The effect of VQVAE's training data on image generation

FoundationVision / LlamaGen

Autoregressive Model Beats Diffusion: 🦙 Llama for Scalable Image Generation

https://arxiv.org/abs/2406.06525

MIT License

1.33k stars 55 forks source link

Open HalvesChen opened 1 month ago

HalvesChen commented 1 month ago

Hi~ Thank you for your work. I would like to know if you have compared the ablation experiments on the LAION-COCO fine-tuning.