anthony-maio/ eve-2-moe-nanofunction-272m:q8-0

1 month ago

... /
model
515597f801f7 · 291MB
    Metadata
  • general.architecture
    deepseek
  • general.file_type
    Q8_0
  • deepseek.attention.head_count
    8
  • deepseek.attention.head_count_kv
    8
  • deepseek.attention.layer_norm_rms_epsilon
    1e-05
  • deepseek.block_count
    12
  • deepseek.context_length
    2048
  • deepseek.embedding_length
    512
  • deepseek.expert_count
    8
  • deepseek.expert_feed_forward_length
    1408
  • deepseek.expert_gating_func
    1
  • deepseek.expert_shared_count
    1
  • deepseek.expert_used_count
    2
  • deepseek.expert_weights_scale
    1
  • deepseek.feed_forward_length
    1408
  • deepseek.leading_dense_block_count
    0
  • deepseek.rope.dimension_count
    64
  • deepseek.rope.freq_base
    10000
  • deepseek.rope.scaling.type
    none
  • deepseek.vocab_size
    50304
  • tokenizer.ggml.bos_token_id
    50256
  • tokenizer.ggml.eos_token_id
    50256
  • tokenizer.ggml.merges
    [Ġ t, Ġ a, h e, i n, r e, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    50256
  • tokenizer.ggml.pre
    gpt-2
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
  • tokenizer.ggml.unknown_token_id
    50256
    • Tensor
  • token_embd.weight
    Q8_0
    [512, 50304]
    • blk.0
  • blk.0.attn_k.weight
    Q8_0
    [512, 512]
  • blk.0.attn_norm.weight
    F32
    [512]
  • blk.0.attn_output.weight
    Q8_0
    [512, 512]
  • blk.0.attn_q.weight
    Q8_0
    [512, 512]
  • blk.0.attn_v.weight
    Q8_0
    [512, 512]
  • blk.0.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.0.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.0.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.0.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.0.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.0.ffn_norm.weight
    F32
    [512]
  • blk.0.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.0.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.1
  • blk.1.attn_k.weight
    Q8_0
    [512, 512]
  • blk.1.attn_norm.weight
    F32
    [512]
  • blk.1.attn_output.weight
    Q8_0
    [512, 512]
  • blk.1.attn_q.weight
    Q8_0
    [512, 512]
  • blk.1.attn_v.weight
    Q8_0
    [512, 512]
  • blk.1.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.1.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.1.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.1.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.1.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.1.ffn_norm.weight
    F32
    [512]
  • blk.1.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.1.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.2
  • blk.2.attn_k.weight
    Q8_0
    [512, 512]
  • blk.2.attn_norm.weight
    F32
    [512]
  • blk.2.attn_output.weight
    Q8_0
    [512, 512]
  • blk.2.attn_q.weight
    Q8_0
    [512, 512]
  • blk.2.attn_v.weight
    Q8_0
    [512, 512]
  • blk.2.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.2.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.2.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.2.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.2.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.2.ffn_norm.weight
    F32
    [512]
  • blk.2.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.2.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.3
  • blk.3.attn_k.weight
    Q8_0
    [512, 512]
  • blk.3.attn_norm.weight
    F32
    [512]
  • blk.3.attn_output.weight
    Q8_0
    [512, 512]
  • blk.3.attn_q.weight
    Q8_0
    [512, 512]
  • blk.3.attn_v.weight
    Q8_0
    [512, 512]
  • blk.3.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.3.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.3.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.3.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.3.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.3.ffn_norm.weight
    F32
    [512]
  • blk.3.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.3.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.4
  • blk.4.attn_k.weight
    Q8_0
    [512, 512]
  • blk.4.attn_norm.weight
    F32
    [512]
  • blk.4.attn_output.weight
    Q8_0
    [512, 512]
  • blk.4.attn_q.weight
    Q8_0
    [512, 512]
  • blk.4.attn_v.weight
    Q8_0
    [512, 512]
  • blk.4.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.4.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.4.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.4.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.4.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.4.ffn_norm.weight
    F32
    [512]
  • blk.4.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.4.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.5
  • blk.5.attn_k.weight
    Q8_0
    [512, 512]
  • blk.5.attn_norm.weight
