4 2 hours ago

vision
363dff2bdbd1 · 45GB
    Metadata
  • general.architecture
    mistral-large4
  • general.file_type
    IQ1_S
  • mistral-large4.attention.head_count
    64
  • mistral-large4.attention.head_count_kv
    1
  • mistral-large4.attention.indexer.head_count
    32
  • mistral-large4.attention.indexer.key_length
    128
  • mistral-large4.attention.indexer.top_k
    2048
  • mistral-large4.attention.indexer.types
    [true, true, true, true, true, ...]
  • mistral-large4.attention.key_length
    576
  • mistral-large4.attention.key_length_mla
    128
  • mistral-large4.attention.kv_lora_rank
    512
  • mistral-large4.attention.kv_source_map
    [0, 1, 2, 3, 4, ...]
  • mistral-large4.attention.layer_norm_epsilon
    1e-06
  • mistral-large4.attention.layer_norm_rms_epsilon
    1e-05
  • mistral-large4.attention.rope_pattern
    [true, true, true, false, true, ...]
  • mistral-large4.attention.value_length
    512
  • mistral-large4.attention.value_length_mla
    128
  • mistral-large4.block_count
    72
  • mistral-large4.context_length
    1048576
  • mistral-large4.embedding_length
    6144
  • mistral-large4.expert_count
    128
  • mistral-large4.expert_feed_forward_length
    6144
  • mistral-large4.expert_gating_func
    1
  • mistral-large4.expert_group_count
    1
  • mistral-large4.expert_group_used_count
    1
  • mistral-large4.expert_shared_count
    1
  • mistral-large4.expert_used_count
    4
  • mistral-large4.expert_weights_norm
    true
  • mistral-large4.expert_weights_scale
    1
  • mistral-large4.feed_forward_length
    24576
  • mistral-large4.final_logit_softcapping
    30
  • mistral-large4.leading_dense_block_count
    1
  • mistral-large4.rope.dimension_count
    64
  • mistral-large4.rope.freq_base
    10000
  • mistral-large4.rope.scaling.factor
    2048
  • mistral-large4.rope.scaling.original_context_length
    512
  • mistral-large4.rope.scaling.type
    yarn
  • mistral-large4.rope.scaling.yarn_beta_fast
    32
  • mistral-large4.rope.scaling.yarn_beta_slow
    1
  • mistral-large4.rope.scaling.yarn_log_multiplier
    1
  • mistral-large4.vocab_size
    204800
  • tokenizer.ggml.add_bos_token
    true
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.bos_token_id
    1
  • tokenizer.ggml.eos_token_id
    2
  • tokenizer.ggml.merges
    [Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, e r, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    11
  • tokenizer.ggml.pre
    tekken-num3
  • tokenizer.ggml.scores
    [0, 0, 0, 0, 0, ...]
  • 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
    6
  • split.no
    0
  • split.tensors.count
    1602
  • Tensor
  • token_embd.weight
    Q3_K
    [6144, 204800]
  • blk.0
  • blk.0.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.0.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.0.attn_kv_a_norm.weight
    F32
    [512]
  • blk.0.attn_norm.weight
    F32
    [6144]
  • blk.0.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.0.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.0.attn_ssmax.bias
    F32
    [64]
  • blk.0.attn_ssmax.weight
    F32
    [64]
  • blk.0.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.0.ffn_down.weight
    Q5_K
    [24576, 6144]
  • blk.0.ffn_gate.weight
    Q5_K
    [6144, 24576]
  • blk.0.ffn_norm.weight
    F32
    [6144]
  • blk.0.ffn_up.weight
    Q5_K
    [6144, 24576]
  • blk.0.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.0.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.0.indexer.k_norm.bias
    F32
    [128]
  • blk.0.indexer.k_norm.weight
    F32
    [128]
  • blk.0.indexer.proj.weight
    F32
    [6144, 32]
  • blk.0.indexer.q_norm.bias
    F32
    [6144]
  • blk.0.indexer.q_norm.weight
    F32
    [6144]
  • blk.0.indexer.ssmax.bias
    F32
    [32]
  • blk.0.indexer.ssmax.weight
    F32
    [32]
  • blk.1
  • blk.1.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.1.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.1.attn_kv_a_norm.weight
    F32
    [512]
  • blk.1.attn_norm.weight
    F32
    [6144]
  • blk.1.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.1.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.1.attn_ssmax.bias
    F32
    [64]
  • blk.1.attn_ssmax.weight
    F32
    [64]
  • blk.1.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.1.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.1.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.1.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.1.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.1.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.1.ffn_norm.weight
    F32
    [6144]
  • blk.1.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.1.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.1.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.1.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.1.indexer.k_norm.bias
    F32
    [128]
  • blk.1.indexer.k_norm.weight
    F32
    [128]
  • blk.1.indexer.proj.weight
    F32
    [6144, 32]
  • blk.1.indexer.q_norm.bias
    F32
    [6144]
  • blk.1.indexer.q_norm.weight
    F32
    [6144]
  • blk.1.indexer.ssmax.bias
    F32
    [32]
  • blk.1.indexer.ssmax.weight
    F32
    [32]
  • blk.2
  • blk.2.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.2.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.2.attn_kv_a_norm.weight
    F32
    [512]
  • blk.2.attn_norm.weight
    F32
    [6144]
  • blk.2.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.2.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.2.attn_ssmax.bias
    F32
    [64]
  • blk.2.attn_ssmax.weight
    F32
    [64]
  • blk.2.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.2.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.2.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.2.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.2.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.2.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.2.ffn_norm.weight
    F32
    [6144]
  • blk.2.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.2.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.2.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.2.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.2.indexer.k_norm.bias
    F32
    [128]
  • blk.2.indexer.k_norm.weight
    F32
    [128]
