bnksys / baia

embedding

119 3 months ago

36 Tags
... /
model
25a1a713cc7e · 618MB
    Metadata
  • bert.attention.causal
    false
  • bert.attention.head_count
    16
  • bert.attention.layer_norm_epsilon
    1e-05
  • bert.block_count
    24
  • bert.context_length
    8194
  • bert.embedding_length
    1024
  • bert.feed_forward_length
    4096
  • bert.pooling_type
    2
  • general.architecture
    bert
  • general.file_type
    7
  • general.name
    bge-m3
  • general.quantization_version
    2
  • tokenizer.ggml.bos_token_id
    0
  • tokenizer.ggml.cls_token_id
    0
  • tokenizer.ggml.eos_token_id
    2
  • tokenizer.ggml.mask_token_id
    250001
  • tokenizer.ggml.model
    llama
  • tokenizer.ggml.padding_token_id
    1
  • tokenizer.ggml.pre
    default
  • tokenizer.ggml.scores
    [0, 0, 0, 0, -3.4635427, ...]
  • tokenizer.ggml.seperator_token_id
    2
  • tokenizer.ggml.token_type
    [3, 2, 3, 2, 1, ...]
  • tokenizer.ggml.token_type_count
    1
  • tokenizer.ggml.tokens
    [<s>, <pad>, </s>, <unk>, ,, ...]
  • tokenizer.ggml.unknown_token_id
    3
    • Tensor
  • Name
    Type
    Shape
  • token_embd.weight
    Q8_0
    [1024, 250002]
    • blk.0
  • blk.0.attn_k.bias
    F32
    [1024]
  • blk.0.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.0.attn_output.bias
    F32
    [1024]
  • blk.0.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.0.attn_output_norm.bias
    F32
    [1024]
  • blk.0.attn_output_norm.weight
    F32
    [1024]
  • blk.0.attn_q.bias
    F32
    [1024]
  • blk.0.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.0.attn_v.bias
    F32
    [1024]
  • blk.0.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.0.ffn_down.bias
    F32
    [1024]
  • blk.0.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.0.ffn_up.bias
    F32
    [4096]
  • blk.0.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.0.layer_output_norm.bias
    F32
    [1024]
  • blk.0.layer_output_norm.weight
    F32
    [1024]
    • blk.1
  • blk.1.attn_k.bias
    F32
    [1024]
  • blk.1.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.1.attn_output.bias
    F32
    [1024]
  • blk.1.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.1.attn_output_norm.bias
    F32
    [1024]
  • blk.1.attn_output_norm.weight
    F32
    [1024]
  • blk.1.attn_q.bias
    F32
    [1024]
  • blk.1.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.1.attn_v.bias
    F32
    [1024]
  • blk.1.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.1.ffn_down.bias
    F32
    [1024]
  • blk.1.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.1.ffn_up.bias
    F32
    [4096]
  • blk.1.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.1.layer_output_norm.bias
    F32
    [1024]
  • blk.1.layer_output_norm.weight
    F32
    [1024]
    • blk.2
  • blk.2.attn_k.bias
    F32
    [1024]
  • blk.2.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.2.attn_output.bias
    F32
    [1024]
  • blk.2.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.2.attn_output_norm.bias
    F32
    [1024]
  • blk.2.attn_output_norm.weight
    F32
    [1024]
  • blk.2.attn_q.bias
    F32
    [1024]
  • blk.2.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.2.attn_v.bias
    F32
    [1024]
  • blk.2.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.2.ffn_down.bias
    F32
    [1024]
  • blk.2.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.2.ffn_up.bias
    F32
    [4096]
  • blk.2.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.2.layer_output_norm.bias
    F32
    [1024]
  • blk.2.layer_output_norm.weight
    F32
    [1024]
    • blk.3
  • blk.3.attn_k.bias
    F32
    [1024]
  • blk.3.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.3.attn_output.bias
    F32
    [1024]
  • blk.3.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.3.attn_output_norm.bias
    F32
    [1024]
  • blk.3.attn_output_norm.weight
    F32
    [1024]
  • blk.3.attn_q.bias
    F32
    [1024]
  • blk.3.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.3.attn_v.bias
    F32
    [1024]
  • blk.3.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.3.ffn_down.bias
    F32
    [1024]
  • blk.3.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.3.ffn_up.bias
