nekincn/ mistra-nemo-full:latest

3 1 week ago

... /
model
fb8d2cf3a120 · 3.0GB
    Metadata
  • general.architecture
    llama
  • general.file_type
    BF16
  • llama.attention.head_count
    32
  • llama.attention.head_count_kv
    8
  • llama.attention.key_length
    128
  • llama.attention.layer_norm_rms_epsilon
    1e-05
  • llama.attention.value_length
    128
  • llama.block_count
    40
  • llama.context_length
    131072
  • llama.embedding_length
    5120
  • llama.feed_forward_length
    14336
  • llama.rope.dimension_count
    128
  • llama.rope.freq_base
    1e+06
  • llama.vocab_size
    131072
  • tokenizer.ggml.add_bos_token
    true
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.add_sep_token
    false
  • tokenizer.ggml.add_space_prefix
    false
  • tokenizer.ggml.bos_token_id
    1
  • tokenizer.ggml.eos_token_id
    2
  • tokenizer.ggml.merges
    [Ġ Ġ, Ġ t, e r, i n, Ġ ĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.pre
    tekken
  • tokenizer.ggml.token_type
    [3, 3, 3, 3, 3, ...]
  • tokenizer.ggml.tokens
    [<unk>, <s>, </s>, [INST], [/INST], ...]
  • tokenizer.ggml.unknown_token_id
    0
  • split.count
    9
  • split.no
    0
  • split.tensors.count
    363
    • Tensor
    blk.0
  • blk.0.attn_k.weight
    BF16
    [5120, 1024]
  • blk.0.attn_norm.weight
    F32
    [5120]
  • blk.0.attn_output.weight
    BF16
    [4096, 5120]
  • blk.0.attn_q.weight
    BF16
    [5120, 4096]
  • blk.0.attn_v.weight
    BF16
    [5120, 1024]
  • blk.0.ffn_down.weight
    BF16
    [14336, 5120]
  • blk.0.ffn_gate.weight
    BF16
    [5120, 14336]
  • blk.0.ffn_norm.weight
    F32
    [5120]
  • blk.0.ffn_up.weight
    BF16
    [5120, 14336]
    • blk.1
  • blk.1.attn_k.weight
    BF16
    [5120, 1024]
  • blk.1.attn_norm.weight
    F32
    [5120]
  • blk.1.attn_output.weight
    BF16
    [4096, 5120]
  • blk.1.attn_q.weight
    BF16
    [5120, 4096]
  • blk.1.attn_v.weight
    BF16
    [5120, 1024]
  • blk.1.ffn_down.weight
    BF16
    [14336, 5120]
  • blk.1.ffn_gate.weight
    BF16
    [5120, 14336]
  • blk.1.ffn_norm.weight
    F32
    [5120]
  • blk.1.ffn_up.weight
    BF16
    [5120, 14336]
    • blk.10
  • blk.10.attn_k.weight
    BF16
    [5120, 1024]
  • blk.10.attn_norm.weight
    F32
    [5120]
  • blk.10.attn_output.weight
    BF16
    [4096, 5120]
  • blk.10.attn_q.weight
    BF16
    [5120, 4096]
  • blk.10.attn_v.weight
    BF16
    [5120, 1024]
  • blk.10.ffn_down.weight
    BF16
    [14336, 5120]
  • blk.10.ffn_gate.weight
    BF16
    [5120, 14336]
  • blk.10.ffn_norm.weight
    F32
    [5120]
  • blk.10.ffn_up.weight
    BF16
    [5120, 14336]
    • blk.11
  • blk.11.attn_k.weight
    BF16
    [5120, 1024]
  • blk.11.attn_norm.weight
    F32
    [5120]
  • blk.11.attn_output.weight
    BF16
    [4096, 5120]
  • blk.11.attn_q.weight
    BF16
    [5120, 4096]
  • blk.11.attn_v.weight
    BF16
    [5120, 1024]
  • blk.11.ffn_down.weight
    BF16
    [14336, 5120]
  • blk.11.ffn_gate.weight
    BF16
    [5120, 14336]
  • blk.11.ffn_norm.weight
    F32
    [5120]
  • blk.11.ffn_up.weight
    BF16
    [5120, 14336]
    • blk.12
  • blk.12.attn_k.weight
    BF16
    [5120, 1024]
  • blk.12.attn_norm.weight
    F32
    [5120]
  • blk.12.attn_output.weight
    BF16
    [4096, 5120]
  • blk.12.attn_q.weight
    BF16
    [5120, 4096]
  • blk.12.attn_v.weight
    BF16
    [5120, 1024]
  • blk.12.ffn_down.weight
    BF16
    [14336, 5120]
  • blk.12.ffn_gate.weight
    BF16
    [5120, 14336]
  • blk.12.ffn_norm.weight
    F32
    [5120]
  • blk.12.ffn_up.weight
    BF16
    [5120, 14336]
    • blk.13
  • blk.13.attn_k.weight
    BF16
    [5120, 1024]
  • blk.13.attn_norm.weight
    F32
    [5120]
  • blk.13.attn_output.weight
    BF16
    [4096, 5120]
  • blk.13.attn_q.weight
    BF16
    [5120, 4096]
  • blk.13.attn_v.weight
    BF16
    [5120, 1024]
  • blk.13.ffn_gate.weight
    BF16
    [5120, 14336]