evilfreelancer / enbeddrus

Enbeddrus - English and Russian embedder

embedding

621 5 months ago

8 Tags
... /
model
3247dd8d4593 · 337MB
    Metadata
  • bert.attention.causal
    false
  • bert.attention.head_count
    12
  • bert.attention.layer_norm_epsilon
    1e-12
  • bert.block_count
    12
  • bert.context_length
    512
  • bert.embedding_length
    768
  • bert.feed_forward_length
    3072
  • bert.pooling_type
    1
  • general.architecture
    bert
  • general.file_type
    1
  • general.name
    enbeddrus-v0.1-domain
  • general.quantization_version
    2
  • tokenizer.ggml.cls_token_id
    101
  • tokenizer.ggml.mask_token_id
    103
  • tokenizer.ggml.model
    bert
  • tokenizer.ggml.padding_token_id
    0
  • tokenizer.ggml.pre
    bert-bmu
  • tokenizer.ggml.seperator_token_id
    102
  • tokenizer.ggml.token_type
    [3, 1, 1, 1, 1, ...]
  • tokenizer.ggml.token_type_count
    2
  • tokenizer.ggml.tokens
    [[PAD], [unused1], [unused2], [unused3], [unused4], ...]
  • tokenizer.ggml.unknown_token_id
    100
    • Tensor
  • Name
    Type
    Shape
  • token_embd.weight
    F16
    [768, 105879]
    • blk.0
  • blk.0.attn_k.bias
    F32
    [768]
  • blk.0.attn_k.weight
    F16
    [768, 768]
  • blk.0.attn_output.bias
    F32
    [768]
  • blk.0.attn_output.weight
    F16
    [768, 768]
  • blk.0.attn_output_norm.bias
    F32
    [768]
  • blk.0.attn_output_norm.weight
    F32
    [768]
  • blk.0.attn_q.bias
    F32
    [768]
  • blk.0.attn_q.weight
    F16
    [768, 768]
  • blk.0.attn_v.bias
    F32
    [768]
  • blk.0.attn_v.weight
    F16
    [768, 768]
  • blk.0.ffn_down.bias
    F32
    [768]
  • blk.0.ffn_down.weight
    F16
    [3072, 768]
  • blk.0.ffn_up.bias
    F32
    [3072]
  • blk.0.ffn_up.weight
    F16
    [768, 3072]
  • blk.0.layer_output_norm.bias
    F32
    [768]
  • blk.0.layer_output_norm.weight
    F32
    [768]
    • blk.1
  • blk.1.attn_k.bias
    F32
    [768]
  • blk.1.attn_k.weight
    F16
    [768, 768]
  • blk.1.attn_output.bias
    F32
    [768]
  • blk.1.attn_output.weight
    F16
    [768, 768]
  • blk.1.attn_output_norm.bias
    F32
    [768]
  • blk.1.attn_output_norm.weight
    F32
    [768]
  • blk.1.attn_q.bias
    F32
    [768]
  • blk.1.attn_q.weight
    F16
    [768, 768]
  • blk.1.attn_v.bias
    F32
    [768]
  • blk.1.attn_v.weight
    F16
    [768, 768]
  • blk.1.ffn_down.bias
    F32
    [768]
  • blk.1.ffn_down.weight
    F16
    [3072, 768]
  • blk.1.ffn_up.bias
    F32
    [3072]
  • blk.1.ffn_up.weight
    F16
    [768, 3072]
  • blk.1.layer_output_norm.bias
    F32
    [768]
  • blk.1.layer_output_norm.weight
    F32
    [768]
    • blk.2
  • blk.2.attn_k.bias
    F32
    [768]
  • blk.2.attn_k.weight
    F16
    [768, 768]
  • blk.2.attn_output.bias
    F32
    [768]
  • blk.2.attn_output.weight
    F16
    [768, 768]
  • blk.2.attn_output_norm.bias
    F32
    [768]
  • blk.2.attn_output_norm.weight
    F32
    [768]
  • blk.2.attn_q.bias
    F32
    [768]
  • blk.2.attn_q.weight
    F16
    [768, 768]
  • blk.2.attn_v.bias
    F32
    [768]
  • blk.2.attn_v.weight
    F16
    [768, 768]
  • blk.2.ffn_down.bias
    F32
    [768]
  • blk.2.ffn_down.weight
    F16
    [3072, 768]
  • blk.2.ffn_up.bias
    F32
    [3072]
  • blk.2.ffn_up.weight
    F16
    [768, 3072]
  • blk.2.layer_output_norm.bias
    F32
    [768]
  • blk.2.layer_output_norm.weight
    F32
    [768]
    • blk.3
  • blk.3.attn_k.bias
    F32
    [768]
  • blk.3.attn_k.weight
