sadiknawazyt/ sadik-ai-nano-multilingual:latest

1 yesterday

... /
model
76de69757ac7 · 33MB
    Metadata
  • general.architecture
    llama
  • general.file_type
    F32
  • llama.attention.head_count
    6
  • llama.attention.head_count_kv
    2
  • llama.attention.layer_norm_rms_epsilon
    1e-06
  • llama.block_count
    4
  • llama.context_length
    256
  • llama.embedding_length
    192
  • llama.feed_forward_length
    512
  • llama.rope.dimension_count
    32
  • llama.rope.freq_base
    10000
  • llama.vocab_size
    32768
  • tokenizer.ggml.add_bos_token
    false
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.bos_token_id
    0
  • tokenizer.ggml.eos_token_id
    0
  • tokenizer.ggml.merges
    [à ¤, Ġ Ġ, à ¥, Ġ à¤, ĠĠ ĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    1
  • tokenizer.ggml.pre
    gpt-2
  • tokenizer.ggml.token_type
    [3, 3, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [<|endoftext|>, <|pad|>, !, ", #, ...]
    • Tensor
  • token_embd.weight
    F32
    [192, 32768]
    • blk.0
  • blk.0.attn_k.weight
    F32
    [192, 64]
  • blk.0.attn_norm.weight
    F32
    [192]
  • blk.0.attn_output.weight
    F32
    [192, 192]
  • blk.0.attn_q.weight
    F32
    [192, 192]
  • blk.0.attn_v.weight
    F32
    [192, 64]
  • blk.0.ffn_down.weight
    F32
    [512, 192]
  • blk.0.ffn_gate.weight
    F32
    [192, 512]
  • blk.0.ffn_norm.weight
    F32
    [192]
  • blk.0.ffn_up.weight
    F32
    [192, 512]
    • blk.1
  • blk.1.attn_k.weight
    F32
    [192, 64]
  • blk.1.attn_norm.weight
    F32
    [192]
  • blk.1.attn_output.weight
    F32
    [192, 192]
  • blk.1.attn_q.weight
    F32
    [192, 192]
  • blk.1.attn_v.weight
    F32
    [192, 64]
  • blk.1.ffn_down.weight
    F32
    [512, 192]
  • blk.1.ffn_gate.weight
    F32
    [192, 512]
  • blk.1.ffn_norm.weight
    F32
    [192]
  • blk.1.ffn_up.weight
    F32
    [192, 512]
    • blk.2
  • blk.2.attn_k.weight
    F32
    [192, 64]
  • blk.2.attn_norm.weight
    F32
    [192]
  • blk.2.attn_output.weight
    F32
    [192, 192]
  • blk.2.attn_q.weight
    F32
    [192, 192]
  • blk.2.attn_v.weight
    F32
    [192, 64]
  • blk.2.ffn_down.weight
    F32
    [512, 192]
  • blk.2.ffn_gate.weight
    F32
    [192, 512]
  • blk.2.ffn_norm.weight
    F32
    [192]
  • blk.2.ffn_up.weight
    F32
    [192, 512]
    • blk.3
  • blk.3.attn_k.weight
    F32
    [192, 64]
  • blk.3.attn_norm.weight
    F32
    [192]
  • blk.3.attn_output.weight
    F32
    [192, 192]
  • blk.3.attn_q.weight
    F32
    [192, 192]
  • blk.3.attn_v.weight
    F32
    [192, 64]
  • blk.3.ffn_down.weight
    F32
    [512, 192]
  • blk.3.ffn_gate.weight
    F32
    [192, 512]
  • blk.3.ffn_norm.weight
    F32
    [192]
  • blk.3.ffn_up.weight
    F32
    [192, 512]
  • output_norm.weight
    F32
    [192]