  • blk.2.indexer.proj.weight
    F32
    [6144, 32]
  • blk.2.indexer.q_norm.bias
    F32
    [6144]
  • blk.2.indexer.q_norm.weight
    F32
    [6144]
  • blk.2.indexer.ssmax.bias
    F32
    [32]
  • blk.2.indexer.ssmax.weight
    F32
    [32]
  • blk.3
  • blk.3.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.3.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.3.attn_kv_a_norm.weight
    F32
    [512]
  • blk.3.attn_norm.weight
    F32
    [6144]
  • blk.3.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.3.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.3.attn_ssmax.bias
    F32
    [64]
  • blk.3.attn_ssmax.weight
    F32
    [64]
  • blk.3.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.3.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.3.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.3.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.3.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.3.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.3.ffn_norm.weight
    F32
    [6144]
  • blk.3.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.3.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.3.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.3.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.3.indexer.k_norm.bias
    F32
    [128]
  • blk.3.indexer.k_norm.weight
    F32
    [128]
  • blk.3.indexer.proj.weight
    F32
    [6144, 32]
  • blk.3.indexer.q_norm.bias
    F32
    [6144]
  • blk.3.indexer.q_norm.weight
    F32
    [6144]
  • blk.3.indexer.ssmax.bias
    F32
    [32]
  • blk.3.indexer.ssmax.weight
    F32
    [32]
  • blk.4
  • blk.4.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.4.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.4.attn_kv_a_norm.weight
    F32
    [512]
  • blk.4.attn_norm.weight
    F32
    [6144]
  • blk.4.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.4.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.4.attn_ssmax.bias
    F32
    [64]
  • blk.4.attn_ssmax.weight
    F32
    [64]
  • blk.4.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.4.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.4.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.4.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.4.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.4.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.4.ffn_norm.weight
    F32
    [6144]
  • blk.4.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.4.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.4.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.4.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.4.indexer.k_norm.bias
    F32
    [128]
  • blk.4.indexer.k_norm.weight
    F32
    [128]
  • blk.4.indexer.proj.weight
    F32
    [6144, 32]
  • blk.4.indexer.q_norm.bias
    F32
    [6144]
  • blk.4.indexer.q_norm.weight
    F32
    [6144]
  • blk.4.indexer.ssmax.bias
    F32
    [32]
  • blk.4.indexer.ssmax.weight
    F32
    [32]
  • blk.5
  • blk.5.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.5.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.5.attn_kv_a_norm.weight
    F32
    [512]
  • blk.5.attn_norm.weight
    F32
    [6144]
  • blk.5.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.5.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.5.attn_ssmax.bias
    F32
    [64]
  • blk.5.attn_ssmax.weight
    F32
    [64]
  • blk.5.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.5.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.5.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.5.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.5.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.5.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.5.ffn_norm.weight
    F32
    [6144]
  • blk.5.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.5.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.5.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.5.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.5.indexer.k_norm.bias
    F32
    [128]
  • blk.5.indexer.k_norm.weight
    F32
    [128]
  • blk.5.indexer.proj.weight
    F32
    [6144, 32]
  • blk.5.indexer.q_norm.bias
    F32
    [6144]
  • blk.5.indexer.q_norm.weight
    F32
    [6144]
  • blk.5.indexer.ssmax.bias
    F32
    [32]
  • blk.5.indexer.ssmax.weight
    F32
    [32]
  • blk.6
  • blk.6.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.6.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.6.attn_kv_a_norm.weight
    F32
    [512]
  • blk.6.attn_norm.weight
    F32
    [6144]
  • blk.6.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.6.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.6.attn_ssmax.bias
    F32
    [64]
  • blk.6.attn_ssmax.weight
    F32
    [64]
  • blk.6.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.6.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.6.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.6.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.6.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.6.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.6.ffn_norm.weight
    F32
    [6144]
  • blk.6.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.6.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.6.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.6.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.6.indexer.k_norm.bias
    F32
    [128]
  • blk.6.indexer.k_norm.weight
    F32
    [128]
  • blk.6.indexer.proj.weight
    F32
    [6144, 32]
  • blk.6.indexer.q_norm.bias
    F32
    [6144]
  • blk.6.indexer.q_norm.weight
    F32
    [6144]
  • blk.6.indexer.ssmax.bias
    F32
    [32]
  • blk.6.indexer.ssmax.weight
    F32
    [32]
  • blk.7
  • blk.7.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.7.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.7.attn_kv_a_norm.weight
    F32
    [512]
  • blk.7.attn_norm.weight
    F32
    [6144]
  • blk.7.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.7.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.7.attn_ssmax.bias
    F32
    [64]
  • blk.7.attn_ssmax.weight
    F32
    [64]
  • blk.7.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.7.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.7.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.7.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.7.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.7.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.7.ffn_norm.weight
