Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

Wonder what the “E” stands for in E2B and E4B models? Look no further! Learn more about these “effective” parameters, Per-Layer Embeddings (PLE), and how they boost a model's power without increasing the actual parameters used during computation in this 2-minute overview.

googlegemma's profile picture

Google Gemma

95,710 subscribers

35,606 Aufrufe • vor 2 Tagen •via X (Twitter)

Wissenschaft & Technologie
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

WHY IS NO ONE TALKING ABOUT THIS?? Gemma 3n model was one of the best surprises for me. The fact that you can run it on edge devices even with just 2GB of RAM is impressive. A few weeks back, I was on holiday and used the Gemini Live feature a lot. But I kept running into issues whenever I was in a place where the network wasn’t reliable. Gemma 3n : > Multimodal: Supports text and image inputs; video/audio in development. > Context Window: Up to 128K tokens (32K for 1B model). > Multilingual: Trained on 140+ languages. > Privacy: Offline, on-device processing for data security. > Model Sizes: E2B (5B parameters, 2GB RAM) and E4B (10–12B parameters, 3GB RAM). Video credit: Google YT
1:34

WHY IS NO ONE TALKING ABOUT THIS?? Gemma 3n model was one of the best surprises for me. The fact that you can run it on edge devices even with just 2GB of RAM is impressive. A few weeks back, I was on holiday and used the Gemini Live feature a lot. But I kept running into issues whenever I was in a place where the network wasn’t reliable. Gemma 3n : > Multimodal: Supports text and image inputs; video/audio in development. > Context Window: Up to 128K tokens (32K for 1B model). > Multilingual: Trained on 140+ languages. > Privacy: Offline, on-device processing for data security. > Model Sizes: E2B (5B parameters, 2GB RAM) and E4B (10–12B parameters, 3GB RAM). Video credit: Google YT

AshutoshShrivastava

210,490 Aufrufe • vor 1 Jahr

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote at #CES2026. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️
0:58

Frontier models that use reasoning to "think" during inference are generating 5X more AI tokens per year. ✨ "Inference is now a thinking process. And in order to teach AI how to think, reinforcement learning and very significant computation was introduced into post-training," said NVIDIA CEO Jensen Huang in a recent keynote at #CES2026. Reinforcement learning is increasing computation demands across all AI scaling laws: pre-training, post-training, and test-time scaling. Learn more about reinforcement learning and AI scaling ➡️

NVIDIA AI Infrastructure

13,329 Aufrufe • vor 6 Monaten

Last week, we launched "Attention in Transformers: Concepts and Code in PyTorch" instructed by Joshua Starmer! In this course, you'll: ✅ Learn how the attention mechanism in LLMs helps convert base token embeddings into rich context-aware embeddings. ✅ Understand the Query, Key, and Value matrices, what they are for, how to produce them, and how to use them in attention. ✅ Learn the difference between self-attention, masked self-attention, and cross-attention, and how multi-head attention scales the algorithm. 🔗 Enroll for free:
1:09

Last week, we launched "Attention in Transformers: Concepts and Code in PyTorch" instructed by Joshua Starmer! In this course, you'll: ✅ Learn how the attention mechanism in LLMs helps convert base token embeddings into rich context-aware embeddings. ✅ Understand the Query, Key, and Value matrices, what they are for, how to produce them, and how to use them in attention. ✅ Learn the difference between self-attention, masked self-attention, and cross-attention, and how multi-head attention scales the algorithm. 🔗 Enroll for free:

DeepLearning.AI

36,832 Aufrufe • vor 1 Jahr

New short course: Attention in Transformers: Concepts and Code in PyTorch. Last week we released a course on how LLM transformers work. This week, go deeper and learn about the technical ideas behind the attention mechanism, and see how to code it in PyTorch. This course is built with Joshua Starmer, Founder and CEO of StatQuest. The attention mechanism was a breakthrough that led to transformers, the architecture powering large language models like ChatGPT. Transformers, introduced in the 2017 paper: "Attention is All You Need" by Viswani and others, took off because of its highly scalable design. In this course, you’ll learn how the attention mechanism, a key element of transformer-based LLMs, works and implement it in PyTorch. You'll develop deep intuition about building reliable, functional, and scalable AI applications. What you will do: - Understand the evolution of the attention mechanism, a key breakthrough that led to transformers. - Learn the relationships between word embeddings, positional embeddings, and attention. - Learn about the Query, Key, and Value matrices, and how to produce and use them in attention. - Walk through the math required to calculate self-attention and masked self-attention to learn why and how they work. - Understand the difference between self-attention and masked self-attention and how one is used in the encoder to build context-aware embeddings and the other is used in the decoder for generative outputs. - Learn the details of the encoder-decoder architecture, cross-attention, and multi-head attention and how they are all incorporated into a transformer. - Use PyTorch to code a class that implements self-attention, masked self-attention, and multi-head attention. There're lots of exciting technical details in this course. Please sign up here:
2:41

