shuiyuPyLab/ xiaoli:latest

6 2 months ago

... /
model
2347c8fc16fd ยท 49MB
    Metadata
  • general.architecture
    llama
  • general.file_type
    F16
  • llama.attention.head_count
    8
  • llama.attention.head_count_kv
    8
  • llama.attention.layer_norm_rms_epsilon
    1e-06
  • llama.block_count
    6
  • llama.context_length
    96
  • llama.embedding_length
    320
  • llama.feed_forward_length
    1536
  • llama.rope.dimension_count
    40
  • llama.rope.freq_base
    10000
  • llama.vocab_size
    20007
  • tokenizer.ggml.add_bos_token
    false
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.bos_token_id
    2
  • tokenizer.ggml.eos_token_id
    5
  • tokenizer.ggml.model
    rwkv
  • tokenizer.ggml.padding_token_id
    0
  • tokenizer.ggml.token_type
    [3, 3, 3, 3, 3, ...]
  • tokenizer.ggml.tokens
    [<PAD>, <UNK>, <SOS>, <EOS>, <START>, ...]
  • tokenizer.ggml.unknown_token_id
    1
    • Tensor
  • token_embd.weight
    F16
    [320, 20007]
    • blk.0
  • blk.0.attn_k.weight
    F16
    [320, 320]
  • blk.0.attn_norm.weight
    F32
    [320]
  • blk.0.attn_output.weight
    F16
    [320, 320]
  • blk.0.attn_q.weight
    F16
    [320, 320]
  • blk.0.attn_v.weight
    F16
    [320, 320]
  • blk.0.ffn_down.weight
    F16
    [1536, 320]
  • blk.0.ffn_gate.weight
    F16
    [320, 1536]
  • blk.0.ffn_norm.weight
    F32
    [320]
  • blk.0.ffn_up.weight
    F16
    [320, 1536]
    • blk.1
  • blk.1.attn_k.weight
    F16
    [320, 320]
  • blk.1.attn_norm.weight
    F32
    [320]
  • blk.1.attn_output.weight
    F16
    [320, 320]
  • blk.1.attn_q.weight
    F16
    [320, 320]
  • blk.1.attn_v.weight
    F16
    [320, 320]
  • blk.1.ffn_down.weight
    F16
    [1536, 320]
  • blk.1.ffn_gate.weight
    F16
    [320, 1536]
  • blk.1.ffn_norm.weight
    F32
    [320]
  • blk.1.ffn_up.weight
    F16
    [320, 1536]
    • blk.2
  • blk.2.attn_k.weight
    F16
    [320, 320]
  • blk.2.attn_norm.weight
    F32
    [320]
  • blk.2.attn_output.weight
    F16
    [320, 320]
  • blk.2.attn_q.weight
    F16
    [320, 320]
  • blk.2.attn_v.weight
    F16
    [320, 320]
  • blk.2.ffn_down.weight
    F16
    [1536, 320]
  • blk.2.ffn_gate.weight
    F16
    [320, 1536]
  • blk.2.ffn_norm.weight
    F32
    [320]
  • blk.2.ffn_up.weight
    F16
    [320, 1536]
    • blk.3
  • blk.3.attn_k.weight
    F16
    [320, 320]
  • blk.3.attn_norm.weight
    F32
    [320]
  • blk.3.attn_output.weight
    F16
    [320, 320]
  • blk.3.attn_q.weight
    F16
    [320, 320]
  • blk.3.attn_v.weight
    F16
    [320, 320]
  • blk.3.ffn_down.weight
    F16
    [1536, 320]
  • blk.3.ffn_gate.weight
    F16
    [320, 1536]
  • blk.3.ffn_norm.weight
    F32
    [320]
  • blk.3.ffn_up.weight
    F16
    [320, 1536]
    • blk.4
  • blk.4.attn_k.weight
    F16
    [320, 320]
  • blk.4.attn_norm.weight
    F32
    [320]
  • blk.4.attn_output.weight
    F16
    [320, 320]
  • blk.4.attn_q.weight
    F16
    [320, 320]
  • blk.4.attn_v.weight
    F16
    [320, 320]
  • blk.4.ffn_down.weight
    F16
    [1536, 320]
  • blk.4.ffn_gate.weight
    F16
    [320, 1536]
  • blk.4.ffn_norm.weight
    F32
    [320]
  • blk.4.ffn_up.weight
    F16
    [320, 1536]
    • blk.5
  • blk.5.attn_k.weight
    F16
    [320, 320]
  • blk.5.attn_norm.weight
    F32
    [320]
  • blk.5.attn_output.weight
    F16
    [320, 320]
  • blk.5.attn_q.weight
    F16
    [320, 320]
  • blk.5.attn_v.weight
    F16
    [320, 320]
  • blk.5.ffn_down.weight
    F16
    [1536, 320]
  • blk.5.ffn_gate.weight
    F16
    [320, 1536]
  • blk.5.ffn_norm.weight
    F32
    [320]
  • blk.5.ffn_up.weight
    F16
    [320, 1536]
  • output.weight
    F16
    [320, 20007]
  • output_norm.weight
    F32
    [320]