TCYZ/ cokertme3:0.2b

6 1 month ago

3. Çökertme Modeli

0.2b
... /
model
f26b74559eea · 835MB
    Metadata
  • general.architecture
    llama
  • llama.attention.head_count
    8
  • llama.attention.head_count_kv
    8
  • llama.attention.layer_norm_rms_epsilon
    1e-06
  • llama.block_count
    8
  • llama.context_length
    2048
  • llama.embedding_length
    2048
  • llama.feed_forward_length
    1376
  • llama.rope.dimension_count
    256
  • tokenizer.ggml.bos_token_id
    0
  • tokenizer.ggml.eos_token_id
    2
  • tokenizer.ggml.merges
    [Ä ±, < /, Ġ </, a r, Ã ¼, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    1
  • tokenizer.ggml.tokens
    [<s>, <pad>, </s>, <unk>, !, ...]
    • Tensor
  • token_embd.weight
    F32
    [2048, 1653]
    • blk.0
  • blk.0.attn_k.weight
    F32
    [2048, 2048]
  • blk.0.attn_norm.weight
    F32
    [2048]
  • blk.0.attn_output.weight
    F32
    [2048, 2048]
  • blk.0.attn_q.weight
    F32
    [2048, 2048]
  • blk.0.attn_v.weight
    F32
    [2048, 2048]
  • blk.0.ffn_down.weight
    F32
    [1376, 2048]
  • blk.0.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.0.ffn_norm.weight
    F32
    [2048]
  • blk.0.ffn_up.weight
    F32
    [2048, 1376]
    • blk.1
  • blk.1.attn_k.weight
    F32
    [2048, 2048]
  • blk.1.attn_norm.weight
    F32
    [2048]
  • blk.1.attn_output.weight
    F32
    [2048, 2048]
  • blk.1.attn_q.weight
    F32
    [2048, 2048]
  • blk.1.attn_v.weight
    F32
    [2048, 2048]
  • blk.1.ffn_down.weight
    F32
    [1376, 2048]
  • blk.1.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.1.ffn_norm.weight
    F32
    [2048]
  • blk.1.ffn_up.weight
    F32
    [2048, 1376]
    • blk.2
  • blk.2.attn_k.weight
    F32
    [2048, 2048]
  • blk.2.attn_norm.weight
    F32
    [2048]
  • blk.2.attn_output.weight
    F32
    [2048, 2048]
  • blk.2.attn_q.weight
    F32
    [2048, 2048]
  • blk.2.attn_v.weight
    F32
    [2048, 2048]
  • blk.2.ffn_down.weight
    F32
    [1376, 2048]
  • blk.2.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.2.ffn_norm.weight
    F32
    [2048]
  • blk.2.ffn_up.weight
    F32
    [2048, 1376]
    • blk.3
  • blk.3.attn_k.weight
    F32
    [2048, 2048]
  • blk.3.attn_norm.weight
    F32
    [2048]
  • blk.3.attn_output.weight
    F32
    [2048, 2048]
  • blk.3.attn_q.weight
    F32
    [2048, 2048]
  • blk.3.attn_v.weight
    F32
    [2048, 2048]
  • blk.3.ffn_down.weight
    F32
    [1376, 2048]
  • blk.3.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.3.ffn_norm.weight
    F32
    [2048]
  • blk.3.ffn_up.weight
    F32
    [2048, 1376]
    • blk.4
  • blk.4.attn_k.weight
    F32
    [2048, 2048]
  • blk.4.attn_norm.weight
    F32
    [2048]
  • blk.4.attn_output.weight
    F32
    [2048, 2048]
  • blk.4.attn_q.weight
    F32
    [2048, 2048]
  • blk.4.attn_v.weight
    F32
    [2048, 2048]
  • blk.4.ffn_down.weight
    F32
    [1376, 2048]
  • blk.4.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.4.ffn_norm.weight
    F32
    [2048]
  • blk.4.ffn_up.weight
    F32
    [2048, 1376]
    • blk.5
  • blk.5.attn_k.weight
    F32
    [2048, 2048]
  • blk.5.attn_norm.weight
    F32
    [2048]
  • blk.5.attn_output.weight
    F32
    [2048, 2048]
  • blk.5.attn_q.weight
    F32
    [2048, 2048]
  • blk.5.attn_v.weight
    F32
    [2048, 2048]
  • blk.5.ffn_down.weight
    F32
    [1376, 2048]
  • blk.5.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.5.ffn_norm.weight
    F32
    [2048]
  • blk.5.ffn_up.weight
    F32
    [2048, 1376]
    • blk.6
  • blk.6.attn_k.weight
    F32
    [2048, 2048]
  • blk.6.attn_norm.weight
    F32
    [2048]
  • blk.6.attn_output.weight
    F32
    [2048, 2048]
  • blk.6.attn_q.weight
    F32
    [2048, 2048]
  • blk.6.attn_v.weight
    F32
    [2048, 2048]
  • blk.6.ffn_down.weight
    F32
    [1376, 2048]
  • blk.6.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.6.ffn_norm.weight
    F32
    [2048]
  • blk.6.ffn_up.weight
    F32
    [2048, 1376]
    • blk.7
  • blk.7.attn_k.weight
    F32
    [2048, 2048]
  • blk.7.attn_norm.weight
    F32
    [2048]
  • blk.7.attn_output.weight
    F32
    [2048, 2048]
  • blk.7.attn_q.weight
    F32
    [2048, 2048]
  • blk.7.attn_v.weight
    F32
    [2048, 2048]
  • blk.7.ffn_down.weight
    F32
    [1376, 2048]
  • blk.7.ffn_gate.weight
    F32
    [2048, 1376]
  • blk.7.ffn_norm.weight
    F32
    [2048]
  • blk.7.ffn_up.weight
    F32
    [2048, 1376]
  • output.weight
    F32
    [2048, 1653]
  • output_norm.weight
    F32
    [2048]