s1gnature/ laya:latest

31 2 days ago

embedding
... /
model
f10551bbe259 · 791MB
    Metadata
  • general.architecture
    modern-bert
  • general.file_type
    F16
  • modern-bert.attention.causal
    false
  • modern-bert.attention.head_count
    16
  • modern-bert.attention.layer_norm_epsilon
    1e-05
  • modern-bert.attention.layer_norm_rms_epsilon
    1e-05
  • modern-bert.attention.sliding_window
    128
  • modern-bert.attention.sliding_window_pattern
    3
  • modern-bert.block_count
    28
  • modern-bert.context_length
    8192
  • modern-bert.embedding_length
    1024
  • modern-bert.feed_forward_length
    2624
  • modern-bert.hidden_activation
    gelu
  • modern-bert.rope.freq_base
    160000
  • modern-bert.rope.freq_base_swa
    10000
  • modern-bert.rope.scaling.type
    none
  • modern-bert.vocab_size
    50368
  • tokenizer.ggml.add_bos_token
    true
  • tokenizer.ggml.add_eos_token
    true
  • tokenizer.ggml.add_sep_token
    false
  • tokenizer.ggml.bos_token_id
    50281
  • tokenizer.ggml.eos_token_id
    50282
  • tokenizer.ggml.mask_token_id
    50284
  • tokenizer.ggml.merges
    [Ġ Ġ, Ġ t, Ġ a, h e, i n, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    50283
  • tokenizer.ggml.pre
    modern-bert
  • tokenizer.ggml.seperator_token_id
    50282
  • tokenizer.ggml.token_type
    [4, 3, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [|||IP_ADDRESS|||, <|padding|>, !, ", #, ...]
  • tokenizer.ggml.unknown_token_id
    50280
    • Tensor
  • token_embd.weight
    F16
    [1024, 50368]
    • blk.0
  • blk.0.attn_output.weight
    F16
    [1024, 1024]
  • blk.0.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.0.ffn_down.weight
    F16
    [2624, 1024]
  • blk.0.ffn_norm.weight
    F32
    [1024]
  • blk.0.ffn_up.weight
    F16
    [1024, 5248]
    • blk.1
  • blk.1.attn_norm.weight
    F32
    [1024]
  • blk.1.attn_output.weight
    F16
    [1024, 1024]
  • blk.1.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.1.ffn_down.weight
    F16
    [2624, 1024]
  • blk.1.ffn_norm.weight
    F32
    [1024]
  • blk.1.ffn_up.weight
    F16
    [1024, 5248]
    • blk.2
  • blk.2.attn_norm.weight
    F32
    [1024]
  • blk.2.attn_output.weight
    F16
    [1024, 1024]
  • blk.2.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.2.ffn_down.weight
    F16
    [2624, 1024]
  • blk.2.ffn_norm.weight
    F32
    [1024]
  • blk.2.ffn_up.weight
    F16
    [1024, 5248]
    • blk.3
  • blk.3.attn_norm.weight
    F32
    [1024]
  • blk.3.attn_output.weight
    F16
    [1024, 1024]
  • blk.3.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.3.ffn_down.weight
    F16
    [2624, 1024]
  • blk.3.ffn_norm.weight
    F32
    [1024]
  • blk.3.ffn_up.weight
    F16
    [1024, 5248]
    • blk.4
  • blk.4.attn_norm.weight
    F32
    [1024]
  • blk.4.attn_output.weight
    F16
    [1024, 1024]
  • blk.4.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.4.ffn_down.weight
    F16
    [2624, 1024]
  • blk.4.ffn_norm.weight
    F32
    [1024]
  • blk.4.ffn_up.weight
    F16
    [1024, 5248]
    • blk.5
  • blk.5.attn_norm.weight
    F32
    [1024]
  • blk.5.attn_output.weight
    F16
    [1024, 1024]
  • blk.5.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.5.ffn_down.weight
    F16
    [2624, 1024]
  • blk.5.ffn_norm.weight
    F32
    [1024]
  • blk.5.ffn_up.weight
    F16
    [1024, 5248]
    • blk.6
  • blk.6.attn_norm.weight
    F32
    [1024]
  • blk.6.attn_output.weight
    F16
    [1024, 1024]
