Loading video...

Video Failed to Load

Go Home

DMax Aggressive Parallel Decoding for dLLMs paper:

_akhaliq's profile picture

AK

514,034 subscribers

23,524 views • 5 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

We are excited to introduce Mercury, the first commercial-grade diffusion large language model (dLLM)! dLLMs push the frontier of intelligence and speed with parallel, coarse-to-fine text generation.
0:19

We are excited to introduce Mercury, the first commercial-grade diffusion large language model (dLLM)! dLLMs push the frontier of intelligence and speed with parallel, coarse-to-fine text generation.

Inception

1,917,263 views • 1 year ago

In the Mean While #Soundariya In Parallel Universe Be like !! Same Answer For Two Different Question paper 😂😂 #BiggBossTamil8 #BiggBossSeason8Tamil #BiggBossSeason8Tamil
0:13

In the Mean While #Soundariya In Parallel Universe Be like !! Same Answer For Two Different Question paper 😂😂 #BiggBossTamil8 #BiggBossSeason8Tamil #BiggBossSeason8Tamil

Joe Michael Praveen

100,519 views • 1 year ago

Can $300 RTX3060 be faster than $10,000 A100 for LLM decoding?🧐Yes Introduce EAGLE-2: - 4.26x🚀vanilla (13B) - 1.4x🚀EAGLE (13B) - No extra training than EAGLE - Maintain distribution Paper: Code: Demo:
0:13

Can $300 RTX3060 be faster than $10,000 A100 for LLM decoding?🧐Yes Introduce EAGLE-2: - 4.26x🚀vanilla (13B) - 1.4x🚀EAGLE (13B) - No extra training than EAGLE - Maintain distribution Paper: Code: Demo:

Hongyang Zhang

35,696 views • 2 years ago

DECODING FOR "ARIRANG" LOCKED RECORDINGS ON SPOTIFY STARTED.. ( #BTS_ARIRANG #BTSxSpotify
0:18

DECODING FOR "ARIRANG" LOCKED RECORDINGS ON SPOTIFY STARTED.. ( #BTS_ARIRANG #BTSxSpotify

BTS Updates, News & Charts ⁷

116,299 views • 5 months ago

Added context to my tiny diffusion model to enable sequential generation of longer outputs! Currently the context is a quarter of the sequence length (seq_len=256, context_len=64). I have a theory that the less semantic-value-per-token, the worse the “curse of parallel decoding” is. With parallel decoding, we independently predict multiple tokens in one step. With the sentence “My poker hand was a ___ ___”, two valid predictions are “two pair” and “straight flush”. Because each token prediction is independent though, we can end up with a nonsensical output like “two flush”. This seems to be exacerbated with low semantic-value-per-token, as now you need more tokens to express the same concept. Instead of needing to independently predict two tokens, we might need to predict 10 instead (which is of course much harder). The model currently has noticeably worse output compared to nanogpt (similar size) and I believe this is a main reason. I’ll try adding confidence-aware parallel decoding (from NVIDIA’s Fast-dLLM paper) and other tricks and see how much they improve generation quality.
0:12

Added context to my tiny diffusion model to enable sequential generation of longer outputs! Currently the context is a quarter of the sequence length (seq_len=256, context_len=64). I have a theory that the less semantic-value-per-token, the worse the “curse of parallel decoding” is. With parallel decoding, we independently predict multiple tokens in one step. With the sentence “My poker hand was a ___ ___”, two valid predictions are “two pair” and “straight flush”. Because each token prediction is independent though, we can end up with a nonsensical output like “two flush”. This seems to be exacerbated with low semantic-value-per-token, as now you need more tokens to express the same concept. Instead of needing to independently predict two tokens, we might need to predict 10 instead (which is of course much harder). The model currently has noticeably worse output compared to nanogpt (similar size) and I believe this is a main reason. I’ll try adding confidence-aware parallel decoding (from NVIDIA’s Fast-dLLM paper) and other tricks and see how much they improve generation quality.

