Klimenty Titov's picture
6

Klimenty Titov

markcda
ยท

AI & ML interests

None yet

Recent Activity

liked a model 4 months ago
AIDC-AI/Marco-o1
liked a model 10 months ago
mistralai/Mistral-7B-Instruct-v0.3
liked a Space 10 months ago
Intel/low_bit_open_llm_leaderboard
View all activity

Organizations

None yet

markcda's activity

reacted to DmitryRyumin's post with ๐Ÿ”ฅ 10 months ago
view post
Post
1829
๐Ÿ”ฅ๐Ÿš€๐ŸŒŸ New Research Alert - YOCO! ๐ŸŒŸ๐Ÿš€๐Ÿ”ฅ
๐Ÿ“„ Title: You Only Cache Once: Decoder-Decoder Architectures for Language Models ๐Ÿ”

๐Ÿ“ Description: YOCO is a novel decoder-decoder architecture for LLMs that reduces memory requirements, speeds up prefilling, and maintains global attention. It consists of a self-decoder for encoding KV caches and a cross-decoder for reusing these caches via cross-attention.

๐Ÿ‘ฅ Authors: Yutao Sun et al.

๐Ÿ“„ Paper: You Only Cache Once: Decoder-Decoder Architectures for Language Models (2405.05254)

๐Ÿ“ Repository: https://github.com/microsoft/unilm/tree/master/YOCO

๐Ÿ“š More Papers: more cutting-edge research presented at other conferences in the DmitryRyumin/NewEraAI-Papers curated by @DmitryRyumin

๐Ÿ” Keywords: #YOCO #DecoderDecoder #LargeLanguageModels #EfficientArchitecture #GPUMemoryReduction #PrefillingSpeedup #GlobalAttention #DeepLearning #Innovation #AI
ยท