LogoThread Easy
  • 発見
  • スレッド作成
LogoThread Easy

Twitter スレッドの万能パートナー

© 2025 Thread Easy All Rights Reserved.

探索

Newest first — browse tweet threads

Keep on to blur preview images; turn off to show them clearly

Russia will not strike China in response to its retaliatory nuclear strike after 3GD. We are very likely to sit that out rubbing our hands, or (unlikely, depends on Genshtab 4D thinking) opportunistically lob a few nukes at local NATO outposts. Consider this a warning.

Russia will not strike China in response to its retaliatory nuclear strike after 3GD. We are very likely to sit that out rubbing our hands, or (unlikely, depends on Genshtab 4D thinking) opportunistically lob a few nukes at local NATO outposts. Consider this a warning.

We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1

avatar for Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Thu Dec 11 00:10:33
RT @pmddomingos: ICML and NeurIPS should merge into a single conference called ICPS (International Conference on Plagiarizing Schmidhuber).

RT @pmddomingos: ICML and NeurIPS should merge into a single conference called ICPS (International Conference on Plagiarizing Schmidhuber).

Building Beneficial AGI - CEO @asi_alliance @singularitynet, @true_agi , Interim CEO @Singularity_Fi, @SophiaVerse_AI, Chair @opencog @HumanityPlus @iCog_Labs

avatar for Ben Goertzel
Ben Goertzel
Thu Dec 11 00:10:00
unsloth 又发力了, 3GB 显存训练Qwen3-4b unsloth 最新的更新提升巨大, 简单来说, 他们把原本Q和K各自需要的2个Triton内核合并为1个，并支持可变长度RoPE, 这样可以很大程度上节省显存和提升计算速度(原本Q和K需要两个Triton Kernel计算，现在合并为了一个), 在长上下文训练上实现了2.3x的训练速度加速. 另外还支持了int64索引, 因为原来的int32索引在500K这种超大上下文训练中会导致CUDA越界错误, 于是换了更大的精度来避免越界, 这样就支持更大上下文了.

unsloth 又发力了, 3GB 显存训练Qwen3-4b unsloth 最新的更新提升巨大, 简单来说, 他们把原本Q和K各自需要的2个Triton内核合并为1个，并支持可变长度RoPE, 这样可以很大程度上节省显存和提升计算速度(原本Q和K需要两个Triton Kernel计算，现在合并为了一个), 在长上下文训练上实现了2.3x的训练速度加速. 另外还支持了int64索引, 因为原来的int32索引在500K这种超大上下文训练中会导致CUDA越界错误, 于是换了更大的精度来避免越界, 这样就支持更大上下文了.

实现细节/1

avatar for karminski-牙医
karminski-牙医
Thu Dec 11 00:09:53
This sperm donor thing will lead to massive issues in Europe. The Incest will be insane in the near future

This sperm donor thing will lead to massive issues in Europe. The Incest will be insane in the near future

Founder | Author | Speaker Building @beltstripe. Healtech/EdTech/Agric I'm Not The Man Of Your Dreams. Your Imagination Wasn't This Great.

avatar for Sani Yusuf
Sani Yusuf
Thu Dec 11 00:06:43
RT @shao__meng: 软件工程：效率与效能 来自 @googledevs 最新视频，由 @addyosmani 主讲，深入的分析了软件工程中，效率和效能的关系，这也是我们职业发展中必须明白和解决的一个关键问题！ 核心主题：效率vs. 效能 · 效率 (Effici…

RT @shao__meng: 软件工程：效率与效能 来自 @googledevs 最新视频，由 @addyosmani 主讲，深入的分析了软件工程中，效率和效能的关系，这也是我们职业发展中必须明白和解决的一个关键问题！ 核心主题：效率vs. 效能 · 效率 (Effici…

邵猛，中年失业程序员 😂 专注 - Context Engineering, AI Agents. 分享 - AI papers, apps and OSS. ex Microsoft MVP 合作 - 私信/邮箱：shaomeng@outlook.com 📢 公众号/小红书: AI 启蒙小伙伴

avatar for meng shao
meng shao
Thu Dec 11 00:06:04
China can also do 4 million drones. Except those will be ones fit to combat American Navy. Yes, in principle they could make hundreds of millions of light FPVs but the world is not an infinite Tanar'ri-Baatezu Slav War. Thankfully. It’s pretty sad how Americans don’t extrapolate

China can also do 4 million drones. Except those will be ones fit to combat American Navy. Yes, in principle they could make hundreds of millions of light FPVs but the world is not an infinite Tanar'ri-Baatezu Slav War. Thankfully. It’s pretty sad how Americans don’t extrapolate

We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1

avatar for Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)
Thu Dec 11 00:02:26
  • Previous
  • 1
  • More pages
  • 1047
  • 1048
  • 1049
  • More pages
  • 5634
  • Next