nub235/ voyage-4-nano:latest

50 1 week ago

Voyage AI’s state of the art embedding model in Q8 GGUF for easy use. Important: See README for more info. I did not make this GGUF, credit goes to jsonMartin for the quant. See https://hf.co/jsonMartin/voyage-4-nano-gguf

embedding tools
... /
model
d0e430fe24fa · 372MB
    Metadata
  • general.architecture
    qwen3
  • general.file_type
    Q8_0
  • qwen3.attention.head_count
    16
  • qwen3.attention.head_count_kv
    8
  • qwen3.attention.key_length
    128
  • qwen3.attention.layer_norm_rms_epsilon
    1e-06
  • qwen3.attention.value_length
    128
  • qwen3.block_count
    12
  • qwen3.context_length
    40960
  • qwen3.embedding_length
    1024
  • qwen3.feed_forward_length
    3072
  • qwen3.pooling_type
    Mean
  • qwen3.rope.freq_base
    1e+06
  • tokenizer.ggml.add_bos_token
    false
  • tokenizer.ggml.bos_token_id
    151643
  • tokenizer.ggml.eos_token_id
    151643
  • tokenizer.ggml.merges
    [Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    151643
  • tokenizer.ggml.pre
    qwen2
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
    • Tensor
  • token_embd.weight
    Q8_0
    [1024, 151936]
    • blk.0
  • blk.0.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.0.attn_k_norm.weight
    F32
    [128]
  • blk.0.attn_norm.weight
    F32
    [1024]
  • blk.0.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.0.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.0.attn_q_norm.weight
    F32
    [128]
  • blk.0.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.0.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.0.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.0.ffn_norm.weight
    F32
    [1024]
  • blk.0.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.1
  • blk.1.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.1.attn_k_norm.weight
    F32
    [128]
  • blk.1.attn_norm.weight
    F32
    [1024]
  • blk.1.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.1.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.1.attn_q_norm.weight
    F32
    [128]
  • blk.1.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.1.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.1.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.1.ffn_norm.weight
    F32
    [1024]
  • blk.1.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.2
  • blk.2.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.2.attn_k_norm.weight
    F32
    [128]
  • blk.2.attn_norm.weight
    F32
    [1024]
  • blk.2.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.2.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.2.attn_q_norm.weight
    F32
    [128]
  • blk.2.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.2.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.2.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.2.ffn_norm.weight
    F32
    [1024]
  • blk.2.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.3
  • blk.3.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.3.attn_k_norm.weight
    F32
    [128]
  • blk.3.attn_norm.weight
    F32
    [1024]
  • blk.3.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.3.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.3.attn_q_norm.weight
    F32
    [128]
  • blk.3.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.3.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.3.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.3.ffn_norm.weight
    F32
    [1024]
  • blk.3.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.4
  • blk.4.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.4.attn_k_norm.weight
    F32
    [128]
  • blk.4.attn_norm.weight
    F32
    [1024]
  • blk.4.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.4.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.4.attn_q_norm.weight
    F32
    [128]
  • blk.4.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.4.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.4.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.4.ffn_norm.weight
    F32
    [1024]
  • blk.4.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.5
  • blk.5.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.5.attn_k_norm.weight
    F32
    [128]
  • blk.5.attn_norm.weight
    F32
    [1024]
  • blk.5.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.5.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.5.attn_q_norm.weight
    F32
    [128]
  • blk.5.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.5.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.5.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.5.ffn_norm.weight
    F32
    [1024]
  • blk.5.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.6
  • blk.6.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.6.attn_k_norm.weight
    F32
    [128]
  • blk.6.attn_norm.weight
    F32
    [1024]
  • blk.6.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.6.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.6.attn_q_norm.weight
    F32
    [128]
  • blk.6.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.6.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.6.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.6.ffn_norm.weight
    F32
    [1024]
  • blk.6.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.7
  • blk.7.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.7.attn_k_norm.weight
    F32
    [128]
  • blk.7.attn_norm.weight
    F32
    [1024]
  • blk.7.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.7.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.7.attn_q_norm.weight
    F32
    [128]
  • blk.7.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.7.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.7.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.7.ffn_norm.weight
    F32
    [1024]
  • blk.7.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.8
  • blk.8.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.8.attn_k_norm.weight
    F32
    [128]
  • blk.8.attn_norm.weight
    F32
    [1024]
  • blk.8.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.8.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.8.attn_q_norm.weight
    F32
    [128]
  • blk.8.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.8.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.8.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.8.ffn_norm.weight
    F32
    [1024]
  • blk.8.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.9
  • blk.9.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.9.attn_k_norm.weight
    F32
    [128]
  • blk.9.attn_norm.weight
    F32
    [1024]
  • blk.9.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.9.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.9.attn_q_norm.weight
    F32
    [128]
  • blk.9.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.9.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.9.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.9.ffn_norm.weight
    F32
    [1024]
  • blk.9.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.10
  • blk.10.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.10.attn_k_norm.weight
    F32
    [128]
  • blk.10.attn_norm.weight
    F32
    [1024]
  • blk.10.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.10.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.10.attn_q_norm.weight
    F32
    [128]
  • blk.10.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.10.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.10.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.10.ffn_norm.weight
    F32
    [1024]
  • blk.10.ffn_up.weight
    Q8_0
    [1024, 3072]
    • blk.11
  • blk.11.attn_k.weight
    Q8_0
    [1024, 1024]
  • blk.11.attn_k_norm.weight
    F32
    [128]
  • blk.11.attn_norm.weight
    F32
    [1024]
  • blk.11.attn_output.weight
    Q8_0
    [2048, 1024]
  • blk.11.attn_q.weight
    Q8_0
    [1024, 2048]
  • blk.11.attn_q_norm.weight
    F32
    [128]
  • blk.11.attn_v.weight
    Q8_0
    [1024, 1024]
  • blk.11.ffn_down.weight
    Q8_0
    [3072, 1024]
  • blk.11.ffn_gate.weight
    Q8_0
    [1024, 3072]
  • blk.11.ffn_norm.weight
    F32
    [1024]
  • blk.11.ffn_up.weight
    Q8_0
    [1024, 3072]
  • output_norm.weight
    F32
    [1024]