Nathan Barry

89,040 views • 10 months ago

Achieving 3,000 tokens/s with 16-parallel inference of Qwen3.6 35B-A3B NVFP4 on a single RTX PRO 6000 WorkStation Edition. I am using ParallelHue—a tool I am developing—to color-code the tokens generated simultaneously via speculative decoding.
0:14

Achieving 3,000 tokens/s with 16-parallel inference of Qwen3.6 35B-A3B NVFP4 on a single RTX PRO 6000 WorkStation Edition. I am using ParallelHue—a tool I am developing—to color-code the tokens generated simultaneously via speculative decoding.

Hikari∣LocalLLM⚡

27,434 views • 1 month ago

Parallel Community Airdrop 🪂 All Parallel users will be eligible for an upcoming points airdrop! More details to be announced soon.
0:13

Parallel Community Airdrop 🪂 All Parallel users will be eligible for an upcoming points airdrop! More details to be announced soon.

Parallel

180,800 views • 2 years ago

BU meets Parallel Web Systems! We're collaborating with Parallel Web Systems for our web search on BU app! Read our blogpost! 👇
0:13

BU meets Parallel Web Systems! We're collaborating with Parallel Web Systems for our web search on BU app! Read our blogpost! 👇

Browser Use

12,237 views • 7 months ago

Origami horse animated by Parallel Studio for Cheval Blanc hotels.
0:17

Origami horse animated by Parallel Studio for Cheval Blanc hotels.

Catsuka

53,486 views • 7 months ago

Press 𝐘 for aggressive catch. 🎮 #STP
0:12

Press 𝐘 for aggressive catch. 🎮 #STP

Mizzou Football

66,311 views • 3 years ago

Aggressive protest in #Kota ! NEET aspirants and their parents held a protest against alleged irregularities & paper leak in #NEET-UG exam results in #Kota, Rajasthan, #NEET_परीक्षा_परिणाम Congress BJP #NEET_पेपर_रद्द_करो #RENEET
0:16

Aggressive protest in #Kota ! NEET aspirants and their parents held a protest against alleged irregularities & paper leak in #NEET-UG exam results in #Kota, Rajasthan, #NEET_परीक्षा_परिणाम Congress BJP #NEET_पेपर_रद्द_करो #RENEET

Indian Doctor🇮🇳

22,947 views • 2 years ago

Revised landing screen for Parallel mobile as per feedback. Cycles different factions. //
0:12

Revised landing screen for Parallel mobile as per feedback. Cycles different factions. //

//Kalos

19,055 views • 5 months ago

New paper with Kenneth Stanley ! We built a scalable fluid engine for artificial life 🧬 Paper:
0:12

New paper with Kenneth Stanley ! We built a scalable fluid engine for artificial life 🧬 Paper:

Ciaran@ALIFE🇨🇦

40,926 views • 2 months ago

Frog paper craft ideas for kids
0:16

Frog paper craft ideas for kids

Craft Gallery

34,378 views • 9 days ago

Paper Cup Bunny Craft for Kids
0:14

Paper Cup Bunny Craft for Kids

Craft Gallery

233,352 views • 1 year ago

Easy Craft for Kids- Paper Firework
0:12

Easy Craft for Kids- Paper Firework

Do It Yourself

38,377 views • 3 years ago

MMaDA-Parallel Multimodal Large Diffusion Language Models for Thinking-Aware Editing and Generation
0:16

MMaDA-Parallel Multimodal Large Diffusion Language Models for Thinking-Aware Editing and Generation

AK

66,158 views • 9 months ago

there is an us the series parallel for everything DONT TOUCH DONT KISS #MoonshadowSeriesEP2
0:15

there is an us the series parallel for everything DONT TOUCH DONT KISS #MoonshadowSeriesEP2

rome

16,728 views • 19 days ago