    F16
    [768, 768]
  • blk.3.attn_output.bias
    F32
    [768]
  • blk.3.attn_output.weight
    F16
    [768, 768]
  • blk.3.attn_output_norm.bias
    F32
    [768]
  • blk.3.attn_output_norm.weight
    F32
    [768]
  • blk.3.attn_q.bias
    F32
    [768]
  • blk.3.attn_q.weight
    F16
    [768, 768]
  • blk.3.attn_v.bias
    F32
    [768]
  • blk.3.attn_v.weight
    F16
    [768, 768]
  • blk.3.ffn_down.bias
    F32
    [768]
  • blk.3.ffn_down.weight
    F16
    [3072, 768]
  • blk.3.ffn_up.bias
    F32
    [3072]
  • blk.3.ffn_up.weight
    F16
    [768, 3072]
  • blk.3.layer_output_norm.bias
    F32
    [768]
  • blk.3.layer_output_norm.weight
    F32
    [768]
    • blk.4
  • blk.4.attn_k.bias
    F32
    [768]
  • blk.4.attn_k.weight
    F16
    [768, 768]
  • blk.4.attn_output.bias
    F32
    [768]
  • blk.4.attn_output.weight
    F16
    [768, 768]
  • blk.4.attn_output_norm.bias
    F32
    [768]
  • blk.4.attn_output_norm.weight
    F32
    [768]
  • blk.4.attn_q.bias
    F32
    [768]
  • blk.4.attn_q.weight
    F16
    [768, 768]
  • blk.4.attn_v.bias
    F32
    [768]
  • blk.4.attn_v.weight
    F16
    [768, 768]
  • blk.4.ffn_down.bias
    F32
    [768]
  • blk.4.ffn_down.weight
    F16
    [3072, 768]
  • blk.4.ffn_up.bias
    F32
    [3072]
  • blk.4.ffn_up.weight
    F16
    [768, 3072]
  • blk.4.layer_output_norm.bias
    F32
    [768]
  • blk.4.layer_output_norm.weight
    F32
    [768]
    • blk.5
  • blk.5.attn_k.bias
    F32
    [768]
  • blk.5.attn_k.weight
    F16
    [768, 768]
  • blk.5.attn_output.bias
    F32
    [768]
  • blk.5.attn_output.weight
    F16
    [768, 768]
  • blk.5.attn_output_norm.bias
    F32
    [768]
  • blk.5.attn_output_norm.weight
    F32
    [768]
  • blk.5.attn_q.bias
    F32
    [768]
  • blk.5.attn_q.weight
    F16
    [768, 768]
  • blk.5.attn_v.bias
    F32
    [768]
  • blk.5.attn_v.weight
    F16
    [768, 768]
  • blk.5.ffn_down.bias
    F32
    [768]
  • blk.5.ffn_down.weight
    F16
    [3072, 768]
  • blk.5.ffn_up.bias
    F32
    [3072]
  • blk.5.ffn_up.weight
    F16
    [768, 3072]
  • blk.5.layer_output_norm.bias
    F32
    [768]
  • blk.5.layer_output_norm.weight
    F32
    [768]
    • blk.6
  • blk.6.attn_k.bias
    F32
    [768]
  • blk.6.attn_k.weight
    F16
    [768, 768]
  • blk.6.attn_output.bias
    F32
    [768]
  • blk.6.attn_output.weight
    F16
    [768, 768]
  • blk.6.attn_output_norm.bias
    F32
    [768]
  • blk.6.attn_output_norm.weight
    F32
    [768]
  • blk.6.attn_q.bias
    F32
    [768]
  • blk.6.attn_q.weight
    F16
    [768, 768]
  • blk.6.attn_v.bias
    F32
    [768]
  • blk.6.attn_v.weight
    F16
    [768, 768]
  • blk.6.ffn_down.bias
    F32
    [768]
  • blk.6.ffn_down.weight
    F16
    [3072, 768]
  • blk.6.ffn_up.bias
    F32
    [3072]
  • blk.6.ffn_up.weight
    F16
    [768, 3072]
  • blk.6.layer_output_norm.bias
    F32
    [768]
  • blk.6.layer_output_norm.weight
    F32
    [768]
    • blk.7
  • blk.7.attn_k.bias
    F32
    [768]
  • blk.7.attn_k.weight
    F16
    [768, 768]
  • blk.7.attn_output.bias
    F32
    [768]
  • blk.7.attn_output.weight
    F16
    [768, 768]
  • blk.7.attn_output_norm.bias
    F32
    [768]
  • blk.7.attn_output_norm.weight
    F32
    [768]
  • blk.7.attn_q.bias
    F32
    [768]
  • blk.7.attn_q.weight
    F16
    [768, 768]
  • blk.7.attn_v.bias
    F32
    [768]
  • blk.7.attn_v.weight
    F16
    [768, 768]
  • blk.7.ffn_down.bias
    F32
    [768]
