sam860/
olmo2:1b-Q6_K

14 1 month ago

Allen Institute's Latest tiny LLM in the Olmo family

1b
7041952b2296 · 1.2GB
    Metadata
  • general.architecture
    olmo2
  • general.file_type
    Q6_K
  • olmo2.attention.head_count
    16
  • olmo2.attention.head_count_kv
    16
  • olmo2.attention.layer_norm_rms_epsilon
    1e-06
  • olmo2.block_count
    16
  • olmo2.context_length
    4096
  • olmo2.embedding_length
    2048
  • olmo2.feed_forward_length
    8192
  • olmo2.rope.freq_base
    500000
  • tokenizer.ggml.bos_token_id
    100257
  • tokenizer.ggml.eos_token_id
    100257
  • tokenizer.ggml.merges
    [Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    100277
  • tokenizer.ggml.pre
    dbrx
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
  • tokenizer.ggml.unknown_token_id
    100257
  • Tensor
  • token_embd.weight
    Q6_K
    [2048, 100352]
  • blk.0
  • blk.0.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.0.attn_k_norm.weight
    F32
    [2048]
  • blk.0.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.0.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.0.attn_q_norm.weight
    F32
    [2048]
  • blk.0.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.0.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.0.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.0.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.0.post_attention_norm.weight
    F32
    [2048]
  • blk.0.post_ffw_norm.weight
    F32
    [2048]
  • blk.1
  • blk.1.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.1.attn_k_norm.weight
    F32
    [2048]
  • blk.1.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.1.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.1.attn_q_norm.weight
    F32
    [2048]
  • blk.1.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.1.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.1.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.1.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.1.post_attention_norm.weight
    F32
    [2048]
  • blk.1.post_ffw_norm.weight
    F32
    [2048]
  • blk.2
  • blk.2.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.2.attn_k_norm.weight
    F32
    [2048]
  • blk.2.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.2.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.2.attn_q_norm.weight
    F32
    [2048]
  • blk.2.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.2.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.2.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.2.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.2.post_attention_norm.weight
    F32
    [2048]
  • blk.2.post_ffw_norm.weight
    F32
    [2048]
  • blk.3
  • blk.3.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.3.attn_k_norm.weight
    F32
    [2048]
  • blk.3.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.3.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.3.attn_q_norm.weight
    F32
    [2048]
  • blk.3.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.3.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.3.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.3.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.3.post_attention_norm.weight
    F32
    [2048]
  • blk.3.post_ffw_norm.weight
    F32
    [2048]
  • blk.4
  • blk.4.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.4.attn_k_norm.weight
    F32
    [2048]
  • blk.4.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.4.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.4.attn_q_norm.weight
    F32
    [2048]
  • blk.4.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.4.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.4.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.4.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.4.post_attention_norm.weight
    F32
    [2048]
  • blk.4.post_ffw_norm.weight
    F32
    [2048]
  • blk.5
  • blk.5.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.5.attn_k_norm.weight
    F32
    [2048]
  • blk.5.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.5.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.5.attn_q_norm.weight
    F32
    [2048]
  • blk.5.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.5.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.5.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.5.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.5.post_attention_norm.weight
    F32
    [2048]
  • blk.5.post_ffw_norm.weight
    F32
    [2048]
  • blk.6
  • blk.6.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.6.attn_k_norm.weight
    F32
    [2048]
  • blk.6.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.6.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.6.attn_q_norm.weight
    F32
    [2048]
  • blk.6.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.6.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.6.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.6.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.6.post_attention_norm.weight
    F32
    [2048]
  • blk.6.post_ffw_norm.weight
    F32
    [2048]
  • blk.7
  • blk.7.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.7.attn_k_norm.weight
    F32
    [2048]
  • blk.7.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.7.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.7.attn_q_norm.weight
    F32
    [2048]
  • blk.7.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.7.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.7.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.7.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.7.post_attention_norm.weight
    F32
    [2048]
  • blk.7.post_ffw_norm.weight
    F32
    [2048]
  • blk.8
  • blk.8.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.8.attn_k_norm.weight
    F32
    [2048]
  • blk.8.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.8.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.8.attn_q_norm.weight
    F32
    [2048]
  • blk.8.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.8.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.8.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.8.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.8.post_attention_norm.weight
    F32
    [2048]
  • blk.8.post_ffw_norm.weight
    F32
    [2048]
  • blk.9
  • blk.9.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.9.attn_k_norm.weight
    F32
    [2048]
  • blk.9.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.9.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.9.attn_q_norm.weight
    F32
    [2048]
  • blk.9.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.9.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.9.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.9.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.9.post_attention_norm.weight
    F32
    [2048]
  • blk.9.post_ffw_norm.weight
    F32
    [2048]
  • blk.10
  • blk.10.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.10.attn_k_norm.weight
    F32
    [2048]
  • blk.10.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.10.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.10.attn_q_norm.weight
    F32
    [2048]
  • blk.10.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.10.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.10.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.10.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.10.post_attention_norm.weight
    F32
    [2048]
  • blk.10.post_ffw_norm.weight
    F32
    [2048]
  • blk.11
  • blk.11.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.11.attn_k_norm.weight
    F32
    [2048]
  • blk.11.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.11.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.11.attn_q_norm.weight
    F32
    [2048]
  • blk.11.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.11.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.11.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.11.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.11.post_attention_norm.weight
    F32
    [2048]
  • blk.11.post_ffw_norm.weight
    F32
    [2048]
  • blk.12
  • blk.12.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.12.attn_k_norm.weight
    F32
    [2048]
  • blk.12.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.12.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.12.attn_q_norm.weight
    F32
    [2048]
  • blk.12.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.12.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.12.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.12.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.12.post_attention_norm.weight
    F32
    [2048]
  • blk.12.post_ffw_norm.weight
    F32
    [2048]
  • blk.13
  • blk.13.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.13.attn_k_norm.weight
    F32
    [2048]
  • blk.13.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.13.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.13.attn_q_norm.weight
    F32
    [2048]
  • blk.13.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.13.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.13.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.13.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.13.post_attention_norm.weight
    F32
    [2048]
  • blk.13.post_ffw_norm.weight
    F32
    [2048]
  • blk.14
  • blk.14.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.14.attn_k_norm.weight
    F32
    [2048]
  • blk.14.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.14.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.14.attn_q_norm.weight
    F32
    [2048]
  • blk.14.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.14.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.14.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.14.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.14.post_attention_norm.weight
    F32
    [2048]
  • blk.14.post_ffw_norm.weight
    F32
    [2048]
  • blk.15
  • blk.15.attn_k.weight
    Q6_K
    [2048, 2048]
  • blk.15.attn_k_norm.weight
    F32
    [2048]
  • blk.15.attn_output.weight
    Q6_K
    [2048, 2048]
  • blk.15.attn_q.weight
    Q6_K
    [2048, 2048]
  • blk.15.attn_q_norm.weight
    F32
    [2048]
  • blk.15.attn_v.weight
    Q6_K
    [2048, 2048]
  • blk.15.ffn_down.weight
    Q6_K
    [8192, 2048]
  • blk.15.ffn_gate.weight
    Q6_K
    [2048, 8192]
  • blk.15.ffn_up.weight
    Q6_K
    [2048, 8192]
  • blk.15.post_attention_norm.weight
    F32
    [2048]
  • blk.15.post_ffw_norm.weight
    F32
    [2048]
  • output.weight
    Q6_K
    [2048, 100352]
  • output_norm.weight
    F32
    [2048]