ECE-ILAB/POIROT-ECE-1.0

Partial fine-tune vs AXCXEPT/Qwen3-EZO-8B-beta. Partial change: 90% of the weights changed (attention, lm_head, mlp, norms), in 35 of 35 layers.

Tensors changed: Embeddings 0/1, Norms 84/141, MLP 102/105, Attention 136/140, lm_head 1/1

1 derivatives checked

• mradermacher/POIROT-ECE-1.0-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse