vicuna

General use chat model based on Llama and Llama 2 with 2K to 16K context sizes.

7b 13b 33b

162.1K 13 months ago

111 Tags
... /
model
f2d50371c678 · 17GB
    Metadata
  • general.architecture
    llama
  • general.file_type
    13
  • general.name
    lmsys
  • general.quantization_version
    2
  • llama.attention.head_count
    52
  • llama.attention.head_count_kv
    52
  • llama.attention.layer_norm_rms_epsilon
    1e-06
  • llama.block_count
    60
  • llama.context_length
    2048
  • llama.embedding_length
    6656
  • llama.feed_forward_length
    17920
  • llama.rope.dimension_count
    128
  • tokenizer.ggml.bos_token_id
    1
  • tokenizer.ggml.eos_token_id
    2
  • tokenizer.ggml.model
    llama
  • tokenizer.ggml.padding_token_id
    0
  • tokenizer.ggml.scores
    [0, 0, 0, 0, 0, ...]
  • tokenizer.ggml.token_type
    [2, 3, 3, 6, 6, ...]
  • tokenizer.ggml.tokens
    [<unk>, <s>, </s>, <0x00>, <0x01>, ...]
    • Tensor
  • Name
    Type
    Shape
  • token_embd.weight
    Q3_K
    [6656, 32000]
    • blk.0
  • blk.0.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.0.attn_norm.weight
    F32
    [6656]
  • blk.0.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.0.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.0.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.0.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.0.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.0.ffn_norm.weight
    F32
    [6656]
  • blk.0.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.1
  • blk.1.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.1.attn_norm.weight
    F32
    [6656]
  • blk.1.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.1.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.1.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.1.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.1.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.1.ffn_norm.weight
    F32
    [6656]
  • blk.1.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.2
  • blk.2.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.2.attn_norm.weight
    F32
    [6656]
  • blk.2.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.2.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.2.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.2.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.2.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.2.ffn_norm.weight
    F32
    [6656]
  • blk.2.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.3
  • blk.3.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.3.attn_norm.weight
    F32
    [6656]
  • blk.3.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.3.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.3.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.3.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.3.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.3.ffn_norm.weight
    F32
    [6656]
  • blk.3.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.4
  • blk.4.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.4.attn_norm.weight
    F32
    [6656]
  • blk.4.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.4.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.4.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.4.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.4.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.4.ffn_norm.weight
    F32
    [6656]
  • blk.4.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.5
  • blk.5.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.5.attn_norm.weight
    F32
    [6656]
  • blk.5.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.5.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.5.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.5.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.5.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.5.ffn_norm.weight
    F32
    [6656]
  • blk.5.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.6
  • blk.6.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.6.attn_norm.weight
    F32
    [6656]
  • blk.6.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.6.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.6.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.6.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.6.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.6.ffn_norm.weight
    F32
    [6656]
  • blk.6.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.7
  • blk.7.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.7.attn_norm.weight
    F32
    [6656]
  • blk.7.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.7.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.7.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.7.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.7.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.7.ffn_norm.weight
    F32
    [6656]
  • blk.7.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.8
  • blk.8.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.8.attn_norm.weight
    F32
    [6656]
  • blk.8.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.8.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.8.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.8.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.8.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.8.ffn_norm.weight
    F32
    [6656]
  • blk.8.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.9
  • blk.9.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.9.attn_norm.weight
    F32
    [6656]
  • blk.9.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.9.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.9.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.9.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.9.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.9.ffn_norm.weight
    F32
    [6656]
  • blk.9.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.10
  • blk.10.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.10.attn_norm.weight
    F32
    [6656]
  • blk.10.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.10.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.10.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.10.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.10.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.10.ffn_norm.weight
    F32
    [6656]
  • blk.10.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.11
  • blk.11.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.11.attn_norm.weight
    F32
    [6656]
  • blk.11.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.11.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.11.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.11.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.11.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.11.ffn_norm.weight
    F32
    [6656]
  • blk.11.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.12
  • blk.12.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.12.attn_norm.weight
    F32
    [6656]
  • blk.12.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.12.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.12.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.12.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.12.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.12.ffn_norm.weight
    F32
    [6656]
  • blk.12.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.13
  • blk.13.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.13.attn_norm.weight
    F32
    [6656]
  • blk.13.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.13.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.13.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.13.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.13.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.13.ffn_norm.weight
    F32
    [6656]
  • blk.13.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.14
  • blk.14.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.14.attn_norm.weight
    F32
    [6656]
  • blk.14.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.14.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.14.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.14.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.14.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.14.ffn_norm.weight
    F32
    [6656]
  • blk.14.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.15
  • blk.15.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.15.attn_norm.weight
    F32
    [6656]
  • blk.15.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.15.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.15.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.15.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.15.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.15.ffn_norm.weight
    F32
    [6656]
  • blk.15.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.16
  • blk.16.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.16.attn_norm.weight
    F32
    [6656]
  • blk.16.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.16.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.16.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.16.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.16.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.16.ffn_norm.weight
    F32
    [6656]
  • blk.16.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.17
  • blk.17.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.17.attn_norm.weight
    F32
    [6656]
  • blk.17.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.17.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.17.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.17.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.17.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.17.ffn_norm.weight
    F32
    [6656]
  • blk.17.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.18
  • blk.18.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.18.attn_norm.weight
    F32
    [6656]
  • blk.18.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.18.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.18.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.18.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.18.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.18.ffn_norm.weight
