Skip to content

coba:1b-fp16

16 TagsUpdated 1.1B params512 contextEnglishApache-2.0by Ollaya

The most accurate size (Ettin 1B): 0.608 on typed decisions, zero-shot.

1.1b
ollaya run coba:1b-fp16 --preset triage "I was charged twice for my subscription this month and want a refund."

Details

  • grapha031573f8daf · 3 MBonnx · encoder · 1.1B · fp32
  • graphd7d2283df2ad · 4 MBonnx · encoder · 1.1B · fp16
  • weights37603159c90c · 4.4 GBhuggingface.co/ollaya-dev/coba-1b/resolve/044c84c…/model.safetensors
  • tokenizer6c8aaa9a5420 · 4 MBhuggingface.co/ollaya-dev/coba-1b/resolve/044c84c…/tokenizer/tokenizer.json
  • decisionb01dafbc1f88 · 457 B{"engine": "onnx", "family": "laya", "encoder": "encoder", "layout": "laya-markers-v1", …}
  • calibration568c4a80539b · 86 B{"temperature": [0.89, 0.8, 0.69]}
  • licenseef14791048a7 · 10 KBCoba by Ollaya (https://huggingface.co/ollaya-dev)
  • archeca91c9a5c40 · 1 KBapplication/vnd.ollaya.arch
  • paramsa1d447ba672f · 21 Bprecision fp16

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.

Variant of

coba readme and all models