rita-cohere's picture
Upload folder using huggingface_hub
4719d8c verified
|
Raw
History Blame Contribute Delete
682 Bytes
metadata
license: apache-2.0
language:
  - multilingual
tags:
  - iol-ai
  - linguistics
  - reasoning
  - multilingual
  - tiny-aya

tiny-aya-reasoning-sft-long

Tiny Aya reasoning SFT (long run) — final checkpoint ckpt-20042 from Mehrnaz sft_default run 20260723_224502.

Source

s3://us-east-01a/30d/post-training/mehrnaz_mofakhami_cohere_com/tiny_aya_sft/sft_default/20260723_224502_run/ckpt-20042

Inference weights are the hf_export/bf16/ export of that checkpoint (uploaded when available).

Decode

IOL Space script.py matches rita-cohere/tya-math-ling-nb-t06 (notebook-style think/answer decode).