Nothing here yet.
Nothing here yet.
18h ago · 20 min read · If you've been following my recent posts, you might have seen my repository and the issue I opened on the NemoClaw repo (#315). In it, I detailed a rather complex workaround—involving TCP relays and injecting iptables rules via nsenter—just to allow ...
Join discussion6d ago · 8 min read · Why This Comparison Exists I've been running Nemotron Nano 9B v2 Japanese on an RTX 5090 with vLLM 0.15.1 for months now. Before settling on vLLM, I evaluated TensorRT-LLM, considered Ollama, and benchmarked llama.cpp. This article captures what I le...
Join discussionMar 13 · 6 min read · I built a multi-engine shogi AI, deployed it to rated games on Floodgate, and watched it lose to programs it should have crushed. I'm a patent lawyer. I'd been writing code for less than two months when that happened. Three months ago, in December 20...
Join discussionMar 8 · 4 min read · Introduction PatentLLM (patent search AI) and Hanrei-DB (case law search AI) are both personal projects built with Python + Streamlit + SQLite. Through their development, I realized that the choice of data format directly impacts an application's mai...
Join discussion
Mar 8 · 2 min read · NVIDIA Nemotron-Nano-9B-v2-Japanese This is a 9B parameter LLM specialized for Japanese, released by NVIDIA. It is based on the Mamba SSM (State Space Model) architecture, which efficiently processes long texts using an approach different from Transf...
Join discussion