Cached Transformers: Improving Transformers with Differentiable Memory Cache
Paper
•
2312.12742
•
Published
•
14
Jinwei Gu
jwgu
AI & ML interests
Computational photography, computational imaging, computer vision, computer graphics, machine learning, AR/VR
Recent Activity
posted
an
update
18 days ago
🎉 NEW RELEASES: Cosmos Predict 2.5 and Transfer 2.5
Cosmos Predict 2.5:
- Combines Text2World, Image2World, and Video2World
- Multimodal, future-state video prediction
Cosmos Transfer 2.5:
- High-fidelity multicontrol world simulations
- Inputs: RGB, depth, segmentation—blended seamlessly
These updates boost development of autonomous vehicles, robotics, and video analytics.
Don’t miss Jensen Huang’s keynote at NVIDIA GTC Washington, D.C. on 10/28 to hear the latest in physical AI.
📺 Watch live: https://nvda.ws/4pUjF4x
🔗 Try Predict 2.5:
https://nvda.ws/4otReZZ
🔗 Try Transfer 2.5:
https://nvda.ws/46GEx7T
liked
a model
19 days ago
nvidia/Cosmos-Transfer2.5-2B
authored
a paper
4 months ago
4DSloMo: 4D Reconstruction for High Speed Scene with Asynchronous
Capture
Organizations
upvoted
a
paper
over 1 year ago