Explore how @Wealthsimple has revolutionized their operations through accelerated #machinelearning model delivery and inference, reducing model deployment times from several months to 15 minutes using NVIDIA's AI #inference platform. Learn more > https://
nvda.ws/424Bliu
MACHINE LEARNING
-
Wealthsimple Cuts ML Model Deployment Time to 15 Minutes
By
–
-
ICLR Acceptance Decisions and Research Community Reflections
By
–
in light of ICLR acceptance decisions yesterday, here's a little reminder from robin williams
-
AlphaGeometry Solves IMO Geometry Problems at Gold Medalist Level
By
–
On a test set of 30 IMO geometry problems from the years 2000-2022, #AlphaGeometry can solve 25 problems. This is approaching that of the average human gold medalists with 25.9 and surpasses the previous state-of-the-art system, which can only solve 10, by a large margin! (Note
-
Synthetic Mathematical Proofs Discovery Through AI Analysis
By
–
Here is a closer look at our synthetic data ordered by proof lengths. We can see that trivial and well-known theorems, such as the Euler circle, were re-discovered. Out of 100M examples, there are 9M with auxiliary constructions or the "insights". Some examples have very long
-
Set-Based Language Models Without Positional Embeddings
By
–
actually I don't need anything with positional embeddings since i'm doing something weird, i just realized this, my objective is 'set based' so there's no ordering. so really i guess i could use most LMs with positional encoding disabled?
-
Anchor Points Problem in Machine Learning Initialization Methods
By
–
how would you do it? check this out: http://
arxiv.org/abs/2303.00721 even these people need 15 or so "starting" anchor points, which i consider cheating in my problem setting -
Multimodal Deep Learning Models for Asynchronous Data Streams
By
–
A deep learning paper from our team on how to build multimodal models, where data for individual modalities arrives at different speeds, by combining powerful pretrained text, image and video models. Useful for robotics. Congrats @konradzolna @serkancabi @yutianc Eric, Claudio,… https://t.co/vPmZpDkve3
— Nando de Freitas (@NandoDF) 17 janvier 2024A deep learning paper from our team on how to build multimodal models, where data for individual modalities arrives at different speeds, by combining powerful pretrained text, image and video models. Useful for robotics. Congrats @konradzolna @serkancabi @yutianc Eric, Claudio,
-
Synthetic Data Generation Using 100K CPU Workers
By
–
These are examples of the random diagrams that we start off with for our synthetic data generation. We use 100K CPU workers to generate one billion random diagrams, then run symbolic deduction and traceback for 3-4 days. After deduplication and filtering by "interestingness", we
-
Long Context Encoder Models and Token Processing Optimization Techniques
By
–
anyone know a model that’s out there that’s encoder-only (BERT-like) but supports a really long context length? also what's the most efficient way of processing many tokens like this? i know about enabling FlashAttention & BetterTransformer. what else is out there?
-
Synthetic Data Generation for Olympiad-Level Mathematical AI Training
By
–
The lack of training data for maths, especially those at the Olympiad level, motivated us to generate everything synthetically. Key to this process is the idea of "symbolic deduction and traceback". We start from random diagrams, try to find all properties through forward