38:19.742061+00:00, sha 419c60984f3d)
- Data as of 2026-08-30T08:39:29.467469+00:00.