    F32
    [6656]
  • blk.18.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.19
  • blk.19.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.19.attn_norm.weight
    F32
    [6656]
  • blk.19.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.19.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.19.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.19.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.19.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.19.ffn_norm.weight
    F32
    [6656]
  • blk.19.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.20
  • blk.20.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.20.attn_norm.weight
    F32
    [6656]
  • blk.20.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.20.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.20.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.20.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.20.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.20.ffn_norm.weight
    F32
    [6656]
  • blk.20.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.21
  • blk.21.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.21.attn_norm.weight
    F32
    [6656]
  • blk.21.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.21.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.21.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.21.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.21.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.21.ffn_norm.weight
    F32
    [6656]
  • blk.21.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.22
  • blk.22.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.22.attn_norm.weight
    F32
    [6656]
  • blk.22.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.22.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.22.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.22.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.22.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.22.ffn_norm.weight
    F32
    [6656]
  • blk.22.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.23
  • blk.23.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.23.attn_norm.weight
    F32
    [6656]
  • blk.23.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.23.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.23.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.23.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.23.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.23.ffn_norm.weight
    F32
    [6656]
  • blk.23.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.24
  • blk.24.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.24.attn_norm.weight
    F32
    [6656]
  • blk.24.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.24.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.24.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.24.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.24.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.24.ffn_norm.weight
    F32
    [6656]
  • blk.24.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.25
  • blk.25.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.25.attn_norm.weight
    F32
    [6656]
  • blk.25.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.25.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.25.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.25.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.25.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.25.ffn_norm.weight
    F32
    [6656]
  • blk.25.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.26
  • blk.26.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.26.attn_norm.weight
    F32
    [6656]
  • blk.26.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.26.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.26.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.26.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.26.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.26.ffn_norm.weight
    F32
    [6656]
  • blk.26.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.27
  • blk.27.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.27.attn_norm.weight
    F32
    [6656]
  • blk.27.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.27.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.27.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.27.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.27.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.27.ffn_norm.weight
    F32
    [6656]
  • blk.27.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.28
  • blk.28.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.28.attn_norm.weight
    F32
    [6656]
  • blk.28.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.28.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.28.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.28.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.28.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.28.ffn_norm.weight
    F32
    [6656]
  • blk.28.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.29
  • blk.29.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.29.attn_norm.weight
    F32
    [6656]
  • blk.29.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.29.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.29.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.29.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.29.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.29.ffn_norm.weight
    F32
    [6656]
  • blk.29.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.30
  • blk.30.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.30.attn_norm.weight
    F32
    [6656]
  • blk.30.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.30.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.30.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.30.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.30.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.30.ffn_norm.weight
    F32
    [6656]
  • blk.30.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.31
  • blk.31.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.31.attn_norm.weight
    F32
    [6656]
  • blk.31.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.31.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.31.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.31.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.31.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.31.ffn_norm.weight
    F32
    [6656]
  • blk.31.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.32
  • blk.32.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.32.attn_norm.weight
    F32
    [6656]
  • blk.32.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.32.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.32.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.32.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.32.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.32.ffn_norm.weight
    F32
    [6656]
  • blk.32.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.33
  • blk.33.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.33.attn_norm.weight
    F32
    [6656]
  • blk.33.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.33.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.33.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.33.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.33.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.33.ffn_norm.weight
    F32
    [6656]
  • blk.33.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.34
  • blk.34.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.34.attn_norm.weight
    F32
    [6656]
  • blk.34.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.34.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.34.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.34.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.34.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.34.ffn_norm.weight
    F32
    [6656]
  • blk.34.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.35
  • blk.35.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.35.attn_norm.weight
    F32
    [6656]
  • blk.35.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.35.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.35.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.35.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.35.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.35.ffn_norm.weight
    F32
    [6656]
  • blk.35.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.36
  • blk.36.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.36.attn_norm.weight
    F32
    [6656]
  • blk.36.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.36.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.36.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.36.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.36.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.36.ffn_norm.weight
    F32
    [6656]
  • blk.36.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.37
  • blk.37.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.37.attn_norm.weight
    F32
    [6656]
  • blk.37.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.37.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.37.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.37.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.37.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.37.ffn_norm.weight
    F32
    [6656]
  • blk.37.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.38
  • blk.38.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.38.attn_norm.weight
    F32
    [6656]
  • blk.38.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.38.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.38.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.38.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.38.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.38.ffn_norm.weight
    F32
    [6656]
  • blk.38.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.39
  • blk.39.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.39.attn_norm.weight
    F32
    [6656]
