Stories by danielhanchen

2.71bit DeepSeek-V3-0324

Gemma 3: Google's new multimodal models

How to Run QwQ-32B effectively

Unsloth AI (YC S24) is hiring ML engineers

Train your own R1 reasoning model

How to run 1.58bit DeepSeek R1 with Open WebUI

Phi-4 Bug Fixes

My take on the Post Pretraining world

Dynamic 4bit Quantization

Show HN: Finetune Llama 3.2 Vision in a Colab

Python 3.11 is 1.25x faster than 3.10

Fixing Gradient Accumulation

Unit Economics of LLM APIs

LoRA Learns Less and Forgets Less Updated

VLLM automatic prefix / prompt caching

Higher Temperatures and Min_p Sampling

Show HN: Open-source fine-tuning in a Colab notebook

Sahm rule signals start of recession

Low Level Technicals of LLMs [video]

Gemma-2 2B beats GPT3.5 on Chatbot Arena

1