New short course: Attention in Transformers: Concepts and Code in PyTorch. Last week we released a course on how LLM transformers work. This week, go deeper and learn about the technical ideas behind the attention mechanism, and see how to code it in PyTorch. This course is built with Joshua Starmer, Founder and CEO of StatQuest. The attention mechanism was a breakthrough that led to transformers, the architecture powering large language models like ChatGPT. Transformers, introduced in the 2017 paper: "Attention is All You Need" by Viswani and others, took off because of its highly scalable design. In this course, you’ll learn how the attention mechanism, a key element of transformer-based LLMs, works and implement it in PyTorch. You'll develop deep intuition about building reliable, functional, and scalable AI applications. What you will do: - Understand the evolution of the attention mechanism, a key breakthrough that led to transformers. - Learn the relationships between word embeddings, positional embeddings, and attention. - Learn about the Query, Key, and Value matrices, and how to produce and use them in attention. - Walk through the math required to calculate self-attention and masked self-attention to learn why and how they work. - Understand the difference between self-attention and masked self-attention and how one is used in the encoder to build context-aware embeddings and the other is used in the decoder for generative outputs. - Learn the details of the encoder-decoder architecture, cross-attention, and multi-head attention and how they are all incorporated into a transformer. - Use PyTorch to code a class that implements self-attention, masked self-attention, and multi-head attention. There're lots of exciting technical details in this course. Please sign up here:

Andrew Ng

132,285 Aufrufe • vor 1 Jahr

This video is alleged to have been posted and later removed by AIPAC. In the video, they reportedly boast about securing $627 million from U.S. taxpayers. Watch the video to learn more about how these funds are used.
1:55

This video is alleged to have been posted and later removed by AIPAC. In the video, they reportedly boast about securing $627 million from U.S. taxpayers. Watch the video to learn more about how these funds are used.

Noah’s Ark 🚢

159,180 Aufrufe • vor 1 Jahr

Like LLMs, autoregressive transformers for action tokens need a reasoning layer to reduce hallucinations and boost reliability. Grounding this layer in the physics of the action space using DVBFs makes for scalable, task-agnostic training—far simpler than creating RL reward functions for each task. Learn more about our novel approach in the video below 👇🏼
1:51

Like LLMs, autoregressive transformers for action tokens need a reasoning layer to reduce hallucinations and boost reliability. Grounding this layer in the physics of the action space using DVBFs makes for scalable, task-agnostic training—far simpler than creating RL reward functions for each task. Learn more about our novel approach in the video below 👇🏼

Sankaet

52,565 Aufrufe • vor 1 Jahr

Henry Kravis' advice for young aspiring investors is not about learning more Excel or building better models. It is about spending time in the actual business to learn the fundamentals At his own firm, KKR, he used to send young people out to work for portfolio companies for a couple of years. Many would come back enlightened by the experience, finally understanding the business This was shocking to him, given these were the same folks giving recommendations to KKR's investment committee and building detailed excel models on the business, without even knowing how they actually function from the inside Must listen
1:05

Henry Kravis' advice for young aspiring investors is not about learning more Excel or building better models. It is about spending time in the actual business to learn the fundamentals At his own firm, KKR, he used to send young people out to work for portfolio companies for a couple of years. Many would come back enlightened by the experience, finally understanding the business This was shocking to him, given these were the same folks giving recommendations to KKR's investment committee and building detailed excel models on the business, without even knowing how they actually function from the inside Must listen

Boring_Business

72,178 Aufrufe • vor 1 Monat

Coach used his internet finger out in public. 😳 After further review, the call stands! 😂 40 bucks a game isn’t worth a grown man screaming in your face. And we wonder why this generation is the way they are. 🤦‍♂️
0:42

Coach used his internet finger out in public. 😳 After further review, the call stands! 😂 40 bucks a game isn’t worth a grown man screaming in your face. And we wonder why this generation is the way they are. 🤦‍♂️

👉M-Û-R-Č-H👈

187,419 Aufrufe • vor 5 Monaten

For those curious about what these Ninja Gaiden Sigma 2 models look like in-game, here's a video showing a quick script I wrote that allows me to toggle them on and off. The more I dig into this, the more I find this approach cool as hell
0:59

For those curious about what these Ninja Gaiden Sigma 2 models look like in-game, here's a video showing a quick script I wrote that allows me to toggle them on and off. The more I dig into this, the more I find this approach cool as hell

