QwenGPU: Simulating GPU Training Inside an LLM

The QwenGPU project trains neural networks inside an LLM using simulated distributed GPUs, achieving 98.8% on XOR, 86.9% on MNIST, and generating recognizable digits, with collective ops matching real NCCL bit-for-bit.

2026-11-03 ~ 2026-11-03 · 4 related posts

• Training neural nets inside LLMs: XOR MLP goes 53% to 98.8% in 40 steps on simulated QwenGPUs — zeeshanp_ · 2026-11-03
• All six collectives on 4 QwenGPUs match real NCCL bit-for-bit on GB200s — zeeshanp_ · 2026-11-03
• Megatron-style tensor parallelism inside LLMs trains an MNIST classifier to 86.9% accuracy — zeeshanp_ · 2026-11-03
• Tiny latent diffusion model trained inside LLMs generates MNIST digits, 76% recognized — zeeshanp_ · 2026-11-03