Skip to content

decima:base-clean

7 TagsUpdated 321M params512 context100+ languagesApache-2.0by A. M. Madani

Decima-base-clean 1.0: decima-base retrained without non-commercial data, for commercial use; 0.474 on typed decisions.

multilingual321m
ollaya run decima:base-clean --preset triage "I was charged twice for my subscription this month and want a refund."

Details

  • graphe11e5fa7abf3 · 5 MBonnx · decima · 321M · fp32
  • weightsbc712693536e · 1.2 GBhuggingface.co/amyrmahdy/decima-base-clean/resolve/35eb218…/pytorch/encoder/model.safetensors
  • weights5e886612d8ba · 57 MBhuggingface.co/amyrmahdy/decima-base-clean/resolve/35eb218…/pytorch/head.safetensors
  • tokenizerd941237f7b66 · 34 MBhuggingface.co/amyrmahdy/decima-base-clean/resolve/35eb218…/pytorch/encoder/tokenizer.json
  • decision5b6900ee0335 · 3 KB{"engine": "onnx", "family": "decima", "encoder": "", "layout": "decima-late-interaction-v2", …}
  • calibration081c56743368 · 291 B{"temperature": [0.693, 1.0, 0.693]}
  • license7eedb1836798 · 11 KBDecima-small, Decima-base, Decima-base-clean, Decima-agent, Decima-guard and Decima-guard-small by A. M. Madani (https://huggingface.co/amyrmahdy, https://github.com/amyrmahdy/decima), Apache-2.0.

Every layer is checked against its sha256 when it is pulled. Weights and tokenizers download from the model author's Hugging Face repository at a pinned commit; Ollaya never re-hosts them.

Decima readme and all models