huahuaapi

便宜中转站

[huahuaapi @slgthinking]

用ai的核心，就是能够让ai不断的去解决更复杂的事情。复杂=长线程+多线程；长可以定义为，不断前进优化。多可以定义为，同时进行。

[huahuaapi @slgthinking]

KiMi

[huahuaapi @slgthinking]

@lidangzzz 根本管不住了。而且以后这些新能容也不一定归广电关，工信部还想管呢？

[huahuaapi @slgthinking]

@joshesye gpt-image2 也行的

[huahuaapi @slgthinking]

提示词结构： 1.时间环境（开头定调） 2.主体 + 服装 + 动作 + 场景 3.道具细节（增加真实感） 4.色调 + 氛围（风格定调） 5.光影美学（宋式美学） 6.动态细节（氛围感点睛） 7.构图层次（前景 + 背景） 8.服装详写（分层描述，保证服装准确） 9.人物状态（表情 + 发型 + 姿态） 10.肤质 + 焦点

[huahuaapi @slgthinking]

中式美学的提示词 ，给你找到了。 评论我贴上母体提示词，你们拿去给你们的ai进行扩展裂变吧。

[huahuaapi @slgthinking]

@yang0001 看上去 很流畅

[huahuaapi @slgthinking]

pretty awesome

[leo 🐾 @synthwavedd]

GPT-Image 2.5 is pretty damn good

[huahuaapi @slgthinking]

@Tz_2022 我们和大语言模型 有什么区别

[huahuaapi @slgthinking]

@traderxiaoxia 故事讲得真好， 有人信吗？

[huahuaapi @slgthinking]

@cholf5 说实话，这个世界有太多娱乐活动比游戏好玩了。

[huahuaapi @slgthinking]

@cholf5 你没看他说的是 这个行业 就业容量太小了。19分化严重，大部分从业者都是步兵炮灰。第九艺术需要艺术家，而不是马龙。

[huahuaapi @slgthinking]

@Formulasearch 便宜啊。自己搭建

[huahuaapi @slgthinking]

@caiyue5 resend

[huahuaapi @slgthinking]

@caizhenghai 没有信仰的 段。

[huahuaapi @slgthinking]

太疯狂了，直接公开大模型训练过程。535B参数

[Percy Liang @percyliang]

🚢 Marin 535B-A23B started training this week! As usual, the whole process is open.
Voyage plan: pretraining (80%) + midtraining (20%) on 18.75T tokens on 11 x GB200 NVL72 for ~3 months (2.7e24 FLOPs). Post-training will follow.
Before kicking off the run, we trained a 4-rung scaling ladder from 1.6B-A61M (48B tokens) to 27.7B-A1.2B (926B tokens) to debug issues, and to make a forecast of our hero run. This is by far our biggest run, so definitely expecting the unexpected.

[Percy Liang @percyliang]

🚢 Marin 535B-A23B started training this week! As usual, the whole process is open.
Voyage plan: pretraining (80%) + midtraining (20%) on 18.75T tokens on 11 x GB200 NVL72 for ~3 months (2.7e24 FLOPs). Post-training will follow.
Before kicking off the run, we trained a 4-rung scaling ladder from 1.6B-A61M (48B tokens) to 27.7B-A1.2B (926B tokens) to debug issues, and to make a forecast of our hero run. This is by far our biggest run, so definitely expecting the unexpected.

[huahuaapi @slgthinking]

@foxshuo 苟有新民，何患无新国

[huahuaapi @slgthinking]

大师出手就是不一样。直接把codex cli 速度提高25倍。

[Charlie Marsh @charliermarsh]

In the latest Codex CLI release, I redid the lifecycle to make `codex` startup instant.

It's now ~25x faster and immediately responsive.