Guide Improving Diffusion Models without Training Through a Teacher's Guide
MACHINE LEARNING
-
Presto: Accelerating Music Generation with Distilled Diffusion Steps
By
–
Presto!
— AK (@_akhaliq) 8 octobre 2024
Distilling Steps and Layers for Accelerating Music Generation
Despite advances in diffusion-based text-to-music (TTM) methods, efficient, high-quality generation remains a challenge. We introduce Presto!, an approach to inference acceleration for score-based diffusion… pic.twitter.com/ZB4UjmEhRSPresto! Distilling Steps and Layers for Accelerating Music Generation Despite advances in diffusion-based text-to-music (TTM) methods, efficient, high-quality generation remains a challenge. We introduce Presto!, an approach to inference acceleration for score-based diffusion
-
Evidently AI: 450 Machine Learning and LLM System Design Case Studies
By
–
Evidently AI – ML and LLM system design: 450 case studies https://
bit.ly/3Y7A1Lw
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Debiasing Large Language Models with Synthetic Data Pipeline
By
–
We are excited to share our latest work on debiasing large language models, accepted to COLM 2024 We introduce a lightweight, simple pipeline using ChatGPT to generate synthetic data for debiasing open-source LLMs through parameter-efficient fine-tuning. Our method outperforms
-
Interpreting AI Model Training Costs from Research Paper
By
–
I don't think it's absurd to think that from the paper. If I look at Table 1 from https://
arxiv.org/pdf/1906.02243 paper, it sure seems like the cost of the last row of the table is the cost of "Training one model (GPU)". How else should one interpret that line? Not to mention that -
Robots Collect Data to Optimize Performance Unlike Standard Machines
By
–
"…Unlike standard machines, robots collect data which can be used to optimize performance". – @AmbiRobotics CTO Jeff Mahler.
-
Neural Architecture Search Methodology Error and Measurement Critique
By
–
Knowing that Neural Architecture Search is a one-time cost per problem domain, not a per-problem cost was definitely widespread knowledge in the ML community at the time of the publication. So that 3261X error was avoidable by simply using their own actual measurements of the
-
Learning from Building an AI Video Editor
By
–
What I’ve Learned in the Past Year Spent Building an AI Editor https://
bit.ly/4eJj3bE
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Llama 3.2 achieves 10x improvement over previous generation
By
–
the new 1 billion parameters Llama model (version 3.2) is head-to-head with the 13 times larger version of one years ago (llama 13B version 2) on lmsys chatbot arena exciting to see such 10x improvements on challenging benchmark it's an amazing sign for small/local/open models
-
Error Analysis Improved Water Forecasting Machine Learning Model
By
–
Fixing Bias: How Error Analysis Improved My Water Forecasting Model https://
bit.ly/3N4RBJI
#AI #MachineLearning #DeepLearning #LLMs #DataScience