coba
16 TagsUpdated Apache-2.0by Ollaya
Ollaya's own decision model: Laya's network on Ettin encoders, distilled from pplx-decider:flash. English. Full-precision 1b and 400m, and ternary sizes small enough for a browser.
zero-shotbrowserfine-tuned1.1b75m164m421m
Tags
| Name |
|---|
| cobalatestSame as coba:400m.1.7 GB · 512 ctx · English |
| coba:150m-ternaryTernary Ettin 150M, in-domain; the model the /try page runs in the browser (66 MB).665 MB · 512 ctx · English |
| coba:1bThe most accurate size (Ettin 1B): 0.608 on typed decisions, zero-shot.4.4 GB · 512 ctx · English |
| coba:400mEttin 400M: 0.591 on typed decisions, zero-shot; 24 ms for five questions on an RTX 5090.1.7 GB · 512 ctx · English |
| coba:400m-ternaryThe 400m with ternary weights, in-domain: trained on the train splits of ten public benchmark tasks.1.7 GB · 512 ctx · English |
| coba:68m-ternaryThe smallest: ternary Ettin 68M, in-domain.307 MB · 512 ctx · English |
| coba:150m-ternary-fp16fp16 variant of coba:150m-ternary665 MB · 512 ctx · English |
| coba:150m-ternary-fp32fp32 variant of coba:150m-ternary665 MB · 512 ctx · English |
| coba:1b-fp16fp16 variant of coba:1b4.4 GB · 512 ctx · English |
| coba:1b-fp32fp32 variant of coba:1b4.4 GB · 512 ctx · English |
| coba:400m-fp16fp16 variant of coba:400m1.7 GB · 512 ctx · English |
| coba:400m-fp32fp32 variant of coba:400m1.7 GB · 512 ctx · English |
| coba:400m-ternary-fp16fp16 variant of coba:400m-ternary1.7 GB · 512 ctx · English |
| coba:400m-ternary-fp32fp32 variant of coba:400m-ternary1.7 GB · 512 ctx · English |
| coba:68m-ternary-fp16fp16 variant of coba:68m-ternary307 MB · 512 ctx · English |
| coba:68m-ternary-fp32fp32 variant of coba:68m-ternary307 MB · 512 ctx · English |
Tags without a suffix load fp16 on a CUDA GPU and fp32 on CPU; add -fp16 or -fp32 to pin one precision.