RESEARCH - AI Dynamics

Good question. Yes, because LLaMA-Adapter v2 is basically LLaMA-Adapter but it also has trainable RMSNorm parameters and trainable bias units (in FC layers)

→ View original post on X — @rasbt

15 June 2023

MACARONS: 3D Reconstruction and Environment Exploration from RGB Images

By

@_akhaliq

–

15 June 2023 17h01

MACARONS: Mapping And Coverage Anticipation with RGB Online Self-Supervision

paper page: https://t.co/3aaipPlMcj

introduce a method that simultaneously learns to explore new large environments and to reconstruct them in 3D from color images only. This is closely related to the… pic.twitter.com/rj7GWrOy9f
— AK (@_akhaliq) 15 juin 2023

MACARONS: Mapping And Coverage Anticipation with RGB Online Self-Supervision paper page: https://
huggingface.co/papers/2303.03
315
… introduce a method that simultaneously learns to explore new large environments and to reconstruct them in 3D from color images only. This is closely related to the

→ View original post on X — @_akhaliq

15 June 2023

LoRA vs Adapter: Performance and Memory Usage Comparison

By

@rasbt

–

15 June 2023 16h58

Good question! On average: LoRA: 21.33 tokens/sec; Memory used: 14.59 GB
Adapter: 26.22 tokens/sec; Memory used: 14.59 GB

→ View original post on X — @rasbt

15 June 2023

Memory reduction: Full finetuning vs Adapter vs LoRA comparison

By

@rasbt

–

15 June 2023 16h37

Btw, it's also a big reduction in memory: 16 gigs instead of 6 x 40 GB. Due to the reduced parameter counts in the backward pass: Full finetuning: 7,217,189,760
Adapter: 1,365,330
Adapter v2: 3,839,186
LoRA: 3,506,176 (I used LoRA rank of 16 to match Adapter v2 above.)

→ View original post on X — @rasbt

15 June 2023

Efficient Falcon Finetuning with LoRA on Single GPU

By

@rasbt

–

15 June 2023 16h24

Have been heads-down porting LoRA (low-rank-adaptation) to finetune Falcon more efficiently and ran some performance benchmarks. My longer write-up here: https://
lightning.ai/pages/communit
y/finetuning-falcon-efficiently/
… Long story short: you can finetune Falcon in 1 h on a 52k dataset using a single GPU with 16 GB RAM.

→ View original post on X — @rasbt

15 June 2023

ChatGPT versus Raphaël Enthoven: Media Experiment Fairness Questioned

By

@dr_l_alexandre

–

15 June 2023 16h18

Cette expérience est une escroquerie médiatique

ChatGPT a eu 90 secondes

Raphael Enthoven a eu 90 minutes

J’aime beaucoup Raphaël malgré tout 😂 https://t.co/OuK0URIusg
— Docteur Laurent Alexandre (@dr_l_alexandre) 15 juin 2023

Cette expérience est une escroquerie médiatique ChatGPT a eu 90 secondes Raphael Enthoven a eu 90 minutes J’aime beaucoup Raphaël malgré tout

→ Voir le post original sur X — @dr_l_alexandre

15 June 2023

AI Anthology: 20 Essays on Beneficial AI Applications

By

@marktabnet

–

15 June 2023 16h07

"I am honored to present the AI Anthology, a compilation of 20 inspiring essays authored by distinguished scholars and professionals from various disciplines. The anthology explores the diverse ways in which AI can be channeled to benefit humanity" https://
bit.ly/3N9m3Si

→ View original post on X — @marktabnet

15 June 2023