    F32
    [4096]
  • blk.3.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.3.layer_output_norm.bias
    F32
    [1024]
  • blk.3.layer_output_norm.weight
    F32
    [1024]
    • blk.4
  • blk.4.attn_k.bias
    F32
    [1024]
  • blk.4.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.4.attn_output.bias
    F32
    [1024]
  • blk.4.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.4.attn_output_norm.bias
    F32
    [1024]
  • blk.4.attn_output_norm.weight
    F32
    [1024]
  • blk.4.attn_q.bias
    F32
    [1024]
  • blk.4.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.4.attn_v.bias
    F32
    [1024]
  • blk.4.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.4.ffn_down.bias
    F32
    [1024]
  • blk.4.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.4.ffn_up.bias
    F32
    [4096]
  • blk.4.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.4.layer_output_norm.bias
    F32
    [1024]
  • blk.4.layer_output_norm.weight
    F32
    [1024]
    • blk.5
  • blk.5.attn_k.bias
    F32
    [1024]
  • blk.5.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.5.attn_output.bias
    F32
    [1024]
  • blk.5.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.5.attn_output_norm.bias
    F32
    [1024]
  • blk.5.attn_output_norm.weight
    F32
    [1024]
  • blk.5.attn_q.bias
    F32
    [1024]
  • blk.5.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.5.attn_v.bias
    F32
    [1024]
  • blk.5.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.5.ffn_down.bias
    F32
    [1024]
  • blk.5.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.5.ffn_up.bias
    F32
    [4096]
  • blk.5.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.5.layer_output_norm.bias
    F32
    [1024]
  • blk.5.layer_output_norm.weight
    F32
    [1024]
    • blk.6
  • blk.6.attn_k.bias
    F32
    [1024]
  • blk.6.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.6.attn_output.bias
    F32
    [1024]
  • blk.6.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.6.attn_output_norm.bias
    F32
    [1024]
  • blk.6.attn_output_norm.weight
    F32
    [1024]
  • blk.6.attn_q.bias
    F32
    [1024]
  • blk.6.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.6.attn_v.bias
    F32
    [1024]
  • blk.6.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.6.ffn_down.bias
    F32
    [1024]
  • blk.6.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.6.ffn_up.bias
    F32
    [4096]
  • blk.6.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.6.layer_output_norm.bias
    F32
    [1024]
  • blk.6.layer_output_norm.weight
    F32
    [1024]
    • blk.7
  • blk.7.attn_k.bias
    F32
    [1024]
  • blk.7.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.7.attn_output.bias
    F32
    [1024]
  • blk.7.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.7.attn_output_norm.bias
    F32
    [1024]
  • blk.7.attn_output_norm.weight
    F32
    [1024]
  • blk.7.attn_q.bias
    F32
    [1024]
  • blk.7.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.7.attn_v.bias
    F32
    [1024]
  • blk.7.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.7.ffn_down.bias
    F32
    [1024]
  • blk.7.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.7.ffn_up.bias
    F32
    [4096]
  • blk.7.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.7.layer_output_norm.bias
    F32
    [1024]
  • blk.7.layer_output_norm.weight
    F32
    [1024]
    • blk.8
  • blk.8.attn_k.bias
    F32
    [1024]
  • blk.8.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.8.attn_output.bias
    F32
    [1024]
  • blk.8.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.8.attn_output_norm.bias
    F32
    [1024]
  • blk.8.attn_output_norm.weight
    F32
    [1024]
  • blk.8.attn_q.bias
    F32
    [1024]
  • blk.8.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.8.attn_v.bias
    F32
    [1024]
  • blk.8.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.8.ffn_down.bias
    F32
    [1024]
  • blk.8.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.8.ffn_up.bias
    F32
    [4096]
  • blk.8.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.8.layer_output_norm.bias
    F32
    [1024]
