▀█▀ █▄ ▄█ █ █ █▀▀ █▀▄ █▀▀ ▀█▀ ▀█▀ █▀▀ █ █ ▀ █ █▀█ █▀▀ █▀▄ █▀▀ █ █ █ ▀▀▀ ▀ ▀ ▀ ▀ ▀▀▀ ▀ ▀ ▀▀▀ ▀ ▀▀▀ ▀▀▀

Check Quants

Refusals (this model): 22/100
Original (MTSAIR/Kodify-Nano): 99/100
KL divergence: 0.0822

Parameters

direction_index = per layer
attn.o_proj.max_weight = 1.09
attn.o_proj.max_weight_position = 24.55
attn.o_proj.min_weight = 0.12
attn.o_proj.min_weight_distance = 12.35
mlp.down_proj.max_weight = 1.22
mlp.down_proj.max_weight_position = 16.37
mlp.down_proj.min_weight = 0.97
mlp.down_proj.min_weight_distance = 7.45

Downloads last month: 2

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for hereticness/Heretic-Kodify-Nano

Base model

MTSAIR/Kodify-Nano

Finetuned

(1)

this model

Quantizations

2 models