Bones / Sage / Nova LoRA -- freshdata lineage, adapters_freshdata_v25

Automated push, real numbers, no cherry-picking. This is whichever checkpoint is newest at push time, good or bad, per Heuremen's own standing policy: consistent visible history over curated-looking results.

Val loss of the checkpoint actually in this repo (the run's real best, iter-promoted): 1.667
Val loss at the run's final iter 1000 (confirms training genuinely completed, not necessarily what's in adapters.safetensors above): 2.127
Real curated corpus size at push time: 2806 rows
Pushed: 2026-08-26 20:21 UTC

Part of the bones-sage-nova-lora family -- this is the actively-iterating line, not the flagship release.

Downloads last month: -; Downloads are not tracked for this model. How to track

MLX

Hardware compatibility

Quantized

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Wayfinder6/bones-sage-freshdata

Base model

Qwen/Qwen2.5-14B

Finetuned

mlx-community/Qwen2.5-14B-Instruct-4bit

Adapter

(2)

this model