TCYZ/ eng:3m

14 1 month ago

Our first English model.

1m 1.8m 3m
... /
model
3bf95e49f1e8 · 13MB
    Metadata
  • general.architecture
    llama
  • llama.attention.head_count
    4
  • llama.attention.head_count_kv
    4
  • llama.attention.layer_norm_rms_epsilon
    1e-06
  • llama.block_count
    4
  • llama.context_length
    256
  • llama.embedding_length
    256
  • llama.feed_forward_length
    512
  • llama.rope.dimension_count
    64
  • tokenizer.ggml.bos_token_id
    0
  • tokenizer.ggml.eos_token_id
    2
  • tokenizer.ggml.merges
    [Ġ a, Ġ i, e r, o n, Ġi s, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    1
  • tokenizer.ggml.tokens
    [<s>, <pad>, </s>, <unk>, <mask>, ...]
    • Tensor
  • token_embd.weight
    F32
    [256, 1001]
    • blk.0
  • blk.0.attn_k.weight
    F32
    [256, 256]
  • blk.0.attn_norm.weight
    F32
    [256]
  • blk.0.attn_output.weight
    F32
    [256, 256]
  • blk.0.attn_q.weight
    F32
    [256, 256]
  • blk.0.attn_v.weight
    F32
    [256, 256]
  • blk.0.ffn_down.weight
    F32
    [512, 256]
  • blk.0.ffn_gate.weight
    F32
    [256, 512]
  • blk.0.ffn_norm.weight
    F32
    [256]
  • blk.0.ffn_up.weight
    F32
    [256, 512]
    • blk.1
  • blk.1.attn_k.weight
    F32
    [256, 256]
  • blk.1.attn_norm.weight
    F32
    [256]
  • blk.1.attn_output.weight
    F32
    [256, 256]
  • blk.1.attn_q.weight
    F32
    [256, 256]
  • blk.1.attn_v.weight
    F32
    [256, 256]
  • blk.1.ffn_down.weight
    F32
    [512, 256]
  • blk.1.ffn_gate.weight
    F32
    [256, 512]
  • blk.1.ffn_norm.weight
    F32
    [256]
  • blk.1.ffn_up.weight
    F32
    [256, 512]
    • blk.2
  • blk.2.attn_k.weight
    F32
    [256, 256]
  • blk.2.attn_norm.weight
    F32
    [256]
  • blk.2.attn_output.weight
    F32
    [256, 256]
  • blk.2.attn_q.weight
    F32
    [256, 256]
  • blk.2.attn_v.weight
    F32
    [256, 256]
  • blk.2.ffn_down.weight
    F32
    [512, 256]
  • blk.2.ffn_gate.weight
    F32
    [256, 512]
  • blk.2.ffn_norm.weight
    F32
    [256]
  • blk.2.ffn_up.weight
    F32
    [256, 512]
    • blk.3
  • blk.3.attn_k.weight
    F32
    [256, 256]
  • blk.3.attn_norm.weight
    F32
    [256]
  • blk.3.attn_output.weight
    F32
    [256, 256]
  • blk.3.attn_q.weight
    F32
    [256, 256]
  • blk.3.attn_v.weight
    F32
    [256, 256]
  • blk.3.ffn_down.weight
    F32
    [512, 256]
  • blk.3.ffn_gate.weight
    F32
    [256, 512]
  • blk.3.ffn_norm.weight
    F32
    [256]
  • blk.3.ffn_up.weight
    F32
    [256, 512]
  • output.weight
    F32
    [256, 1001]
  • output_norm.weight
    F32
    [256]