  • blk.7.ffn_down.weight
    F16
    [3072, 768]
  • blk.7.ffn_up.bias
    F32
    [3072]
  • blk.7.ffn_up.weight
    F16
    [768, 3072]
  • blk.7.layer_output_norm.bias
    F32
    [768]
  • blk.7.layer_output_norm.weight
    F32
    [768]
    • blk.8
  • blk.8.attn_k.bias
    F32
    [768]
  • blk.8.attn_k.weight
    F16
    [768, 768]
  • blk.8.attn_output.bias
    F32
    [768]
  • blk.8.attn_output.weight
    F16
    [768, 768]
  • blk.8.attn_output_norm.bias
    F32
    [768]
  • blk.8.attn_output_norm.weight
    F32
    [768]
  • blk.8.attn_q.bias
    F32
    [768]
  • blk.8.attn_q.weight
    F16
    [768, 768]
  • blk.8.attn_v.bias
    F32
    [768]
  • blk.8.attn_v.weight
    F16
    [768, 768]
  • blk.8.ffn_down.bias
    F32
    [768]
  • blk.8.ffn_down.weight
    F16
    [3072, 768]
  • blk.8.ffn_up.bias
    F32
    [3072]
  • blk.8.ffn_up.weight
    F16
    [768, 3072]
  • blk.8.layer_output_norm.bias
    F32
    [768]
  • blk.8.layer_output_norm.weight
    F32
    [768]
    • blk.9
  • blk.9.attn_k.bias
    F32
    [768]
  • blk.9.attn_k.weight
    F16
    [768, 768]
  • blk.9.attn_output.bias
    F32
    [768]
  • blk.9.attn_output.weight
    F16
    [768, 768]
  • blk.9.attn_output_norm.bias
    F32
    [768]
  • blk.9.attn_output_norm.weight
    F32
    [768]
  • blk.9.attn_q.bias
    F32
    [768]
  • blk.9.attn_q.weight
    F16
    [768, 768]
  • blk.9.attn_v.bias
    F32
    [768]
  • blk.9.attn_v.weight
    F16
    [768, 768]
  • blk.9.ffn_down.bias
    F32
    [768]
  • blk.9.ffn_down.weight
    F16
    [3072, 768]
  • blk.9.ffn_up.bias
    F32
    [3072]
  • blk.9.ffn_up.weight
    F16
    [768, 3072]
  • blk.9.layer_output_norm.bias
    F32
    [768]
  • blk.9.layer_output_norm.weight
    F32
    [768]
    • blk.10
  • blk.10.attn_k.bias
    F32
    [768]
  • blk.10.attn_k.weight
    F16
    [768, 768]
  • blk.10.attn_output.bias
    F32
    [768]
  • blk.10.attn_output.weight
    F16
    [768, 768]
  • blk.10.attn_output_norm.bias
    F32
    [768]
  • blk.10.attn_output_norm.weight
    F32
    [768]
  • blk.10.attn_q.bias
    F32
    [768]
  • blk.10.attn_q.weight
    F16
    [768, 768]
  • blk.10.attn_v.bias
    F32
    [768]
  • blk.10.attn_v.weight
    F16
    [768, 768]
  • blk.10.ffn_down.bias
    F32
    [768]
  • blk.10.ffn_down.weight
    F16
    [3072, 768]
  • blk.10.ffn_up.bias
    F32
    [3072]
  • blk.10.ffn_up.weight
    F16
    [768, 3072]
  • blk.10.layer_output_norm.bias
    F32
    [768]
  • blk.10.layer_output_norm.weight
    F32
    [768]
    • blk.11
  • blk.11.attn_k.bias
    F32
    [768]
  • blk.11.attn_k.weight
    F16
    [768, 768]
  • blk.11.attn_output.bias
    F32
    [768]
  • blk.11.attn_output.weight
    F16
    [768, 768]
  • blk.11.attn_output_norm.bias
    F32
    [768]
  • blk.11.attn_output_norm.weight
    F32
    [768]
  • blk.11.attn_q.bias
    F32
    [768]
  • blk.11.attn_q.weight
    F16
    [768, 768]
  • blk.11.attn_v.bias
    F32
    [768]
  • blk.11.attn_v.weight
    F16
    [768, 768]
  • blk.11.ffn_down.bias
    F32
    [768]
  • blk.11.ffn_down.weight
    F16
    [3072, 768]
  • blk.11.ffn_up.bias
    F32
    [3072]
  • blk.11.ffn_up.weight
    F16
    [768, 3072]
  • blk.11.layer_output_norm.bias
    F32
    [768]
  • blk.11.layer_output_norm.weight
    F32
    [768]
  • position_embd.weight
    F32
    [768, 512]
  • token_embd_norm.bias
    F32
    [768]
  • token_embd_norm.weight
    F32
    [768]
  • token_types.weight
    F32
    [768, 2]