abeja/ABEJA-QwQ32b-Reasoning-Japanese-v1.0

Full fine-tune vs Qwen/QwQ-32B. Fine-tune: 100% of the weights changed, across 64 of 64 layers.

Tensors changed: Embeddings 1/1, Norms 129/129, MLP 192/192, Attention 448/448, lm_head 1/1

3 derivatives checked