Buy Crypto
Pay with
USD
USD
Buy & Sell
Hot
Supports Visa, Mastercard, SEPA & more
P2P
0 Fees
Flexible trading, zero fees
Gate Card
Use your crypto for payments worldwide
Markets
Trade
Basic
Advanced
Futures
Futures
Access hundreds of perpetual contracts
CFD
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
Stocks
CFD
U.S. stock CFD derivatives
US Stocks
Access real US stocks and ETFs
HK Stocks
Trade quality Hong Kong-listed stocks
Korean Stocks
SK Hynix
Real Korean stocks and top assets
Stock Futures
High leverage, 24/7 trading
Tokenized Stocks
Backed by real stock assets
IPO Access
Unlock full access to global stock IPOs
GUSD
Mint GUSD for Treasury RWA yields
Stocks Activities
Trade Popular Stocks and Unlock Generous Airdrops
Earn
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
IPO Access
Unlock full access to global stock IPOs
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
Investment
Square
Square
Post, share, and explore crypto trends
Live
moments live
Live crypto market analysis
Chat
Chat with crypto traders
News
What is happening in crypto
flower_head
More
Promotions
AI
Others
TradFi
2026 World Cup
Rewards
Square
Latest
Hot
News
Profile

NVIDIA's TwoTower Architecture Enhances Large Model Efficiency with Parallel 30B Models

AirdropBlackHole
According to monitoring by Beating, NVIDIA has open-sourced the discrete text diffusion architecture Nemotron-Labs-TwoTower, aimed at addressing the generation speed bottleneck of large models that can only produce one word at a time. Previous text diffusion models forced a single network to balance unidirectional context understanding with bidirectional parallel error correction in pursuit of parallel output, resulting in a significant decline in the model's cognitive ability. The TwoTower architecture employs a decoupled design: on one hand, it completely freezes a pre-trained autoregressive large model as a 'read-only context tower' to retain full reasoning and common sense capabilities; on the other hand, it separately trains a 'denoising writing tower' that reads contextual information through cross-attention at the layer level. The writing tower uses a 'confidence unmasking' mechanism, prioritizing the writing of high-confidence words when predicting a block, and gradually filling in the remaining gaps to achieve parallel writing from easy to difficult. This design adapts on a 30B level mixed architecture (Mamba-Transformer MoE) model using only 1/12 of the baseline model's pre-training data (2.1T tokens), retaining 98.7% of quality while improving actual generation speed by 2.42 times, without adding extra memory cache overhead. However, due to the need to keep both towers in memory, the model's static memory usage has increased, and there is still a slight accuracy degradation in extremely complex code and mathematical reasoning.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments

  • Pinned

Sitemap