  • blk.6.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.6.ffn_down.weight
    F16
    [2624, 1024]
  • blk.6.ffn_norm.weight
    F32
    [1024]
  • blk.6.ffn_up.weight
    F16
    [1024, 5248]
    • blk.7
  • blk.7.attn_norm.weight
    F32
    [1024]
  • blk.7.attn_output.weight
    F16
    [1024, 1024]
  • blk.7.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.7.ffn_down.weight
    F16
    [2624, 1024]
  • blk.7.ffn_norm.weight
    F32
    [1024]
  • blk.7.ffn_up.weight
    F16
    [1024, 5248]
    • blk.8
  • blk.8.attn_norm.weight
    F32
    [1024]
  • blk.8.attn_output.weight
    F16
    [1024, 1024]
  • blk.8.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.8.ffn_down.weight
    F16
    [2624, 1024]
  • blk.8.ffn_norm.weight
    F32
    [1024]
  • blk.8.ffn_up.weight
    F16
    [1024, 5248]
    • blk.9
  • blk.9.attn_norm.weight
    F32
    [1024]
  • blk.9.attn_output.weight
    F16
    [1024, 1024]
  • blk.9.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.9.ffn_down.weight
    F16
    [2624, 1024]
  • blk.9.ffn_norm.weight
    F32
    [1024]
  • blk.9.ffn_up.weight
    F16
    [1024, 5248]
    • blk.10
  • blk.10.attn_norm.weight
    F32
    [1024]
  • blk.10.attn_output.weight
    F16
    [1024, 1024]
  • blk.10.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.10.ffn_down.weight
    F16
    [2624, 1024]
  • blk.10.ffn_norm.weight
    F32
    [1024]
  • blk.10.ffn_up.weight
    F16
    [1024, 5248]
    • blk.11
  • blk.11.attn_norm.weight
    F32
    [1024]
  • blk.11.attn_output.weight
    F16
    [1024, 1024]
  • blk.11.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.11.ffn_down.weight
    F16
    [2624, 1024]
  • blk.11.ffn_norm.weight
    F32
    [1024]
  • blk.11.ffn_up.weight
    F16
    [1024, 5248]
    • blk.12
  • blk.12.attn_norm.weight
    F32
    [1024]
  • blk.12.attn_output.weight
    F16
    [1024, 1024]
  • blk.12.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.12.ffn_down.weight
    F16
    [2624, 1024]
  • blk.12.ffn_norm.weight
    F32
    [1024]
  • blk.12.ffn_up.weight
    F16
    [1024, 5248]
    • blk.13
  • blk.13.attn_norm.weight
    F32
    [1024]
  • blk.13.attn_output.weight
    F16
    [1024, 1024]
  • blk.13.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.13.ffn_down.weight
    F16
    [2624, 1024]
  • blk.13.ffn_norm.weight
    F32
    [1024]
  • blk.13.ffn_up.weight
    F16
    [1024, 5248]
    • blk.14
  • blk.14.attn_norm.weight
    F32
    [1024]
  • blk.14.attn_output.weight
    F16
    [1024, 1024]
  • blk.14.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.14.ffn_down.weight
    F16
    [2624, 1024]
  • blk.14.ffn_norm.weight
    F32
    [1024]
  • blk.14.ffn_up.weight
    F16
    [1024, 5248]
    • blk.15
  • blk.15.attn_norm.weight
    F32
    [1024]
  • blk.15.attn_output.weight
    F16
    [1024, 1024]
  • blk.15.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.15.ffn_down.weight
    F16
    [2624, 1024]
  • blk.15.ffn_norm.weight
    F32
    [1024]
  • blk.15.ffn_up.weight
    F16
    [1024, 5248]
    • blk.16
  • blk.16.attn_norm.weight
    F32
    [1024]
  • blk.16.attn_output.weight
    F16
    [1024, 1024]
  • blk.16.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.16.ffn_down.weight
    F16
    [2624, 1024]
  • blk.16.ffn_norm.weight
    F32
    [1024]
  • blk.16.ffn_up.weight
    F16
    [1024, 5248]
    • blk.17
  • blk.17.attn_norm.weight
    F32
    [1024]
  • blk.17.attn_output.weight
    F16
    [1024, 1024]