    F32
    [6144]
  • blk.7.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.7.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.7.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.7.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.7.indexer.k_norm.bias
    F32
    [128]
  • blk.7.indexer.k_norm.weight
    F32
    [128]
  • blk.7.indexer.proj.weight
    F32
    [6144, 32]
  • blk.7.indexer.q_norm.bias
    F32
    [6144]
  • blk.7.indexer.q_norm.weight
    F32
    [6144]
  • blk.7.indexer.ssmax.bias
    F32
    [32]
  • blk.7.indexer.ssmax.weight
    F32
    [32]
  • blk.8
  • blk.8.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.8.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.8.attn_kv_a_norm.weight
    F32
    [512]
  • blk.8.attn_norm.weight
    F32
    [6144]
  • blk.8.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.8.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.8.attn_ssmax.bias
    F32
    [64]
  • blk.8.attn_ssmax.weight
    F32
    [64]
  • blk.8.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.8.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.8.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.8.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.8.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.8.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.8.ffn_norm.weight
    F32
    [6144]
  • blk.8.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.8.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.8.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.8.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.8.indexer.k_norm.bias
    F32
    [128]
  • blk.8.indexer.k_norm.weight
    F32
    [128]
  • blk.8.indexer.proj.weight
    F32
    [6144, 32]
  • blk.8.indexer.q_norm.bias
    F32
    [6144]
  • blk.8.indexer.q_norm.weight
    F32
    [6144]
  • blk.8.indexer.ssmax.bias
    F32
    [32]
  • blk.8.indexer.ssmax.weight
    F32
    [32]
  • blk.9
  • blk.9.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.9.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.9.attn_kv_a_norm.weight
    F32
    [512]
  • blk.9.attn_norm.weight
    F32
    [6144]
  • blk.9.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.9.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.9.attn_ssmax.bias
    F32
    [64]
  • blk.9.attn_ssmax.weight
    F32
    [64]
  • blk.9.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.9.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.9.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.9.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.9.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.9.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.9.ffn_norm.weight
    F32
    [6144]
  • blk.9.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.9.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.9.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.9.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.9.indexer.k_norm.bias
    F32
    [128]
  • blk.9.indexer.k_norm.weight
    F32
    [128]
  • blk.9.indexer.proj.weight
    F32
    [6144, 32]
  • blk.9.indexer.q_norm.bias
    F32
    [6144]
  • blk.9.indexer.q_norm.weight
    F32
    [6144]
  • blk.9.indexer.ssmax.bias
    F32
    [32]
  • blk.9.indexer.ssmax.weight
    F32
    [32]
  • blk.10
  • blk.10.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.10.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.10.attn_kv_a_norm.weight
    F32
    [512]
  • blk.10.attn_norm.weight
    F32
    [6144]
  • blk.10.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.10.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.10.attn_ssmax.bias
    F32
    [64]
  • blk.10.attn_ssmax.weight
    F32
    [64]
  • blk.10.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.10.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.10.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.10.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.10.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.10.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.10.ffn_norm.weight
    F32
    [6144]
  • blk.10.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.10.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.10.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.10.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.10.indexer.k_norm.bias
    F32
    [128]
  • blk.10.indexer.k_norm.weight
    F32
    [128]
  • blk.10.indexer.proj.weight
    F32
    [6144, 32]
  • blk.10.indexer.q_norm.bias
    F32
    [6144]
  • blk.10.indexer.q_norm.weight
    F32
    [6144]
  • blk.10.indexer.ssmax.bias
    F32
    [32]
  • blk.10.indexer.ssmax.weight
    F32
    [32]
  • blk.11
  • blk.11.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.11.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.11.attn_kv_a_norm.weight
    F32
    [512]
  • blk.11.attn_norm.weight
    F32
    [6144]
  • blk.11.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.11.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.11.attn_ssmax.bias
    F32
    [64]
  • blk.11.attn_ssmax.weight
    F32
    [64]
  • blk.11.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.11.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.11.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.11.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.11.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.11.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.11.ffn_norm.weight
    F32
    [6144]
  • blk.11.ffn_up_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.11.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.11.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.11.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.11.indexer.k_norm.bias
    F32
    [128]
  • blk.11.indexer.k_norm.weight
    F32
    [128]
  • blk.11.indexer.proj.weight
    F32
    [6144, 32]
  • blk.11.indexer.q_norm.bias
    F32
    [6144]
  • blk.11.indexer.q_norm.weight
    F32
    [6144]
  • blk.11.indexer.ssmax.bias
    F32
    [32]
  • blk.11.indexer.ssmax.weight
    F32
    [32]
  • blk.12
