Nanochat performance metrics

[obriensystems]

• Nanochat

• andrej.karpathy Nanochat on NVIDIA DGX Spark GB10 - Nanochat frontier LLM from scratch - andrej.karpathy Nanochat on NVIDIA DGX Spark GB10, Apple Silicon and NVIDIA RTX-A6000/A4500/A4000/3500Ada blog#170
• https://github.com/obriensystems/nanochat/blob/dev/runs/runcpu.sh#L79
• DGX Spark andrej.karpathy Nanochat on NVIDIA DGX Spark GB10 - 900M or 1B model via PyTorch on CUDA foundation-transformer-llm#7
• Apple Silicon nanochat on Apple Silicon M2Ultra, M3Ultra, M4Pro, M4Max via PyTorch and mps foundation-transformer-llm#8
• Windows/WSL NVIDIA nanochat on windows for RTX-A4000, Dual RTX-A4040, RTX-A6000, Dual GTX-4090 via PyTorch on CUDA foundation-transformer-llm#9
• https://github.com/ObrienlabsDev/blog/blob/main/nvidia-dgx-spark.md
• Add --model-tag to scripts.chat_cli call in runcpu.sh/speedrun.sh example karpathy/nanochat#542