Koenji

57,679 Aufrufe • vor 1 Jahr

getting to see the actual conversation happening and learn more about what goes on in situations like this was so cool. shoutout mark for always capturing everything
0:25

getting to see the actual conversation happening and learn more about what goes on in situations like this was so cool. shoutout mark for always capturing everything

autumn ψ

12,934 Aufrufe • vor 10 Monaten

Guys, in the last week, I've heard of 3 of my friends get "#vbridger" on their rigs and it's like, not rly vbridger?? u_u I understand that other riggers do not charge as much as I do for it. It's not easy to rig - but some transparency of what you're actually rigging would be kind to your clients! I want to pls request that if ur offering vbridger - but you only actually rig a restricted amount of the parameters - to state that in ur comm page! My friends have been given rigs where they had paid for "vbridger" and then they use the rigs and they're just utterly disappointed that their mouth doesn't seem to look or function like vbridger. Either because only some vbridger parameters were rigged or they were not put in or calibrated into VTS for the client. Some have never even actually opened the vbridger app to make sure it works! They just rig it in l2d and call it good. VBridger adds like 5 new parameters to the default 2 required for mouth function (4 are shown in the video attached because Jaw Open kinda goes with Mouth Open). If you will not be adding all 5 of these new parameters on your paying client's rig, please be transparent either on your price list or in your comms. QnQ Clients!!: Learn what your vbridger mouth should look like too! Don't be angry at your rigger if maybe it doesn't look like what you expected, but at least make sure you have the parameters that are exclusive to vbridger included in your rig. Just do a nice safety check. The video is so clients can understand what vbridger includes, what the forms look like, and hopefully will know how to make the faces to test the models they get. I'm sorry to cause trouble and more work for riggers, but I do very much care about client transparency and if you feel what you're charging isn't enough to motivate you to rig all of vbridger or learn it, either raise your pricing or state transparency of it on your site! (Or just like, don't offer it maybe qq) #vts #vtubestudio #live2d
0:49

Guys, in the last week, I've heard of 3 of my friends get "#vbridger" on their rigs and it's like, not rly vbridger?? u_u I understand that other riggers do not charge as much as I do for it. It's not easy to rig - but some transparency of what you're actually rigging would be kind to your clients! I want to pls request that if ur offering vbridger - but you only actually rig a restricted amount of the parameters - to state that in ur comm page! My friends have been given rigs where they had paid for "vbridger" and then they use the rigs and they're just utterly disappointed that their mouth doesn't seem to look or function like vbridger. Either because only some vbridger parameters were rigged or they were not put in or calibrated into VTS for the client. Some have never even actually opened the vbridger app to make sure it works! They just rig it in l2d and call it good. VBridger adds like 5 new parameters to the default 2 required for mouth function (4 are shown in the video attached because Jaw Open kinda goes with Mouth Open). If you will not be adding all 5 of these new parameters on your paying client's rig, please be transparent either on your price list or in your comms. QnQ Clients!!: Learn what your vbridger mouth should look like too! Don't be angry at your rigger if maybe it doesn't look like what you expected, but at least make sure you have the parameters that are exclusive to vbridger included in your rig. Just do a nice safety check. The video is so clients can understand what vbridger includes, what the forms look like, and hopefully will know how to make the faces to test the models they get. I'm sorry to cause trouble and more work for riggers, but I do very much care about client transparency and if you feel what you're charging isn't enough to motivate you to rig all of vbridger or learn it, either raise your pricing or state transparency of it on your site! (Or just like, don't offer it maybe qq) #vts #vtubestudio #live2d

Novaj 👑🌸 God Queen

158,007 Aufrufe • vor 3 Jahren

"Every offseason is critical, but we're excited here about the possibilities with the draft capital we now have, taking a hard look at free agency and where we can supplement given the parameters in the league." -Mark Wilf on what makes this a critical offseason
0:53

"Every offseason is critical, but we're excited here about the possibilities with the draft capital we now have, taking a hard look at free agency and where we can supplement given the parameters in the league." -Mark Wilf on what makes this a critical offseason

Minnesota Vikings

72,487 Aufrufe • vor 6 Monaten

So we know Gemma 4 is good at tool calling, but what about web coding? I threw 4 UI screenshots at three Gemma 4 models and said rebuild this, one shot, no hand-holding, just image in, code out. Model lineup: - E4B - 26B A4B (MoE) - 31B Dense (skipped the E2B this round) Let me know which one you think cooked the hardest
0:21