  • blk.12.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.12.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.12.attn_kv_a_norm.weight
    F32
    [512]
  • blk.12.attn_norm.weight
    F32
    [6144]
  • blk.12.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.12.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.12.attn_ssmax.bias
    F32
    [64]
  • blk.12.attn_ssmax.weight
    F32
    [64]
  • blk.12.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.12.ffn_down_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.12.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.12.ffn_gate_exps.weight
    IQ1_S
    [6144, 6144, 128]
  • blk.12.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.12.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.12.ffn_norm.weight
    F32
    [6144]
  • blk.12.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.12.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.12.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.12.indexer.k_norm.bias
    F32
    [128]
  • blk.12.indexer.k_norm.weight
    F32
    [128]
  • blk.12.indexer.proj.weight
    F32
    [6144, 32]
  • blk.12.indexer.q_norm.bias
    F32
    [6144]
  • blk.12.indexer.q_norm.weight
    F32
    [6144]
  • blk.12.indexer.ssmax.bias
    F32
    [32]
  • blk.12.indexer.ssmax.weight
    F32
    [32]
  • blk.13
  • blk.13.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.13.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.13.attn_kv_a_norm.weight
    F32
    [512]
  • blk.13.attn_norm.weight
    F32
    [6144]
  • blk.13.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.13.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.13.attn_ssmax.bias
    F32
    [64]
  • blk.13.attn_ssmax.weight
    F32
    [64]
  • blk.13.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.13.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.13.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.13.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.13.ffn_norm.weight
    F32
    [6144]
  • blk.13.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.13.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.13.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.13.indexer.k_norm.bias
    F32
    [128]
  • blk.13.indexer.k_norm.weight
    F32
    [128]
  • blk.13.indexer.proj.weight
    F32
    [6144, 32]
  • blk.13.indexer.q_norm.bias
    F32
    [6144]
  • blk.13.indexer.q_norm.weight
    F32
    [6144]
  • blk.13.indexer.ssmax.bias
    F32
    [32]
  • blk.13.indexer.ssmax.weight
    F32
    [32]
  • blk.14
  • blk.14.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.14.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.14.attn_kv_a_norm.weight
    F32
    [512]
  • blk.14.attn_norm.weight
    F32
    [6144]
  • blk.14.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.14.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.14.attn_ssmax.bias
    F32
    [64]
  • blk.14.attn_ssmax.weight
    F32
    [64]
  • blk.14.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.14.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.14.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.14.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.14.ffn_norm.weight
    F32
    [6144]
  • blk.14.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.14.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.14.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.14.indexer.k_norm.bias
    F32
    [128]
  • blk.14.indexer.k_norm.weight
    F32
    [128]
  • blk.14.indexer.proj.weight
    F32
    [6144, 32]
  • blk.14.indexer.q_norm.bias
    F32
    [6144]
  • blk.14.indexer.q_norm.weight
    F32
    [6144]
  • blk.14.indexer.ssmax.bias
    F32
    [32]
  • blk.14.indexer.ssmax.weight
    F32
    [32]
  • blk.15
  • blk.15.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.15.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.15.attn_kv_a_norm.weight
    F32
    [512]
  • blk.15.attn_norm.weight
    F32
    [6144]
  • blk.15.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.15.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.15.attn_ssmax.bias
    F32
    [64]
  • blk.15.attn_ssmax.weight
    F32
    [64]
  • blk.15.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.15.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.15.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.15.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.15.ffn_norm.weight
    F32
    [6144]
  • blk.15.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.15.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.15.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.15.indexer.k_norm.bias
    F32
    [128]
  • blk.15.indexer.k_norm.weight
    F32
    [128]
  • blk.15.indexer.proj.weight
    F32
    [6144, 32]
  • blk.15.indexer.q_norm.bias
    F32
    [6144]
  • blk.15.indexer.q_norm.weight
    F32
    [6144]
  • blk.15.indexer.ssmax.bias
    F32
    [32]
  • blk.15.indexer.ssmax.weight
    F32
    [32]
  • blk.16
  • blk.16.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.16.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.16.attn_kv_a_norm.weight
    F32
    [512]
  • blk.16.attn_norm.weight
    F32
    [6144]
  • blk.16.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.16.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.16.attn_ssmax.bias
    F32
    [64]
  • blk.16.attn_ssmax.weight
    F32
    [64]
  • blk.16.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.16.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.16.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.16.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.16.ffn_norm.weight
    F32
    [6144]
  • blk.16.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.16.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.16.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.16.indexer.k_norm.bias
    F32
    [128]
  • blk.16.indexer.k_norm.weight
    F32
    [128]
  • blk.16.indexer.proj.weight
    F32
    [6144, 32]
  • blk.16.indexer.q_norm.bias
    F32
    [6144]
  • blk.16.indexer.q_norm.weight
    F32
    [6144]
  • blk.16.indexer.ssmax.bias
    F32
    [32]
  • blk.16.indexer.ssmax.weight
    F32
    [32]
  • blk.17
  • blk.17.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.17.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.17.attn_kv_a_norm.weight
