rpn-gpt - a markab Collection

markab 's Collections

rpn-gpt

updated about 13 hours ago

An exact Forth machine in Nemotron 3.5 Lightning: merged model, LoRA, NVFP4, GGUF (Ollama), MLX, the kernel in the weights, and the data.

markab/rpn-gpt-lightning

Text Generation • 32B • Updated about 13 hours ago

Note Start here: merged BF16 (vLLM, transformers)
markab/rpn-gpt-lightning-Forth

Text Generation • 4B • Updated about 12 hours ago

Note The Forth kernel in the weights
markab/rpn-gpt-lightning-NVFP4

Text Generation • 33B • Updated about 13 hours ago

Note Blackwell, vLLM: ~BF16 accuracy at 22 GB
markab/rpn-gpt-lightning-GGUF

Text Generation • 32B • Updated about 12 hours ago

Note Ollama, llama.cpp, LM Studio
markab/rpn-gpt-lightning-MLX-4bit

Text Generation • 32B • Updated about 13 hours ago • 1

Note Apple Silicon
markab/rpn-gpt-lightning-MLX-8bit

Text Generation • 32B • Updated about 13 hours ago

Note Apple Silicon, 8-bit
markab/rpn-gpt-lightning-LoRA

Text Generation • Updated about 13 hours ago

Note The adapter, for hot-swap
markab/rpn-gpt-data

Viewer • Updated about 13 hours ago • 13.5k

Note Training rows, demonstrations, held-out evals