Resources
OpenCerebral
- Hugging Face organization — all 17 models and their cards
- Model index — the same list with release dates and sizes
- Benchmark tables
- How we train
Training data
Tools we use
- Transformers — the models load with stock Transformers, no custom code
- llama.cpp — GGUF builds and CPU inference
- GPT-NeoX-20B tokenizer — the BPE vocabulary Boris uses
Related
- Klondike Software Project — parent project
- KMOS — the x86_64 kernel developed alongside this work