    F32
    [512]
  • blk.17.attn_norm.weight
    F32
    [6144]
  • blk.17.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.17.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.17.attn_ssmax.bias
    F32
    [64]
  • blk.17.attn_ssmax.weight
    F32
    [64]
  • blk.17.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.17.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.17.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.17.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.17.ffn_norm.weight
    F32
    [6144]
  • blk.17.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.18
  • blk.18.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.18.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.18.attn_kv_a_norm.weight
    F32
    [512]
  • blk.18.attn_norm.weight
    F32
    [6144]
  • blk.18.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.18.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.18.attn_ssmax.bias
    F32
    [64]
  • blk.18.attn_ssmax.weight
    F32
    [64]
  • blk.18.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.18.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.18.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.18.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.18.ffn_norm.weight
    F32
    [6144]
  • blk.18.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.19
  • blk.19.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.19.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.19.attn_kv_a_norm.weight
    F32
    [512]
  • blk.19.attn_norm.weight
    F32
    [6144]
  • blk.19.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.19.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.19.attn_ssmax.bias
    F32
    [64]
  • blk.19.attn_ssmax.weight
    F32
    [64]
  • blk.19.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.19.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.19.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.19.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.19.ffn_norm.weight
    F32
    [6144]
  • blk.19.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.19.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.19.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.19.indexer.k_norm.bias
    F32
    [128]
  • blk.19.indexer.k_norm.weight
    F32
    [128]
  • blk.19.indexer.proj.weight
    F32
    [6144, 32]
  • blk.19.indexer.q_norm.bias
    F32
    [6144]
  • blk.19.indexer.q_norm.weight
    F32
    [6144]
  • blk.19.indexer.ssmax.bias
    F32
    [32]
  • blk.19.indexer.ssmax.weight
    F32
    [32]
  • blk.20
  • blk.20.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.20.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.20.attn_kv_a_norm.weight
    F32
    [512]
  • blk.20.attn_norm.weight
    F32
    [6144]
  • blk.20.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.20.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.20.attn_ssmax.bias
    F32
    [64]
  • blk.20.attn_ssmax.weight
    F32
    [64]
  • blk.20.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.20.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.20.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.20.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.20.ffn_norm.weight
    F32
    [6144]
  • blk.20.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.20.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.20.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.20.indexer.k_norm.bias
    F32
    [128]
  • blk.20.indexer.k_norm.weight
    F32
    [128]
  • blk.20.indexer.proj.weight
    F32
    [6144, 32]
  • blk.20.indexer.q_norm.bias
    F32
    [6144]
  • blk.20.indexer.q_norm.weight
    F32
    [6144]
  • blk.20.indexer.ssmax.bias
    F32
    [32]
  • blk.20.indexer.ssmax.weight
    F32
    [32]
  • blk.21
  • blk.21.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.21.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.21.attn_kv_a_norm.weight
    F32
    [512]
  • blk.21.attn_norm.weight
    F32
    [6144]
  • blk.21.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.21.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.21.attn_ssmax.bias
    F32
    [64]
  • blk.21.attn_ssmax.weight
    F32
    [64]
  • blk.21.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.21.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.21.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.21.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.21.ffn_norm.weight
    F32
    [6144]
  • blk.21.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.22
  • blk.22.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.22.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.22.attn_kv_a_norm.weight
    F32
    [512]
  • blk.22.attn_norm.weight
    F32
    [6144]
  • blk.22.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.22.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.22.attn_ssmax.bias
    F32
    [64]
  • blk.22.attn_ssmax.weight
    F32
    [64]
  • blk.22.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.22.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.22.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.22.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.22.ffn_norm.weight
    F32
    [6144]
  • blk.22.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.23
  • blk.23.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.23.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.23.attn_kv_a_norm.weight
    F32
    [512]
  • blk.23.attn_norm.weight
    F32
    [6144]
  • blk.23.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.23.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.23.attn_ssmax.bias
    F32
    [64]
  • blk.23.attn_ssmax.weight
    F32
    [64]
  • blk.23.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.23.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.23.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.23.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.23.ffn_norm.weight
    F32
    [6144]
  • blk.23.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.23.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.23.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.23.indexer.k_norm.bias
    F32
    [128]
  • blk.23.indexer.k_norm.weight
    F32
    [128]
  • blk.23.indexer.proj.weight
    F32
