Mechanical Dirk

Mechanical Dreams

Science EN ↓ 170 episodes

An automatically generated podcast about machine learning and natural language processing. The two fictional hosts talk about papers that I want to learn more about on my way to work. It's not good, but it's useful.

Be sure to visit the podcast's website and support the creator: mechanicaldreams.mechanicaldirk.com

Author

Mechanical Dirk

Episodes

The Zamba2 Suite 29.11.2024 13:04

Small-scale proxies for large-scale Transformer training instabilities 26.11.2024 10:07

Hardware Scaling Trends and Diminishing Returns in Large-Scale Distributed Training 25.11.2024 10:29

I slightly tweaked the personality of the hosts.

Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations 19.11.2024 8:34

Understanding WSD Learning Rates 18.11.2024 9:10

Toward Understanding Why Adam Converges Faster Than SGD for Transformers 16.11.2024 7:30

New generation algorithm! Should make the episodes longer, more detailed, and more coherent.

Amuro & Char - Analyzing the Relationship between Pre-Training and Fine-Tuning of Large Language Models 08.11.2024 13:36

Evaluation data contamination in LLMs: How do we measure it and (when) does it matter? 07.11.2024 6:28

How Does Critical Batch Size Scale in Pre-training? 04.11.2024 7:41

The Road Less Scheduled 01.11.2024 8:54

Learning-Rate-Free Learning by D-Adaptation 31.10.2024 4:37

Scaling FP8 Training to Trillion Token LLMs 30.10.2024 9:55

Geometric Dynamics of Signal Propagation Predict Trainability of Transformers 29.10.2024 15:44

A Survey on Model MoErging 28.10.2024 8:58

Liquid Time-constant Networks 27.10.2024 8:11

Scaling Laws for Predicting Downstream Performance in LLMs 26.10.2024 10:07

A Spectral Condition for Feature Learning 25.10.2024 16:53

Don't decay the learning rate 24.10.2024 7:07

A classic paper about learning rates.

OLMoE 23.10.2024 6:59

professor norris: Welcome back to Mechanical Dreams, the podcast where we delve into the exciting world of machine learning and natural language processing. I'm Professor Norris, and as always, I'm joined by my brilliant student, Linda. linda: It's great to be back, Professor. And I'm particularly excited about today's paper. It tackles a topic that's been buzzing in...

An Empirical Model of Large Batch Training 23.10.2024 11:32

First attempt to automatically generate a podcast from a paper. This one is way too short, but it's a start.