minte1431/Qwen3.6-27B-OBLITERATED

Partial fine-tune vs Qwen/Qwen3.6-27B. Partial change: 26% of the weights changed (attention, lm_head, mlp), in 18 of 64 layers.

Tensors changed: Embeddings 0/1, MLP 48/192, Attention 44/448, lm_head 1/1, Norms 0/209

9 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse