正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Past Model - Memory

goth music videos

226,097 subscribers

13,069 次观看 • 5 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Memory from the past!!! 🤣

Memory from the past!!! 🤣

Frank Khalid OBE

74,908 次观看 • 1 年前

Nothing slips past Sachin Tendulkar - elephant-level memory 🤯

Nothing slips past Sachin Tendulkar - elephant-level memory 🤯

Cricketopia

13,469 次观看 • 3 个月前

We're releasing Letta Code, a memory-first coding agent - open source (apache 2.0) - model agnostic - portable agent learning and memory

We're releasing Letta Code, a memory-first coding agent - open source (apache 2.0) - model agnostic - portable agent learning and memory

Letta

372,778 次观看 • 8 个月前

This Amphibia moment has been a core memory for me for the past 4 years

This Amphibia moment has been a core memory for me for the past 4 years

Marngel 🐸🦉👻

36,448 次观看 • 1 年前

Lots happened in the past year with 1 model.

Lots happened in the past year with 1 model.

𝗬unie 🦋💀

135,051 次观看 • 1 年前

A memory of past but a lesson for the future, for the invaders US army vehic in Afghanistan

A memory of past but a lesson for the future, for the invaders US army vehic in Afghanistan

W.A. Mubariz - وکیل احمد مبارز

35,192 次观看 • 5 天前

Does ASICS Thailand know who their latest model is and her past?

Does ASICS Thailand know who their latest model is and her past?

Chrome Crumpet

13,740 次观看 • 2 年前

We've made it past Year 2 for Street Fighter 6! 🥳 Let's take a trip down memory lane with some stats, our past trailers, and creative works by Capcom Creators.

We've made it past Year 2 for Street Fighter 6! 🥳 Let's take a trip down memory lane with some stats, our past trailers, and creative works by Capcom Creators.

Street Fighter 6

334,260 次观看 • 1 年前

researchers gave a tiny local model human-style memory and its context limit basically stopped existing a team from MBZUAI, Princeton and Weizmann took a 1B model and rebuilt how it reads. instead of attending to everything at once, the model reads in 1,024 token chunks and passes the important stuff forward through an associative memory, the same way you carry the plot of a book between chapters without rereading them. the design mirrors human memory on purpose. full attention inside a chunk works as short-term memory. the module that carries information between chunks works as long-term memory. they even trained it like a person, starting with short easy texts and raising the difficulty gradually, because memory thrown into the deep end learns nothing. the numbers back it up. the normal model burns 40GB of GPU memory on a long document and collapses hard past its limit, dropping from 0.86 to 0.32 accuracy. the memory version holds 0.71 at double that length while using a flat 12GB no matter how long the input gets. it also needs about 30% fewer FLOPs. the part i keep thinking about is that nobody scaled anything here. they didn't build a bigger model, didn't stretch the window, didn't add compute. they looked at how a brain handles a long day and copied the architecture. a model small enough to run on a consumer gpu now survives documents its own architecture used to choke on. we keep treating intelligence as a compute problem. sometimes it's a memory problem.

researchers gave a tiny local model human-style memory and its context limit basically stopped existing a team from MBZUAI, Princeton and Weizmann took a 1B model and rebuilt how it reads. instead of attending to everything at once, the model reads in 1,024 token chunks and passes the important stuff forward through an associative memory, the same way you carry the plot of a book between chapters without rereading them. the design mirrors human memory on purpose. full attention inside a chunk works as short-term memory. the module that carries information between chunks works as long-term memory. they even trained it like a person, starting with short easy texts and raising the difficulty gradually, because memory thrown into the deep end learns nothing. the numbers back it up. the normal model burns 40GB of GPU memory on a long document and collapses hard past its limit, dropping from 0.86 to 0.32 accuracy. the memory version holds 0.71 at double that length while using a flat 12GB no matter how long the input gets. it also needs about 30% fewer FLOPs. the part i keep thinking about is that nobody scaled anything here. they didn't build a bigger model, didn't stretch the window, didn't add compute. they looked at how a brain handles a long day and copied the architecture. a model small enough to run on a consumer gpu now survives documents its own architecture used to choke on. we keep treating intelligence as a compute problem. sometimes it's a memory problem.

Alex Veremeyenko

16,147 次观看 • 21 天前

Memory-V2V by Adobe: diffusion model with memory for consistent iterative editing; - ReCamMaster+LucyEdit; - reduces FLOPs by 30-90%; outperforming FIFO-Diffusion. high fidelity across long durations like consistent style transfer

Memory-V2V by Adobe: diffusion model with memory for consistent iterative editing; - ReCamMaster+LucyEdit; - reduces FLOPs by 30-90%; outperforming FIFO-Diffusion. high fidelity across long durations like consistent style transfer

Wildminder

10,248 次观看 • 6 个月前

alexandr wang talks about my favorite topic: model welfare full episode on the core memory pod this week

alexandr wang talks about my favorite topic: model welfare full episode on the core memory pod this week

Kylie Robison

43,093 次观看 • 3 个月前

Past Lives (2023) is heartbreaking not through tragedy, but through time, memory, and the lives we don’t get to live.

Past Lives (2023) is heartbreaking not through tragedy, but through time, memory, and the lives we don’t get to live.

cinesthetic.

128,549 次观看 • 1 年前

We've been testing Memory (short-term and long-term) on Perplexity for a while. The results are great, and we are rolling it out widely. You can ask personalized questions, questions about past chats, and use any model or search mode with personal context (both apps and web).

We've been testing Memory (short-term and long-term) on Perplexity for a while. The results are great, and we are rolling it out widely. You can ask personalized questions, questions about past chats, and use any model or search mode with personal context (both apps and web).

Aravind Srinivas

96,497 次观看 • 8 个月前

That one time I drove the Model 3 Performance 9km past empty 🤦😅

That one time I drove the Model 3 Performance 9km past empty 🤦😅

Oli

45,142 次观看 • 2 个月前

Automatically create Agent Skills from past sessions. Gemini CLI can now comb through past session data and suggest new skills based on past patterns of things you do frequently! Helps the agent self-improve. 🧠 Enable Auto Memory in /settings to try it out.

Automatically create Agent Skills from past sessions. Gemini CLI can now comb through past session data and suggest new skills based on past patterns of things you do frequently! Helps the agent self-improve. 🧠 Enable Auto Memory in /settings to try it out.

Jack Wotherspoon

24,793 次观看 • 3 个月前

.Savannah Chrisley reacts to the new docuseries exploring past controversies on 'America's Next Top Model.'.

.Savannah Chrisley reacts to the new docuseries exploring past controversies on 'America's Next Top Model.'.

The View

19,576 次观看 • 5 个月前

"Having a short memory would be really good in these situations, whatever's past is in the past, forget about it and try to move forward." Juan Soto on what's most important in the playoffs:

"Having a short memory would be really good in these situations, whatever's past is in the past, forget about it and try to move forward." Juan Soto on what's most important in the playoffs:

Yankees Videos

28,804 次观看 • 1 年前

1-bit Bonsai 8B running locally on an M4 Pro (MLX) alongside a standard 16-bit 8B model. Same class of model, very different deployment profile: far lower memory use and substantially higher throughput.

PrismML

93,021 次观看 • 4 个月前