Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

YAGAOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Caedrel's profile picture

Caedrel

526,425 subscribers

418,452 görüntüleme • 2 yıl önce •via X (Twitter)

KomediOyun
LIVE

Anya Rossi• Live Now

Private livecam show

0 Yorum

Yorum bulunmuyor

Orijinal gönderinin yorumları burada görünecek

Benzer Videolar

actual good one FNATIC
0:14

actual good one FNATIC

Caedrel

952,706 görüntüleme • 2 yıl önce

how the fuck did no one come up with this before
0:17

how the fuck did no one come up with this before

Caedrel

4,266,835 görüntüleme • 3 yıl önce

nah the LPL account is out of control
0:15

nah the LPL account is out of control

Caedrel

566,052 görüntüleme • 2 yıl önce

THIS IS IT - T1 & FAKER TAKE ON THE ENTIRE LPL IT BEGINS NOWWWWWW
0:14

THIS IS IT - T1 & FAKER TAKE ON THE ENTIRE LPL IT BEGINS NOWWWWWW

Caedrel

520,597 görüntüleme • 3 yıl önce

this blastcone instantly respawning made it so G2 could chase and get an extra 3 kills which affected the game quite alot bug or no bug?
0:15

this blastcone instantly respawning made it so G2 could chase and get an extra 3 kills which affected the game quite alot bug or no bug?

Caedrel

728,446 görüntüleme • 2 yıl önce

TABE VS REF HAHHA LEGEND
0:17

TABE VS REF HAHHA LEGEND

Caedrel

546,201 görüntüleme • 3 yıl önce

If you make games without AI, you're wasting your time. Here's how I build Ladder Knight. I leave a note right on the map and keep working. A minute later, the change is in the game. From small things like basic decor or a slightly higher jump, to entirely new mechanics like a block that moves on any path and connects to any buttons. Each one is just a note. I test it right in the editor. The editor is made completely by AI too. Even this video was made with a feature AI added, because I didn't want to waste time cropping it. Years ago I spent more time making editors than making games (I worked on mobile games in 2020). Now the tools grow with the game. If you ever dreamed of making your own game, there was never a better time to start.
0:16

If you make games without AI, you're wasting your time. Here's how I build Ladder Knight. I leave a note right on the map and keep working. A minute later, the change is in the game. From small things like basic decor or a slightly higher jump, to entirely new mechanics like a block that moves on any path and connects to any buttons. Each one is just a note. I test it right in the editor. The editor is made completely by AI too. Even this video was made with a feature AI added, because I didn't want to waste time cropping it. Years ago I spent more time making editors than making games (I worked on mobile games in 2020). Now the tools grow with the game. If you ever dreamed of making your own game, there was never a better time to start.

Alex Kunchevsky

58,148 görüntüleme • 1 gün önce

happy day 🎀
0:11

happy day 🎀

sam

761,225 görüntüleme • 2 gün önce

Santi And they tried defending all this😭😭😭
0:13

Santi And they tried defending all this😭😭😭

GWN

52,125 görüntüleme • 4 ay önce

jinny my rap god! SECRET NUMBER (시크릿넘버) #SECRET_NUMBER #CHICAGO #SECRETNUMBER_UNLOCK #SECRETNUMBER_US_TOUR
0:14

jinny my rap god! SECRET NUMBER (시크릿넘버) #SECRET_NUMBER #CHICAGO #SECRETNUMBER_UNLOCK #SECRETNUMBER_US_TOUR

em

23,354 görüntüleme • 2 yıl önce

NEW FANCAM OF TAEHYUNG PULLING JUNGKOOK OVER TO DANCE TOGETHER 🤍 THEY’RE ADORABLE
0:18

NEW FANCAM OF TAEHYUNG PULLING JUNGKOOK OVER TO DANCE TOGETHER 🤍 THEY’RE ADORABLE

taekook 𑣲𝒌⁷

172,807 görüntüleme • 1 gün önce

If you thought the Gemma 4 31B (dense) model was fast, sit down. I just benched the updated Gemma 4 26B A4B MoE on a single RTX 4090 (24 GB VRAM) 9,200 t/s prefill. 160 t/s decode. 250,000 context window. All on a single consumer RTX 4090. The numbers are completely unhinged. The 31B is a dense behemoth. But the 26B is a Mixture of Experts (MoE), specifically an Active 4 Billion (A4B). It holds 26B parameters of knowledge but only activates 4B per token. Because its inference memory footprint is so light, I didn’t even need KV cache quantization to hit a quarter million context. Compiled the latest llama.cpp from source on Ubuntu 22 (CUDA 13). Fed it a 28k token prompt, and manually cranked the batch sizes (-b 2048 -ub 2048) to absolutely redline the Tensor Cores. Here is the benchmarking breakdown: # 1. The Baseline (No MTP) Even without speculative decoding, the A4B architecture flies. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (21.5 GB VRAM) Prefill: 9,200 t/s (Absurd) Decode: 124 t/s # 2. The MTP Overdrive Injected the new MTP draft model to enable Speculative Decoding. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --spec-type draft-mtp --spec-draft-model mtp-gemma-4-26B-A4B-it.gguf --spec-draft-n-max 4 --spec-draft-p-min 0.7 -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (22.96 GB VRAM) Prefill: 7,054 t/s (MTP draft overhead slightly caps prefill) Decode: 156 t/s # The Agentic Architecture Insight Why does this matter? Because you can now build a killer local agentic loop on a consumer desktop. Use the 31B dense model (from the previous post) as your heavy, deliberate Orchestrator / Verifier / Planner. Pass the actual execution tasks to this 26B MoE. At 160 t/s, this MoE can chew through code generation, tool calling, and massive RAG document retrieval over a 250k context window almost instantly, drastically speeding up your agentic loop. If you own a single RTX 3090 or 4090 and haven't tried this specific stack yet, you need to pull these latest updates and run it. Local inference just leveled up. Hugging Face links to the Unsloth 26B QAT quants and MTP drafters are in the replies. performance graphs also available in the replies.
0:17