  • blk.17.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.17.ffn_down.weight
    F16
    [2624, 1024]
  • blk.17.ffn_norm.weight
    F32
    [1024]
  • blk.17.ffn_up.weight
    F16
    [1024, 5248]
    • blk.18
  • blk.18.attn_norm.weight
    F32
    [1024]
  • blk.18.attn_output.weight
    F16
    [1024, 1024]
  • blk.18.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.18.ffn_down.weight
    F16
    [2624, 1024]
  • blk.18.ffn_norm.weight
    F32
    [1024]
  • blk.18.ffn_up.weight
    F16
    [1024, 5248]
    • blk.19
  • blk.19.attn_norm.weight
    F32
    [1024]
  • blk.19.attn_output.weight
    F16
    [1024, 1024]
  • blk.19.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.19.ffn_down.weight
    F16
    [2624, 1024]
  • blk.19.ffn_norm.weight
    F32
    [1024]
  • blk.19.ffn_up.weight
    F16
    [1024, 5248]
    • blk.20
  • blk.20.attn_norm.weight
    F32
    [1024]
  • blk.20.attn_output.weight
    F16
    [1024, 1024]
  • blk.20.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.20.ffn_down.weight
    F16
    [2624, 1024]
  • blk.20.ffn_norm.weight
    F32
    [1024]
  • blk.20.ffn_up.weight
    F16
    [1024, 5248]
    • blk.21
  • blk.21.attn_norm.weight
    F32
    [1024]
  • blk.21.attn_output.weight
    F16
    [1024, 1024]
  • blk.21.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.21.ffn_down.weight
    F16
    [2624, 1024]
  • blk.21.ffn_norm.weight
    F32
    [1024]
  • blk.21.ffn_up.weight
    F16
    [1024, 5248]
    • blk.22
  • blk.22.attn_norm.weight
    F32
    [1024]
  • blk.22.attn_output.weight
    F16
    [1024, 1024]
  • blk.22.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.22.ffn_down.weight
    F16
    [2624, 1024]
  • blk.22.ffn_norm.weight
    F32
    [1024]
  • blk.22.ffn_up.weight
    F16
    [1024, 5248]
    • blk.23
  • blk.23.attn_norm.weight
    F32
    [1024]
  • blk.23.attn_output.weight
    F16
    [1024, 1024]
  • blk.23.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.23.ffn_down.weight
    F16
    [2624, 1024]
  • blk.23.ffn_norm.weight
    F32
    [1024]
  • blk.23.ffn_up.weight
    F16
    [1024, 5248]
    • blk.24
  • blk.24.attn_norm.weight
    F32
    [1024]
  • blk.24.attn_output.weight
    F16
    [1024, 1024]
  • blk.24.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.24.ffn_down.weight
    F16
    [2624, 1024]
  • blk.24.ffn_norm.weight
    F32
    [1024]
  • blk.24.ffn_up.weight
    F16
    [1024, 5248]
    • blk.25
  • blk.25.attn_norm.weight
    F32
    [1024]
  • blk.25.attn_output.weight
    F16
    [1024, 1024]
  • blk.25.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.25.ffn_down.weight
    F16
    [2624, 1024]
  • blk.25.ffn_norm.weight
    F32
    [1024]
  • blk.25.ffn_up.weight
    F16
    [1024, 5248]
    • blk.26
  • blk.26.attn_norm.weight
    F32
    [1024]
  • blk.26.attn_output.weight
    F16
    [1024, 1024]
  • blk.26.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.26.ffn_down.weight
    F16
    [2624, 1024]
  • blk.26.ffn_norm.weight
    F32
    [1024]
  • blk.26.ffn_up.weight
    F16
    [1024, 5248]
    • blk.27
  • blk.27.attn_norm.weight
    F32
    [1024]
  • blk.27.attn_output.weight
    F16
    [1024, 1024]
  • blk.27.attn_qkv.weight
    F16
    [1024, 3072]
  • blk.27.ffn_down.weight
    F16
    [2624, 1024]
  • blk.27.ffn_norm.weight
    F32
    [1024]
  • blk.27.ffn_up.weight
    F16
    [1024, 5248]
  • token_embd_norm.weight
    F32
    [1024]
  • output_norm.weight
    F32
    [1024]