  • blk.8.layer_output_norm.weight
    F32
    [1024]
    • blk.9
  • blk.9.attn_k.bias
    F32
    [1024]
  • blk.9.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.9.attn_output.bias
    F32
    [1024]
  • blk.9.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.9.attn_output_norm.bias
    F32
    [1024]
  • blk.9.attn_output_norm.weight
    F32
    [1024]
  • blk.9.attn_q.bias
    F32
    [1024]
  • blk.9.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.9.attn_v.bias
    F32
    [1024]
  • blk.9.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.9.ffn_down.bias
    F32
    [1024]
  • blk.9.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.9.ffn_up.bias
    F32
    [4096]
  • blk.9.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.9.layer_output_norm.bias
    F32
    [1024]
  • blk.9.layer_output_norm.weight
    F32
    [1024]
    • blk.10
  • blk.10.attn_k.bias
    F32
    [1024]
  • blk.10.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.10.attn_output.bias
    F32
    [1024]
  • blk.10.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.10.attn_output_norm.bias
    F32
    [1024]
  • blk.10.attn_output_norm.weight
    F32
    [1024]
  • blk.10.attn_q.bias
    F32
    [1024]
  • blk.10.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.10.attn_v.bias
    F32
    [1024]
  • blk.10.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.10.ffn_down.bias
    F32
    [1024]
  • blk.10.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.10.ffn_up.bias
    F32
    [4096]
  • blk.10.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.10.layer_output_norm.bias
    F32
    [1024]
  • blk.10.layer_output_norm.weight
    F32
    [1024]
    • blk.11
  • blk.11.attn_k.bias
    F32
    [1024]
  • blk.11.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.11.attn_output.bias
    F32
    [1024]
  • blk.11.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.11.attn_output_norm.bias
    F32
    [1024]
  • blk.11.attn_output_norm.weight
    F32
    [1024]
  • blk.11.attn_q.bias
    F32
    [1024]
  • blk.11.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.11.attn_v.bias
    F32
    [1024]
  • blk.11.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.11.ffn_down.bias
    F32
    [1024]
  • blk.11.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.11.ffn_up.bias
    F32
    [4096]
  • blk.11.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.11.layer_output_norm.bias
    F32
    [1024]
  • blk.11.layer_output_norm.weight
    F32
    [1024]
    • blk.12
  • blk.12.attn_k.bias
    F32
    [1024]
  • blk.12.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.12.attn_output.bias
    F32
    [1024]
  • blk.12.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.12.attn_output_norm.bias
    F32
    [1024]
  • blk.12.attn_output_norm.weight
    F32
    [1024]
  • blk.12.attn_q.bias
    F32
    [1024]
  • blk.12.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.12.attn_v.bias
    F32
    [1024]
  • blk.12.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.12.ffn_down.bias
    F32
    [1024]
  • blk.12.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.12.ffn_up.bias
    F32
    [4096]
  • blk.12.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.12.layer_output_norm.bias
    F32
    [1024]
  • blk.12.layer_output_norm.weight
    F32
    [1024]
    • blk.13
  • blk.13.attn_k.bias
    F32
    [1024]
  • blk.13.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.13.attn_output.bias
    F32
    [1024]
  • blk.13.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.13.attn_output_norm.bias
    F32
    [1024]
  • blk.13.attn_output_norm.weight
    F32
    [1024]
  • blk.13.attn_q.bias
    F32
    [1024]
  • blk.13.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.13.attn_v.bias
    F32
    [1024]
  • blk.13.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.13.ffn_down.bias
    F32
    [1024]
  • blk.13.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.13.ffn_up.bias
    F32
    [4096]
  • blk.13.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.13.layer_output_norm.bias
    F32
    [1024]
  • blk.13.layer_output_norm.weight
    F32
    [1024]
    • blk.14
  • blk.14.attn_k.bias