    F32
    [512]
  • blk.5.attn_output.weight
    Q8_0
    [512, 512]
  • blk.5.attn_q.weight
    Q8_0
    [512, 512]
  • blk.5.attn_v.weight
    Q8_0
    [512, 512]
  • blk.5.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.5.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.5.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.5.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.5.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.5.ffn_norm.weight
    F32
    [512]
  • blk.5.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.5.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.6
  • blk.6.attn_k.weight
    Q8_0
    [512, 512]
  • blk.6.attn_norm.weight
    F32
    [512]
  • blk.6.attn_output.weight
    Q8_0
    [512, 512]
  • blk.6.attn_q.weight
    Q8_0
    [512, 512]
  • blk.6.attn_v.weight
    Q8_0
    [512, 512]
  • blk.6.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.6.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.6.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.6.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.6.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.6.ffn_norm.weight
    F32
    [512]
  • blk.6.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.6.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.7
  • blk.7.attn_k.weight
    Q8_0
    [512, 512]
  • blk.7.attn_norm.weight
    F32
    [512]
  • blk.7.attn_output.weight
    Q8_0
    [512, 512]
  • blk.7.attn_q.weight
    Q8_0
    [512, 512]
  • blk.7.attn_v.weight
    Q8_0
    [512, 512]
  • blk.7.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.7.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.7.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.7.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.7.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.7.ffn_norm.weight
    F32
    [512]
  • blk.7.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.7.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.8
  • blk.8.attn_k.weight
    Q8_0
    [512, 512]
  • blk.8.attn_norm.weight
    F32
    [512]
  • blk.8.attn_output.weight
    Q8_0
    [512, 512]
  • blk.8.attn_q.weight
    Q8_0
    [512, 512]
  • blk.8.attn_v.weight
    Q8_0
    [512, 512]
  • blk.8.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.8.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.8.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.8.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.8.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.8.ffn_norm.weight
    F32
    [512]
  • blk.8.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.8.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.9
  • blk.9.attn_k.weight
    Q8_0
    [512, 512]
  • blk.9.attn_norm.weight
    F32
    [512]
  • blk.9.attn_output.weight
    Q8_0
    [512, 512]
  • blk.9.attn_q.weight
    Q8_0
    [512, 512]
  • blk.9.attn_v.weight
    Q8_0
    [512, 512]
  • blk.9.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.9.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.9.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.9.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.9.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.9.ffn_norm.weight
    F32
    [512]
  • blk.9.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.9.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.10
  • blk.10.attn_k.weight
    Q8_0
    [512, 512]
  • blk.10.attn_norm.weight
    F32
    [512]
  • blk.10.attn_output.weight
    Q8_0
    [512, 512]
  • blk.10.attn_q.weight
    Q8_0
    [512, 512]
  • blk.10.attn_v.weight
    Q8_0
    [512, 512]
  • blk.10.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.10.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.10.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.10.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.10.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.10.ffn_norm.weight
    F32
    [512]
  • blk.10.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.10.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
    • blk.11
  • blk.11.attn_k.weight
    Q8_0
    [512, 512]
  • blk.11.attn_norm.weight
    F32
    [512]
  • blk.11.attn_output.weight
    Q8_0
    [512, 512]
  • blk.11.attn_q.weight
    Q8_0
    [512, 512]
  • blk.11.attn_v.weight
    Q8_0
    [512, 512]
  • blk.11.ffn_down_exps.weight
    Q8_0
    [1408, 512, 8]
  • blk.11.ffn_down_shexp.weight
    Q8_0
    [1408, 512]
  • blk.11.ffn_gate_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.11.ffn_gate_inp.weight
    F32
    [512, 8]
  • blk.11.ffn_gate_shexp.weight
    Q8_0
    [512, 1408]
  • blk.11.ffn_norm.weight
    F32
    [512]
  • blk.11.ffn_up_exps.weight
    Q8_0
    [512, 1408, 8]
  • blk.11.ffn_up_shexp.weight
    Q8_0
    [512, 1408]
  • output_norm.weight
    F32
    [512]