So we know Gemma 4 is good at tool calling, but what about web coding? I threw 4 UI screenshots at three Gemma 4 models and said rebuild this, one shot, no hand-holding, just image in, code out. Model lineup: - E4B - 26B A4B (MoE) - 31B Dense (skipped the E2B this round) Let me know which one you think cooked the hardest

stevibe

124,942 Aufrufe • vor 4 Monaten

Current robotics models really underscore how wildly efficient the human brain is. Robotics models are relatively small (low billions of parameters), are only storing a second of context, and doing 100ms inference steps. By contrast the human brain has 100 trillion synapses (roughly analogous to parameters), processes 24 frames per second, while considering hours of previous context. So how do we expect to get many order-of-magnitude improvements in robotics across all these dimensions (which are trading off against each other in terms inference compute) at the same time? And how do we explain what's going on in the brain? Is it more analogous to a software advantage or a hardware advantage? Full episode with Sergey Levine out now!
1:33

Current robotics models really underscore how wildly efficient the human brain is. Robotics models are relatively small (low billions of parameters), are only storing a second of context, and doing 100ms inference steps. By contrast the human brain has 100 trillion synapses (roughly analogous to parameters), processes 24 frames per second, while considering hours of previous context. So how do we expect to get many order-of-magnitude improvements in robotics across all these dimensions (which are trading off against each other in terms inference compute) at the same time? And how do we explain what's going on in the brain? Is it more analogous to a software advantage or a hardware advantage? Full episode with Sergey Levine out now!

Dwarkesh Patel

87,247 Aufrufe • vor 10 Monaten

Wow, what a vibrant, diverse, and dynamic monarchy...🤭 No wonder they're obsessed with Harry and Meghan. They make these people look dull and boring. Kate, dressed like someone's grandma in what I'm guessing is a homage to the late queen. How touching.#PrincessofWales
0:20

Wow, what a vibrant, diverse, and dynamic monarchy...🤭 No wonder they're obsessed with Harry and Meghan. They make these people look dull and boring. Kate, dressed like someone's grandma in what I'm guessing is a homage to the late queen. How touching.#PrincessofWales

Benjamin Eley Cannell Parody (Fake/Parody)

24,278 Aufrufe • vor 3 Monaten

This was in Nagpur last night. Look how they plan a riot and communal tension meticulously. The images speak for themselves. No mercy for these rioters.
6:49

This was in Nagpur last night. Look how they plan a riot and communal tension meticulously. The images speak for themselves. No mercy for these rioters.

Aditya Raj Kaul

75,502 Aufrufe • vor 1 Jahr

This Earth Day, we’re taking a look at the Library’s incredibly cute pocket globe collection in the Geography and Map Division. 🌎 Learn more about these small treasures:
0:58

This Earth Day, we’re taking a look at the Library’s incredibly cute pocket globe collection in the Geography and Map Division. 🌎 Learn more about these small treasures:

Library of Congress

10,828 Aufrufe • vor 1 Jahr

Ravi River, such a beautiful one that singers used to make it part of their songs, now feels more like a drain! No wonder these rivers take a virat roop during the monsoon. They clean the filth and the filth creators every year, yet people never learn.
1:54

Ravi River, such a beautiful one that singers used to make it part of their songs, now feels more like a drain! No wonder these rivers take a virat roop during the monsoon. They clean the filth and the filth creators every year, yet people never learn.

Nikhil saini

70,180 Aufrufe • vor 1 Monat

🚨 FASCINATING that these senators don't have a CLUE what they are talking about. HASSAN: Scientists wanted to brief you and understand why you unilaterally changed the parameters for giving vaccines, making it possible to make them go off-label! RFK JR: This is CRAZY TALK, you are MAKING THINGS UP! HASSAN: Why was this behind closed doors? RFK JR: It's NOT. The industry makes the studies, but they could not provide us studies saying the COVID vax is effective for healthy kids. HASSAN: When have you produced data to change these parameters? RFK JR: The data is all PUBLIC! You are just MAKING STUFF UP, senator! 🔥🔥
1:19

🚨 FASCINATING that these senators don't have a CLUE what they are talking about. HASSAN: Scientists wanted to brief you and understand why you unilaterally changed the parameters for giving vaccines, making it possible to make them go off-label! RFK JR: This is CRAZY TALK, you are MAKING THINGS UP! HASSAN: Why was this behind closed doors? RFK JR: It's NOT. The industry makes the studies, but they could not provide us studies saying the COVID vax is effective for healthy kids. HASSAN: When have you produced data to change these parameters? RFK JR: The data is all PUBLIC! You are just MAKING STUFF UP, senator! 🔥🔥

Eric Daugherty

38,133 Aufrufe • vor 11 Monaten

What is a prominence or filament eruption? Well, there was a good one early today, 2 Mar, so to learn more about what these are and why they are important, check out this latest SWPC educational video about the latest eruption and how this one is not Earth-directed.
2:54

What is a prominence or filament eruption? Well, there was a good one early today, 2 Mar, so to learn more about what these are and why they are important, check out this latest SWPC educational video about the latest eruption and how this one is not Earth-directed.

NOAA Space Weather Prediction Center

28,333 Aufrufe • vor 5 Monaten