    F32
    [1024]
  • blk.14.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.14.attn_output.bias
    F32
    [1024]
  • blk.14.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.14.attn_output_norm.bias
    F32
    [1024]
  • blk.14.attn_output_norm.weight
    F32
    [1024]
  • blk.14.attn_q.bias
    F32
    [1024]
  • blk.14.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.14.attn_v.bias
    F32
    [1024]
  • blk.14.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.14.ffn_down.bias
    F32
    [1024]
  • blk.14.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.14.ffn_up.bias
    F32
    [4096]
  • blk.14.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.14.layer_output_norm.bias
    F32
    [1024]
  • blk.14.layer_output_norm.weight
    F32
    [1024]
    • blk.15
  • blk.15.attn_k.bias
    F32
    [1024]
  • blk.15.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.15.attn_output.bias
    F32
    [1024]
  • blk.15.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.15.attn_output_norm.bias
    F32
    [1024]
  • blk.15.attn_output_norm.weight
    F32
    [1024]
  • blk.15.attn_q.bias
    F32
    [1024]
  • blk.15.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.15.attn_v.bias
    F32
    [1024]
  • blk.15.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.15.ffn_down.bias
    F32
    [1024]
  • blk.15.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.15.ffn_up.bias
    F32
    [4096]
  • blk.15.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.15.layer_output_norm.bias
    F32
    [1024]
  • blk.15.layer_output_norm.weight
    F32
    [1024]
    • blk.16
  • blk.16.attn_k.bias
    F32
    [1024]
  • blk.16.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.16.attn_output.bias
    F32
    [1024]
  • blk.16.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.16.attn_output_norm.bias
    F32
    [1024]
  • blk.16.attn_output_norm.weight
    F32
    [1024]
  • blk.16.attn_q.bias
    F32
    [1024]
  • blk.16.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.16.attn_v.bias
    F32
    [1024]
  • blk.16.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.16.ffn_down.bias
    F32
    [1024]
  • blk.16.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.16.ffn_up.bias
    F32
    [4096]
  • blk.16.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.16.layer_output_norm.bias
    F32
    [1024]
  • blk.16.layer_output_norm.weight
    F32
    [1024]
    • blk.17
  • blk.17.attn_k.bias
    F32
    [1024]
  • blk.17.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.17.attn_output.bias
    F32
    [1024]
  • blk.17.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.17.attn_output_norm.bias
    F32
    [1024]
  • blk.17.attn_output_norm.weight
    F32
    [1024]
  • blk.17.attn_q.bias
    F32
    [1024]
  • blk.17.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.17.attn_v.bias
    F32
    [1024]
  • blk.17.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.17.ffn_down.bias
    F32
    [1024]
  • blk.17.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.17.ffn_up.bias
    F32
    [4096]
  • blk.17.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.17.layer_output_norm.bias
    F32
    [1024]
  • blk.17.layer_output_norm.weight
    F32
    [1024]
    • blk.18
  • blk.18.attn_k.bias
    F32
    [1024]
  • blk.18.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.18.attn_output.bias
    F32
    [1024]
  • blk.18.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.18.attn_output_norm.bias
    F32
    [1024]
  • blk.18.attn_output_norm.weight
    F32
    [1024]
  • blk.18.attn_q.bias
    F32
    [1024]
  • blk.18.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.18.attn_v.bias
    F32
    [1024]
  • blk.18.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.18.ffn_down.bias
    F32
    [1024]
  • blk.18.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.18.ffn_up.bias
    F32
    [4096]
  • blk.18.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.18.layer_output_norm.bias
    F32
    [1024]
  • blk.18.layer_output_norm.weight
    F32
    [1024]
    • blk.19
  • blk.19.attn_k.bias
    F32
    [1024]