    [6144, 32]
  • blk.23.indexer.q_norm.bias
    F32
    [6144]
  • blk.23.indexer.q_norm.weight
    F32
    [6144]
  • blk.23.indexer.ssmax.bias
    F32
    [32]
  • blk.23.indexer.ssmax.weight
    F32
    [32]
  • blk.24
  • blk.24.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.24.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.24.attn_kv_a_norm.weight
    F32
    [512]
  • blk.24.attn_norm.weight
    F32
    [6144]
  • blk.24.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.24.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.24.attn_ssmax.bias
    F32
    [64]
  • blk.24.attn_ssmax.weight
    F32
    [64]
  • blk.24.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.24.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.24.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.24.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.24.ffn_norm.weight
    F32
    [6144]
  • blk.24.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.24.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.24.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.24.indexer.k_norm.bias
    F32
    [128]
  • blk.24.indexer.k_norm.weight
    F32
    [128]
  • blk.24.indexer.proj.weight
    F32
    [6144, 32]
  • blk.24.indexer.q_norm.bias
    F32
    [6144]
  • blk.24.indexer.q_norm.weight
    F32
    [6144]
  • blk.24.indexer.ssmax.bias
    F32
    [32]
  • blk.24.indexer.ssmax.weight
    F32
    [32]
  • blk.25
  • blk.25.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.25.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.25.attn_kv_a_norm.weight
    F32
    [512]
  • blk.25.attn_norm.weight
    F32
    [6144]
  • blk.25.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.25.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.25.attn_ssmax.bias
    F32
    [64]
  • blk.25.attn_ssmax.weight
    F32
    [64]
  • blk.25.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.25.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.25.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.25.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.25.ffn_norm.weight
    F32
    [6144]
  • blk.25.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.26
  • blk.26.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.26.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.26.attn_kv_a_norm.weight
    F32
    [512]
  • blk.26.attn_norm.weight
    F32
    [6144]
  • blk.26.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.26.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.26.attn_ssmax.bias
    F32
    [64]
  • blk.26.attn_ssmax.weight
    F32
    [64]
  • blk.26.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.26.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.26.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.26.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.26.ffn_norm.weight
    F32
    [6144]
  • blk.26.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.27
  • blk.27.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.27.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.27.attn_kv_a_norm.weight
    F32
    [512]
  • blk.27.attn_norm.weight
    F32
    [6144]
  • blk.27.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.27.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.27.attn_ssmax.bias
    F32
    [64]
  • blk.27.attn_ssmax.weight
    F32
    [64]
  • blk.27.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.27.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.27.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.27.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.27.ffn_norm.weight
    F32
    [6144]
  • blk.27.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.27.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.27.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.27.indexer.k_norm.bias
    F32
    [128]
  • blk.27.indexer.k_norm.weight
    F32
    [128]
  • blk.27.indexer.proj.weight
    F32
    [6144, 32]
  • blk.27.indexer.q_norm.bias
    F32
    [6144]
  • blk.27.indexer.q_norm.weight
    F32
    [6144]
  • blk.27.indexer.ssmax.bias
    F32
    [32]
  • blk.27.indexer.ssmax.weight
    F32
    [32]
  • blk.28
  • blk.28.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.28.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.28.attn_kv_a_norm.weight
    F32
    [512]
  • blk.28.attn_norm.weight
    F32
    [6144]
  • blk.28.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.28.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.28.attn_ssmax.bias
    F32
    [64]
  • blk.28.attn_ssmax.weight
    F32
    [64]
  • blk.28.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.28.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.28.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.28.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.28.ffn_norm.weight
    F32
    [6144]
  • blk.28.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.28.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.28.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.28.indexer.k_norm.bias
    F32
    [128]
  • blk.28.indexer.k_norm.weight
    F32
    [128]
  • blk.28.indexer.proj.weight
    F32
    [6144, 32]
  • blk.28.indexer.q_norm.bias
    F32
    [6144]
  • blk.28.indexer.q_norm.weight
    F32
    [6144]
  • blk.28.indexer.ssmax.bias
    F32
    [32]
  • blk.28.indexer.ssmax.weight
    F32
    [32]
  • blk.29
  • blk.29.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.29.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.29.attn_kv_a_norm.weight
    F32
    [512]
  • blk.29.attn_norm.weight
    F32
    [6144]
  • blk.29.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.29.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.29.attn_ssmax.bias
    F32
    [64]
  • blk.29.attn_ssmax.weight
    F32
    [64]
  • blk.29.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.29.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.29.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.29.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.29.ffn_norm.weight
    F32
    [6144]
  • blk.29.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.30
  • blk.30.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.30.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.30.attn_kv_a_norm.weight
    F32
    [512]
