pdevine / discollama-unsloth

tools

6 2 months ago

1 Tag
... /
adapter
544a8eb1c4ff · 14MB
    Metadata
  • adapter.lora.alpha
    32
  • adapter.type
    lora
  • general.architecture
    llama
  • general.file_type
    1
  • general.type
    adapter
  • general.version
    v0.2
  • llama.attention.head_count
    32
  • llama.attention.head_count_kv
    8
    • Tensor
  • Name
    Type
    Shape
    • blk.0
  • blk.0.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.0.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.0.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.0.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.1
  • blk.1.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.1.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.1.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.1.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.2
  • blk.2.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.2.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.2.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.2.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.3
  • blk.3.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.3.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.3.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.3.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.4
  • blk.4.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.4.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.4.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.4.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.5
  • blk.5.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.5.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.5.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.5.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.6
  • blk.6.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.6.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.6.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.6.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.7
  • blk.7.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.7.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.7.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.7.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.8
  • blk.8.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.8.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.8.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.8.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.9
  • blk.9.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.9.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.9.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.9.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.10
  • blk.10.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.10.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.10.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.10.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.11
  • blk.11.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.11.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.11.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.11.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.12
  • blk.12.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.12.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.12.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.12.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.13
  • blk.13.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.13.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.13.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.13.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.14
  • blk.14.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.14.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.14.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.14.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.15
  • blk.15.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.15.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.15.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.15.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.16
  • blk.16.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.16.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.16.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.16.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.17
  • blk.17.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.17.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.17.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.17.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.18
  • blk.18.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.18.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.18.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.18.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.19
  • blk.19.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.19.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.19.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.19.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.20
  • blk.20.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.20.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.20.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.20.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.21
  • blk.21.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.21.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.21.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.21.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.22
  • blk.22.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.22.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.22.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.22.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.23
  • blk.23.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.23.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.23.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.23.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.24
  • blk.24.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.24.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.24.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.24.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.25
  • blk.25.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.25.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.25.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.25.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.26
  • blk.26.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.26.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.26.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.26.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.27
  • blk.27.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.27.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.27.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.27.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.28
  • blk.28.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.28.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.28.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.28.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.29
  • blk.29.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.29.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.29.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.29.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.30
  • blk.30.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.30.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.30.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.30.attn_v.weight.lora_b
    F16
    [16, 1024]
    • blk.31
  • blk.31.attn_q.weight.lora_a
    F16
    [4096, 16]
  • blk.31.attn_q.weight.lora_b
    F16
    [16, 4096]
  • blk.31.attn_v.weight.lora_a
    F16
    [4096, 16]
  • blk.31.attn_v.weight.lora_b
    F16
    [16, 1024]