Stories by tosh

LLMs: Common HParam Settings

Iron Lung

AirLLM

Capybara dataset is now open-source and available

Pressure-tested the most popular open-source LLMs

Why do countries change their name?

Yi-34B-200K-DARE-merge-v5

Extreme brainstorming questions to trigger new, better ideas

Kamal and Hetzner 50x cheaper than Heroku

Ultra-Low Inference Latency with LLaMA 65B on PyTorch/XLA

MiniMA-2-3B

Kaggle: GPU Notebooks with 4 CPUs, 29 GBs of RAM

Solar 10.7B

Japan Preparing EU-Style Law to Force Apple to Allow App Sideloading

Nous-Hermes-2-Yi-34B

Understanding and Using Supervised Fine-Tuning (SFT) for Language Models

YAYI 2

Metal-flash-attention: Faster alternative to Metal Performance Shaders

SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis

Mixtral_7Bx2_MoE