  • blk.30.attn_norm.weight
    F32
    [6144]
  • blk.30.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.30.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.30.attn_ssmax.bias
    F32
    [64]
  • blk.30.attn_ssmax.weight
    F32
    [64]
  • blk.30.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.30.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.30.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.30.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.30.ffn_norm.weight
    F32
    [6144]
  • blk.30.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.31
  • blk.31.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.31.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.31.attn_kv_a_norm.weight
    F32
    [512]
  • blk.31.attn_norm.weight
    F32
    [6144]
  • blk.31.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.31.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.31.attn_ssmax.bias
    F32
    [64]
  • blk.31.attn_ssmax.weight
    F32
    [64]
  • blk.31.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.31.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.31.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.31.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.31.ffn_norm.weight
    F32
    [6144]
  • blk.31.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.31.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.31.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.31.indexer.k_norm.bias
    F32
    [128]
  • blk.31.indexer.k_norm.weight
    F32
    [128]
  • blk.31.indexer.proj.weight
    F32
    [6144, 32]
  • blk.31.indexer.q_norm.bias
    F32
    [6144]
  • blk.31.indexer.q_norm.weight
    F32
    [6144]
  • blk.31.indexer.ssmax.bias
    F32
    [32]
  • blk.31.indexer.ssmax.weight
    F32
    [32]
  • blk.32
  • blk.32.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.32.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.32.attn_kv_a_norm.weight
    F32
    [512]
  • blk.32.attn_norm.weight
    F32
    [6144]
  • blk.32.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.32.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.32.attn_ssmax.bias
    F32
    [64]
  • blk.32.attn_ssmax.weight
    F32
    [64]
  • blk.32.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.32.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.32.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.32.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.32.ffn_norm.weight
    F32
    [6144]
  • blk.32.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.32.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.32.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.32.indexer.k_norm.bias
    F32
    [128]
  • blk.32.indexer.k_norm.weight
    F32
    [128]
  • blk.32.indexer.proj.weight
    F32
    [6144, 32]
  • blk.32.indexer.q_norm.bias
    F32
    [6144]
  • blk.32.indexer.q_norm.weight
    F32
    [6144]
  • blk.32.indexer.ssmax.bias
    F32
    [32]
  • blk.32.indexer.ssmax.weight
    F32
    [32]
  • blk.33
  • blk.33.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.33.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.33.attn_kv_a_norm.weight
    F32
    [512]
  • blk.33.attn_norm.weight
    F32
    [6144]
  • blk.33.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.33.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.33.attn_ssmax.bias
    F32
    [64]
  • blk.33.attn_ssmax.weight
    F32
    [64]
  • blk.33.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.33.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.33.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.33.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.33.ffn_norm.weight
    F32
    [6144]
  • blk.33.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.34
  • blk.34.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.34.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.34.attn_kv_a_norm.weight
    F32
    [512]
  • blk.34.attn_norm.weight
    F32
    [6144]
  • blk.34.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.34.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.34.attn_ssmax.bias
    F32
    [64]
  • blk.34.attn_ssmax.weight
    F32
    [64]
  • blk.34.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.34.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.34.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.34.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.34.ffn_norm.weight
    F32
    [6144]
  • blk.34.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.35
  • blk.35.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.35.attn_kv_a_mqa.weight
    Q8_0
    [6144, 512]
  • blk.35.attn_kv_a_norm.weight
    F32
    [512]
  • blk.35.attn_norm.weight
    F32
    [6144]
  • blk.35.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.35.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.35.attn_ssmax.bias
    F32
    [64]
  • blk.35.attn_ssmax.weight
    F32
    [64]
  • blk.35.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.35.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.35.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.35.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.35.ffn_norm.weight
    F32
    [6144]
  • blk.35.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.35.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.35.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.35.indexer.k_norm.bias
    F32
    [128]
  • blk.35.indexer.k_norm.weight
    F32
    [128]
  • blk.35.indexer.proj.weight
    F32
    [6144, 32]
  • blk.35.indexer.q_norm.bias
    F32
    [6144]
  • blk.35.indexer.q_norm.weight
    F32
    [6144]
  • blk.35.indexer.ssmax.bias
    F32
    [32]
  • blk.35.indexer.ssmax.weight
    F32
    [32]
  • blk.36
  • blk.36.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.36.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.36.attn_kv_a_norm.weight
    F32
    [512]
  • blk.36.attn_norm.weight
    F32
    [6144]
  • blk.36.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.36.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.36.attn_ssmax.bias
    F32
    [64]
  • blk.36.attn_ssmax.weight
    F32
    [64]
  • blk.36.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.36.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.36.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.36.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.36.ffn_norm.weight
