本周大事记
Stable Diffusion XL发布
ControlNet 1.1 发布
databricks 发布Dolly 2.0
Free Dolly: Introducing the World's First Truly Open Instruction-Tuned LLM
微软发布DeepSpeed-Chat
DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales
最新技术:
RedPajama, a project to create leading open-source models, starts by reproducing LLaMA training dataset of over 1.2 trillion tokens
主页: www.together.xyz
Inpaint Anything: Segment Anything Meets Image Inpainting
论文: arxiv.org
代码:github.com
MiniGPT-4
Delta Denoising Score
论文: arxiv.org
Power-seeking can be probable and predictive for trained agents
论文:arxiv.org
Segment Everything Everywhere All at Once
论文: arxiv.org/abs/2304.06718
Expressive Text-to-Image Generation with Rich Text
论文: arxiv.org
A new, unique AI dataset for animating amateur drawings
Training Large Language Models Efficiently with Sparsity and Dataflow
论文: arxiv.org
DreamPose: Fashion Image-to-Video Synthesis via Stable Diffusion
论文: arxiv.org
Toxicity in ChatGPT: Analyzing Persona-assigned Language Models
论文: arxiv.org
Teaching Large Language Models to Self-Debug
论文: arxiv.org
paella: 一个新的文生图的模型
主页:laion.ai
Building LLM applications for production
多个ChatGPT合作完成指定任务,迷你AGI控制世界要来了?
课程:
开源图像模型Stable Diffusion入门手册
使用 LoRA 和 Hugging Face 高效训练大语言模型
商业:
当前炼制“大语言模型”的两个现象
一张估值20亿的“笑脸”,正在拆掉OpenAI的围墙
马斯克加入AIGC大战!火速抢购1万张GPU,挖Deepmind墙角
谷歌研究科学家:ChatGPT秘密武器的演进与局限
国家互联网信息办公室关于《生成式人工智能服务管理办法(征求意见稿)》公开征求意见的通知
案例:
3A将贵到做不起必须用AI!15分钟速成场景的Promethean AI又是啥?
chatGPT 插件例子:weathergpt
揭秘 Auto-GPT 喧嚣背后的残酷真相