Ideogram 4.0 发布:9.3B Diffusion Transformer 可在 24GB 显卡运行

Today we published a technical blog post about Ideogram 4.0 — our goal is to enable more innovation ...

精选理由

Ideogram 4.0 把 9.3B 参数的大模型压缩到 24GB 显卡就能跑,做图像生成或 AI 绘画的开发者可以直接在本地尝试,不用再依赖昂贵云端算力。

AI 摘要

Ideogram 4.0 是一个 9.3B 参数的 Diffusion Transformer 模型,从零开始训练,并搭配了一个冻结的 8B 视觉语言模型作为文本编码器。该模型通过 nf4 量化检查点可在 24GB 消费级 GPU 上运行,大幅降低了硬件门槛。团队表示目标是推动更多创新和创造力。这一发布意味着高质量图像生成模型向个人开发者和小团队开放了可能性。

原文 · Ideogram

Today we published a technical blog post about Ideogram 4.0 — our goal is to enable more innovation ...

Today we published a technical blog post about Ideogram 4.0 — our goal is to enable more innovation and creativity. It's a 9.3B Diffusion Transformer trained from scratch, paired with a frozen 8B VLM as text encoder. The nf4 checkpoint runs on a 24GB consumer GPU. Thread 🧵 💬 12 🔄 24 ❤️ 115 👀 4997 📊 40 ⚡