birgermoell/oellm-9b-math-rlvr-canary-step16-experimental
Partial fine-tune vs Neonkraft/oellm-9b-256k-theta64m-prelude-anneal300b-instruct-sft. Partial change: 76% of the weights changed (attention, mlp, norms), in 36 of 36 layers.
Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 51/145, MLP 108/108, Attention 144/144
1 derivatives checked
- birgermoell/oellm-9b-math-rlvr-phase-a-32step-experimental: Partial fine-tune