AaryanDahal/aaryans_personal

AaryanDahal/

aaryans_personal_llm:latest

4 Downloads Updated 5 months ago

aaryans_personal_llm:latest ... /

adapter

a2ab0ae9c9a5 · 361MB

Metadata

general.architecture

llama

llama
general.file_type

F16

F16
llama.attention.head_count

32

32
llama.attention.head_count_kv

8

8
adapter.lora.alpha

512

512
adapter.type

lora

lora

Tensor

Name

Type

Shape

blk.0

blk.0.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.0.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.0.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.0.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.0.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.0.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.0.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.0.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.0.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.0.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.0.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.0.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.0.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.0.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.1

blk.1.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.1.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.1.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.1.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.1.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.1.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.1.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.1.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.1.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.1.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.1.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.1.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.1.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.1.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.2

blk.2.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.2.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.2.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.2.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.2.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.2.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.2.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.2.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.2.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.2.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.2.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.2.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.2.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.2.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.3

blk.3.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.3.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.3.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.3.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.3.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.3.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.3.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.3.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.3.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.3.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.3.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.3.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.3.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.3.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.4

blk.4.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.4.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.4.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.4.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.4.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.4.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.4.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.4.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.4.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.4.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.4.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.4.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.4.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.4.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.5

blk.5.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.5.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.5.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.5.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.5.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.5.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.5.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.5.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.5.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.5.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.5.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.5.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.5.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.5.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.6

blk.6.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.6.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.6.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.6.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.6.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.6.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.6.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.6.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.6.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.6.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.6.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.6.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.6.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.6.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.7

blk.7.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.7.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.7.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.7.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.7.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.7.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.7.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.7.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.7.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.7.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.7.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.7.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.7.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.7.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.8

blk.8.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.8.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.8.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.8.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.8.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.8.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.8.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.8.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.8.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.8.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.8.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.8.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.8.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.8.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.9

blk.9.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.9.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.9.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.9.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.9.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.9.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.9.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.9.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.9.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.9.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.9.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.9.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.9.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.9.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.10

blk.10.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.10.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.10.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.10.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.10.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.10.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.10.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.10.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.10.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.10.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.10.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.10.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.10.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.10.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.11

blk.11.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.11.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.11.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.11.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.11.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.11.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.11.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.11.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.11.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.11.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.11.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.11.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.11.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.11.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.12

blk.12.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.12.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.12.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.12.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.12.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.12.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.12.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.12.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.12.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.12.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.12.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.12.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.12.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.12.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.13

blk.13.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.13.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.13.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.13.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.13.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.13.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.13.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.13.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.13.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.13.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.13.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.13.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.13.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.13.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.14

blk.14.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.14.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.14.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.14.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.14.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.14.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.14.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.14.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.14.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.14.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.14.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.14.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.14.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.14.ffn_up.weight.lora_b

F16

F16

[256, 8192]

blk.15

blk.15.attn_k.weight.lora_a

F16

F16

[2048, 256]
blk.15.attn_k.weight.lora_b

F16

F16

[256, 512]
blk.15.attn_output.weight.lora_a

F16

F16

[2048, 256]
blk.15.attn_output.weight.lora_b

F16

F16

[256, 2048]
blk.15.attn_q.weight.lora_a

F16

F16

[2048, 256]
blk.15.attn_q.weight.lora_b

F16

F16

[256, 2048]
blk.15.attn_v.weight.lora_a

F16

F16

[2048, 256]
blk.15.attn_v.weight.lora_b

F16

F16

[256, 512]
blk.15.ffn_down.weight.lora_a

F16

F16

[8192, 256]
blk.15.ffn_down.weight.lora_b

F16

F16

[256, 2048]
blk.15.ffn_gate.weight.lora_a

F16

F16

[2048, 256]
blk.15.ffn_gate.weight.lora_b

F16

F16

[256, 8192]
blk.15.ffn_up.weight.lora_a

F16

F16

[2048, 256]
blk.15.ffn_up.weight.lora_b

F16

F16

[256, 8192]