If you thought the Gemma 4 31B (dense) model was fast, sit down. I just benched the updated Gemma 4 26B A4B MoE on a single RTX 4090 (24 GB VRAM) 9,200 t/s prefill. 160 t/s decode. 250,000 context window. All on a single consumer RTX 4090. The numbers are completely unhinged. The 31B is a dense behemoth. But the 26B is a Mixture of Experts (MoE), specifically an Active 4 Billion (A4B). It holds 26B parameters of knowledge but only activates 4B per token. Because its inference memory footprint is so light, I didn’t even need KV cache quantization to hit a quarter million context. Compiled the latest llama.cpp from source on Ubuntu 22 (CUDA 13). Fed it a 28k token prompt, and manually cranked the batch sizes (-b 2048 -ub 2048) to absolutely redline the Tensor Cores. Here is the benchmarking breakdown: # 1. The Baseline (No MTP) Even without speculative decoding, the A4B architecture flies. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (21.5 GB VRAM) Prefill: 9,200 t/s (Absurd) Decode: 124 t/s # 2. The MTP Overdrive Injected the new MTP draft model to enable Speculative Decoding. llama.cpp flags: ./build/bin/llama-server -m gemma-4-26B-A4B-it-qat-UD-Q4_K_XL.gguf --spec-type draft-mtp --spec-draft-model mtp-gemma-4-26B-A4B-it.gguf --spec-draft-n-max 4 --spec-draft-p-min 0.7 -c 250000 -ngl 99 -fa on -b 2048 -ub 2048 --port 8080 -v Context Ceiling: 250,000 tokens (22.96 GB VRAM) Prefill: 7,054 t/s (MTP draft overhead slightly caps prefill) Decode: 156 t/s # The Agentic Architecture Insight Why does this matter? Because you can now build a killer local agentic loop on a consumer desktop. Use the 31B dense model (from the previous post) as your heavy, deliberate Orchestrator / Verifier / Planner. Pass the actual execution tasks to this 26B MoE. At 160 t/s, this MoE can chew through code generation, tool calling, and massive RAG document retrieval over a 250k context window almost instantly, drastically speeding up your agentic loop. If you own a single RTX 3090 or 4090 and haven't tried this specific stack yet, you need to pull these latest updates and run it. Local inference just leveled up. Hugging Face links to the Unsloth 26B QAT quants and MTP drafters are in the replies. performance graphs also available in the replies.

Alok

40,993 görüntüleme • 3 ay önce

Mum says a mole on the chest is a sign of true love
0:16

Mum says a mole on the chest is a sign of true love

Vandana

87,942 görüntüleme • 17 gün önce

Over two weeks, Prime Agent orchestrated a swarm of over 2,000 agents to rewrite itself in Rust. With 10,000+ sandboxes, 200B+ GLM-5.3 tokens, and 16,000 agent-to-agent messages, this was our largest test yet of Prime Agent’s multi-agent capabilities and our infrastructure powering these agent swarms. Prime Agent now reaches usable input ~13× faster and uses 83% less startup memory than before.
0:15

Over two weeks, Prime Agent orchestrated a swarm of over 2,000 agents to rewrite itself in Rust. With 10,000+ sandboxes, 200B+ GLM-5.3 tokens, and 16,000 agent-to-agent messages, this was our largest test yet of Prime Agent’s multi-agent capabilities and our infrastructure powering these agent swarms. Prime Agent now reaches usable input ~13× faster and uses 83% less startup memory than before.

Prime Intellect

557,377 görüntüleme • 20 gün önce

10/10 would recommend this link in bio hehe
0:16

10/10 would recommend this link in bio hehe

Savanah Nickole

20,379 görüntüleme • 4 ay önce

class dismissed
0:11

class dismissed

Mat 🖤

66,244 görüntüleme • 2 ay önce

When mom saw you masturbating yesterday night, but she didn't hesitate to stare at you. And you have no idea what mom saw yesterday. #momson #bustymom #desimom #sexysaree #mominsaree #motherson
0:11

When mom saw you masturbating yesterday night, but she didn't hesitate to stare at you. And you have no idea what mom saw yesterday. #momson #bustymom #desimom #sexysaree #mominsaree #motherson

MommiesBoy18

11,144 görüntüleme • 2 ay önce

Turns out the future does have flying cars
0:15

Turns out the future does have flying cars

Elon Musk

79,590,281 görüntüleme • 14 gün önce

CLOCK IT HARRY😭😭
0:10

CLOCK IT HARRY😭😭

bridget semi inactive facts

204,060 görüntüleme • 1 yıl önce

also Ahmed Irtiza why are you lurking on his ig??? girl you're a weirdo
0:15

also Ahmed Irtiza why are you lurking on his ig??? girl you're a weirdo

mel purnell ✨

29,073 görüntüleme • 2 ay önce