Dog-Breed Text-to-image Flow Matching from Pretraining to Fine-Tuning on 4 Consumer GPUs
We develop MiniDog, a framework that enables pretraining diffusion transformers in 55 minutes and fine-tuning in another 15 minutes on four 3090 GPUs.
ProofPaper ↗
Key points
- We generally perceive that training text-to-image diffusion models requires heavy compute: it is true if we aim at a comprehensive frontier model.
- For data, we use 26k dog images from ImageNet and caption them for pretraining and use 2k images generated by FLUX.1-schnell for fine-tuning.
- We precompute the image and text features and cache them for efficiency.
- For architecture, we use E2E-INVAE for its semantic latent space, a small transformer with 22M parameters, a small Qwen model for text encoding, and in-context conditioning.
Sources (1)
- [1]Dog-Breed Text-to-image Flow Matching from Pretraining to Fine-Tuning on 4 Consumer GPUsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 05:59 PM
We develop MiniDog, a framework that enables pretraining diffusion transformers in 55 minutes and fine-tuning in another 15 minutes on four 3090 GPUs.
We generally perceive that training text-to-image diffusion models requires heavy compute: it is true if we aim at a comprehensive frontier model.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 13, 2026PixelMoE: Scaling Pixel-Space Diffusion with LLM-Style Mixture-of-Experts
- Oct 13, 2026Run fast, private agents on-device with LFM2.5-24B-A2B
- Oct 13, 2026Gemini 4 Argon: Built to solve your hardest problems
- Oct 13, 2026black-forest-labs/flux-3-image-klein-comfy
- Oct 13, 2026LoopREG: Representation Supervision Enables Test-Time Recurrent Depth in Diffusion Transformers
- Oct 13, 2026PrT-Flow: Probability-Tractable Flow Matching for Reinforcement Learning