  • blk.19.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.19.attn_output.bias
    F32
    [1024]
  • blk.19.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.19.attn_output_norm.bias
    F32
    [1024]
  • blk.19.attn_output_norm.weight
    F32
    [1024]
  • blk.19.attn_q.bias
    F32
    [1024]
  • blk.19.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.19.attn_v.bias
    F32
    [1024]
  • blk.19.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.19.ffn_down.bias
    F32
    [1024]
  • blk.19.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.19.ffn_up.bias
    F32
    [4096]
  • blk.19.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.19.layer_output_norm.bias
    F32
    [1024]
  • blk.19.layer_output_norm.weight
    F32
    [1024]
    • blk.20
  • blk.20.attn_k.bias
    F32
    [1024]
  • blk.20.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.20.attn_output.bias
    F32
    [1024]
  • blk.20.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.20.attn_output_norm.bias
    F32
    [1024]
  • blk.20.attn_output_norm.weight
    F32
    [1024]
  • blk.20.attn_q.bias
    F32
    [1024]
  • blk.20.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.20.attn_v.bias
    F32
    [1024]
  • blk.20.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.20.ffn_down.bias
    F32
    [1024]
  • blk.20.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.20.ffn_up.bias
    F32
    [4096]
  • blk.20.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.20.layer_output_norm.bias
    F32
    [1024]
  • blk.20.layer_output_norm.weight
    F32
    [1024]
    • blk.21
  • blk.21.attn_k.bias
    F32
    [1024]
  • blk.21.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.21.attn_output.bias
    F32
    [1024]
  • blk.21.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.21.attn_output_norm.bias
    F32
    [1024]
  • blk.21.attn_output_norm.weight
    F32
    [1024]
  • blk.21.attn_q.bias
    F32
    [1024]
  • blk.21.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.21.attn_v.bias
    F32
    [1024]
  • blk.21.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.21.ffn_down.bias
    F32
    [1024]
  • blk.21.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.21.ffn_up.bias
    F32
    [4096]
  • blk.21.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.21.layer_output_norm.bias
    F32
    [1024]
  • blk.21.layer_output_norm.weight
    F32
    [1024]
    • blk.22
  • blk.22.attn_k.bias
    F32
    [1024]
  • blk.22.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.22.attn_output.bias
    F32
    [1024]
  • blk.22.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.22.attn_output_norm.bias
    F32
    [1024]
  • blk.22.attn_output_norm.weight
    F32
    [1024]
  • blk.22.attn_q.bias
    F32
    [1024]
  • blk.22.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.22.attn_v.bias
    F32
    [1024]
  • blk.22.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.22.ffn_down.bias
    F32
    [1024]
  • blk.22.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.22.ffn_up.bias
    F32
    [4096]
  • blk.22.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.22.layer_output_norm.bias
    F32
    [1024]
  • blk.22.layer_output_norm.weight
    F32
    [1024]
    • blk.23
  • blk.23.attn_k.bias
    F32
    [1024]
  • blk.23.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.23.attn_output.bias
    F32
    [1024]
  • blk.23.attn_output.weight
    Q8_0
    [1024, 1024]
  • blk.23.attn_output_norm.bias
    F32
    [1024]
  • blk.23.attn_output_norm.weight
    F32
    [1024]
  • blk.23.attn_q.bias
    F32
    [1024]
  • blk.23.attn_q.weight
    Q8_0
    [1024, 1024]
  • blk.23.attn_v.bias
    F32
    [1024]
  • blk.23.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.23.ffn_down.bias
    F32
    [1024]
  • blk.23.ffn_down.weight
    Q8_0
    [4096, 1024]
  • blk.23.ffn_up.bias
    F32
    [4096]
  • blk.23.ffn_up.weight
    Q8_0
    [1024, 4096]
  • blk.23.layer_output_norm.bias
    F32
    [1024]
  • blk.23.layer_output_norm.weight
    F32
    [1024]
  • position_embd.weight
    F16
    [1024, 8194]
  • token_embd_norm.bias
    F32
    [1024]
  • token_embd_norm.weight
    F32
    [1024]
  • token_types.weight
    F32
    [1024]