GPU
Hello! In the previous article, "TensorRT 10 × Blackwell Migration Guide (Part 1): Inference Assets Won't Run on RTX 50 — The Basics and the First Walls," we covered why existing inference assets stop working in the move to the Blackwell generation, and how to get a minimal
GPU
Hello! You get a new GPU, excitedly move your existing inference stack over, and everything that ran fine yesterday now stops with errors across the board — sound familiar? In migrations to the NVIDIA RTX 50 series and NVIDIA RTX PRO series (the Blackwell generation), this happens with rather high probability.
GPU
Hello! This is the Qualiteg Product Development Team! Today we describe what happened when we put two RTX 4090s into a Windows + WSL2 machine and tried to run a large open model with vLLM — and got thoroughly stuck at NCCL initialization. Information on this is scattered and fragmentary, and it
DIY GPU Machines
NVIDIA GPU List Search Series All SM_120 (Blackwell) SM_103 (Blackwell Ultra) SM_100 (Blackwell) SM_90 (Hopper) SM_89 (Ada Lovelace) SM_86/SM_87 (Ampere) SM_75 (Turing) SM_70/SM_72 (Volta) SM_60/SM_61/SM_62 (Pascal) SM_50/SM_52/SM_53 (Maxwell)
AI-Business
— Where to take position in the age of the SaaS shake-up Hello! This is Qualiteg Consulting! Over the past few years, we have seen a steady increase in consultations and discussions on the theme of "Japan's AI strategy." Since the arrival of generative AI, everyone from
LLM
Introduction This report provides a comprehensive analysis of the performance of Japanese-capable LLMs, based on benchmark data from Nejumi Leaderboard 4 (March 6, 2026 edition). Last time, we published our analysis report on the December 18, 2025 edition, and in roughly three months the landscape has shifted dramatically once again.
Generative AI Frontlines
Hello! Greetings from the Qualiteg consulting team! As of 2026, we believe coding is one of the biggest use cases for LLMs. Indeed, Menlo Ventures' research positions coding as a flagship use case for enterprise AI. Having recognized this, AI platform providers have been launching their own AI coding
LLM
Hello! Welcome to a special extra installment of our LLM Inference Infrastructure Provisioning Course! Part 3, "Estimating Inference-Time Memory Consumption for Your Model", we introduced the two major consumers of GPU memory, the model footprint and the KV cache, and explained how to calculate the KV cache size
AI-Business
Hello! Between 2025 and 2026, AI-related investment in Japan has been moving fast. Domestic LLM development, a data center construction boom, expanding government support programs. The sense of urgency—"Japan cannot afford to fall behind in AI"—shows up clearly in the numbers. This momentum is welcome, and
LLM
Hello! Why can OpenAI and Anthropic build world-class LLMs while no LLM from Japan stands shoulder to shoulder with them? Is it a gap in technical capability? A talent problem? The answer is simpler than that: the number of GPUs, and money. Today we focus on "LLM training"
Daily Dev Tips
Hello! When designing parallel processing systems that use multiple GPUs, it is easy to design the whole system without giving much thought to the CPU. "The GPU is the star of a machine learning system, so surely the CPU doesn't need much attention?" Actually, that is
Generative AI Frontlines
Why can't the Transformer that powers GPT simply be dropped into lipsync? We unpack three challenges — data, computation, and overfitting — how to choose between LSTMs and Transformers in practice, and transfer learning as a third option, in a dialogue between Professor Quo and Manabu.
Daily Dev Tips
Hello! In this post, we look at how to resolve the "libcudnn.so.9: cannot open shared object file" error that can occur during GPU inference with ONNX Runtime. When running GPU inference with ONNX Runtime, you may run into a CUDA provider initialization error. This article explains
PyTorch
Hello! In this post, we explain how to deal with the following problem in a PyTorch 2.6.0 environment. NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_70
Daily Dev Tips
Hello! This is the Qualiteg Product Development team! This is the story of how simply converting a synchronous method into an asynchronous one (async) hit us with an unexpected memory leak*. It happened while we were developing a video processing system that uses deep learning models. We had a requirement
AI Mathematics
1. Position Loss (L_position) - Accuracy of the Mouth Shape Time Mouth opening Truth Predicted L_position = Σᵢ wᵢ × ||y_pred - y_true||² Computes the difference between predicted and ground-truth values at each time step. Important parameters (jaw opening, mouth opening) receive larger weights. jaw_open: ×2.0