angelfencer/llama32-steer/adapter

angelfencer/

llama32-steer:latest

46 Downloads Updated 1 year ago

Llama 3.2 with nvidia/HelpSteer2 LoRA this thing is one smart llama

tools

llama32-steer:latest ... /

adapter

9e764df656ff · 4.6MB

Metadata

general.architecture

llama

llama
general.file_type

F16

F16
llama.attention.head_count

24

24
llama.attention.head_count_kv

8

8
adapter.lora.alpha

32

32
adapter.type

lora

lora

Tensor

Name

Type

Shape

blk.0

blk.0.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.0.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.0.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.0.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.1

blk.1.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.1.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.1.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.1.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.2

blk.2.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.2.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.2.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.2.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.3

blk.3.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.3.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.3.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.3.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.4

blk.4.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.4.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.4.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.4.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.5

blk.5.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.5.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.5.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.5.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.6

blk.6.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.6.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.6.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.6.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.7

blk.7.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.7.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.7.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.7.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.8

blk.8.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.8.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.8.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.8.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.9

blk.9.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.9.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.9.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.9.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.10

blk.10.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.10.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.10.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.10.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.11

blk.11.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.11.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.11.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.11.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.12

blk.12.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.12.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.12.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.12.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.13

blk.13.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.13.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.13.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.13.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.14

blk.14.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.14.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.14.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.14.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.15

blk.15.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.15.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.15.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.15.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.16

blk.16.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.16.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.16.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.16.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.17

blk.17.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.17.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.17.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.17.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.18

blk.18.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.18.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.18.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.18.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.19

blk.19.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.19.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.19.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.19.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.20

blk.20.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.20.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.20.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.20.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.21

blk.21.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.21.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.21.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.21.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.22

blk.22.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.22.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.22.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.22.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.23

blk.23.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.23.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.23.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.23.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.24

blk.24.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.24.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.24.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.24.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.25

blk.25.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.25.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.25.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.25.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.26

blk.26.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.26.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.26.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.26.attn_v.weight.lora_b

F16

F16

[8, 1024]

blk.27

blk.27.attn_q.weight.lora_a

F16

F16

[3072, 8]
blk.27.attn_q.weight.lora_b

F16

F16

[8, 3072]
blk.27.attn_v.weight.lora_a

F16

F16

[3072, 8]
blk.27.attn_v.weight.lora_b

F16

F16

[8, 1024]