  • blk.39.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.39.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.39.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.39.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.39.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.39.ffn_norm.weight
    F32
    [6656]
  • blk.39.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.40
  • blk.40.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.40.attn_norm.weight
    F32
    [6656]
  • blk.40.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.40.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.40.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.40.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.40.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.40.ffn_norm.weight
    F32
    [6656]
  • blk.40.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.41
  • blk.41.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.41.attn_norm.weight
    F32
    [6656]
  • blk.41.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.41.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.41.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.41.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.41.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.41.ffn_norm.weight
    F32
    [6656]
  • blk.41.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.42
  • blk.42.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.42.attn_norm.weight
    F32
    [6656]
  • blk.42.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.42.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.42.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.42.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.42.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.42.ffn_norm.weight
    F32
    [6656]
  • blk.42.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.43
  • blk.43.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.43.attn_norm.weight
    F32
    [6656]
  • blk.43.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.43.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.43.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.43.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.43.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.43.ffn_norm.weight
    F32
    [6656]
  • blk.43.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.44
  • blk.44.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.44.attn_norm.weight
    F32
    [6656]
  • blk.44.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.44.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.44.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.44.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.44.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.44.ffn_norm.weight
    F32
    [6656]
  • blk.44.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.45
  • blk.45.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.45.attn_norm.weight
    F32
    [6656]
  • blk.45.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.45.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.45.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.45.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.45.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.45.ffn_norm.weight
    F32
    [6656]
  • blk.45.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.46
  • blk.46.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.46.attn_norm.weight
    F32
    [6656]
  • blk.46.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.46.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.46.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.46.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.46.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.46.ffn_norm.weight
    F32
    [6656]
  • blk.46.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.47
  • blk.47.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.47.attn_norm.weight
    F32
    [6656]
  • blk.47.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.47.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.47.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.47.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.47.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.47.ffn_norm.weight
    F32
    [6656]
  • blk.47.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.48
  • blk.48.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.48.attn_norm.weight
    F32
    [6656]
  • blk.48.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.48.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.48.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.48.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.48.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.48.ffn_norm.weight
    F32
    [6656]
  • blk.48.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.49
  • blk.49.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.49.attn_norm.weight
    F32
    [6656]
  • blk.49.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.49.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.49.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.49.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.49.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.49.ffn_norm.weight
    F32
    [6656]
  • blk.49.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.50
  • blk.50.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.50.attn_norm.weight
    F32
    [6656]
  • blk.50.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.50.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.50.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.50.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.50.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.50.ffn_norm.weight
    F32
    [6656]
  • blk.50.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.51
  • blk.51.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.51.attn_norm.weight
    F32
    [6656]
  • blk.51.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.51.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.51.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.51.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.51.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.51.ffn_norm.weight
    F32
    [6656]
  • blk.51.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.52
  • blk.52.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.52.attn_norm.weight
    F32
    [6656]
  • blk.52.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.52.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.52.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.52.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.52.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.52.ffn_norm.weight
    F32
    [6656]
  • blk.52.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.53
  • blk.53.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.53.attn_norm.weight
    F32
    [6656]
  • blk.53.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.53.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.53.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.53.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.53.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.53.ffn_norm.weight
    F32
    [6656]
  • blk.53.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.54
  • blk.54.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.54.attn_norm.weight
    F32
    [6656]
  • blk.54.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.54.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.54.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.54.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.54.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.54.ffn_norm.weight
    F32
    [6656]
  • blk.54.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.55
  • blk.55.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.55.attn_norm.weight
    F32
    [6656]
  • blk.55.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.55.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.55.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.55.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.55.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.55.ffn_norm.weight
    F32
    [6656]
  • blk.55.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.56
  • blk.56.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.56.attn_norm.weight
    F32
    [6656]
  • blk.56.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.56.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.56.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.56.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.56.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.56.ffn_norm.weight
    F32
    [6656]
  • blk.56.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.57
  • blk.57.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.57.attn_norm.weight
    F32
    [6656]
  • blk.57.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.57.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.57.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.57.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.57.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.57.ffn_norm.weight
    F32
    [6656]
  • blk.57.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.58
  • blk.58.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.58.attn_norm.weight
    F32
    [6656]
  • blk.58.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.58.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.58.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.58.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.58.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.58.ffn_norm.weight
    F32
    [6656]
  • blk.58.ffn_up.weight
    Q3_K
    [6656, 17920]
    • blk.59
  • blk.59.attn_k.weight
    Q3_K
    [6656, 6656]
  • blk.59.attn_norm.weight
    F32
    [6656]
  • blk.59.attn_output.weight
    Q5_K
    [6656, 6656]
  • blk.59.attn_q.weight
    Q3_K
    [6656, 6656]
  • blk.59.attn_v.weight
    Q5_K
    [6656, 6656]
  • blk.59.ffn_down.weight
    Q5_K
    [17920, 6656]
  • blk.59.ffn_gate.weight
    Q3_K
    [6656, 17920]
  • blk.59.ffn_norm.weight
    F32
    [6656]
  • blk.59.ffn_up.weight
    Q3_K
    [6656, 17920]
  • output.weight
    Q6_K
    [6656, 32000]
  • output_norm.weight
    F32
    [6656]