    F32
    [6144]
  • blk.36.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.36.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.36.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.36.indexer.k_norm.bias
    F32
    [128]
  • blk.36.indexer.k_norm.weight
    F32
    [128]
  • blk.36.indexer.proj.weight
    F32
    [6144, 32]
  • blk.36.indexer.q_norm.bias
    F32
    [6144]
  • blk.36.indexer.q_norm.weight
    F32
    [6144]
  • blk.36.indexer.ssmax.bias
    F32
    [32]
  • blk.36.indexer.ssmax.weight
    F32
    [32]
  • blk.37
  • blk.37.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.37.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.37.attn_kv_a_norm.weight
    F32
    [512]
  • blk.37.attn_norm.weight
    F32
    [6144]
  • blk.37.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.37.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.37.attn_ssmax.bias
    F32
    [64]
  • blk.37.attn_ssmax.weight
    F32
    [64]
  • blk.37.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.37.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.37.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.37.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.37.ffn_norm.weight
    F32
    [6144]
  • blk.37.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.38
  • blk.38.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.38.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.38.attn_kv_a_norm.weight
    F32
    [512]
  • blk.38.attn_norm.weight
    F32
    [6144]
  • blk.38.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.38.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.38.attn_ssmax.bias
    F32
    [64]
  • blk.38.attn_ssmax.weight
    F32
    [64]
  • blk.38.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.38.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.38.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.38.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.38.ffn_norm.weight
    F32
    [6144]
  • blk.38.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.39
  • blk.39.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.39.attn_norm.weight
    F32
    [6144]
  • blk.39.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.39.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.39.attn_ssmax.bias
    F32
    [64]
  • blk.39.attn_ssmax.weight
    F32
    [64]
  • blk.39.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.39.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.39.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.39.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.39.ffn_norm.weight
    F32
    [6144]
  • blk.39.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.39.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.39.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.39.indexer.k_norm.bias
    F32
    [128]
  • blk.39.indexer.k_norm.weight
    F32
    [128]
  • blk.39.indexer.proj.weight
    F32
    [6144, 32]
  • blk.39.indexer.q_norm.bias
    F32
    [6144]
  • blk.39.indexer.q_norm.weight
    F32
    [6144]
  • blk.39.indexer.ssmax.bias
    F32
    [32]
  • blk.39.indexer.ssmax.weight
    F32
    [32]
  • blk.40
  • blk.40.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.40.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.40.attn_kv_a_norm.weight
    F32
    [512]
  • blk.40.attn_norm.weight
    F32
    [6144]
  • blk.40.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.40.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.40.attn_ssmax.bias
    F32
    [64]
  • blk.40.attn_ssmax.weight
    F32
    [64]
  • blk.40.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.40.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.40.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.40.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.40.ffn_norm.weight
    F32
    [6144]
  • blk.40.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.40.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.40.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.40.indexer.k_norm.bias
    F32
    [128]
  • blk.40.indexer.k_norm.weight
    F32
    [128]
  • blk.40.indexer.proj.weight
    F32
    [6144, 32]
  • blk.40.indexer.q_norm.bias
    F32
    [6144]
  • blk.40.indexer.q_norm.weight
    F32
    [6144]
  • blk.40.indexer.ssmax.bias
    F32
    [32]
  • blk.40.indexer.ssmax.weight
    F32
    [32]
  • blk.41
  • blk.41.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.41.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.41.attn_kv_a_norm.weight
    F32
    [512]
  • blk.41.attn_norm.weight
    F32
    [6144]
  • blk.41.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.41.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.41.attn_ssmax.bias
    F32
    [64]
  • blk.41.attn_ssmax.weight
    F32
    [64]
  • blk.41.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.41.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.41.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.41.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.41.ffn_norm.weight
    F32
    [6144]
  • blk.41.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.42
  • blk.42.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.42.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.42.attn_kv_a_norm.weight
    F32
    [512]
  • blk.42.attn_norm.weight
    F32
    [6144]
  • blk.42.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.42.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.42.attn_ssmax.bias
    F32
    [64]
  • blk.42.attn_ssmax.weight
    F32
    [64]
  • blk.42.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.42.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.42.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.42.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.42.ffn_norm.weight
    F32
    [6144]
  • blk.42.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.43
  • blk.43.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.43.attn_norm.weight
    F32
    [6144]
  • blk.43.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.43.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.43.attn_ssmax.bias
    F32
    [64]
  • blk.43.attn_ssmax.weight
    F32
    [64]
  • blk.43.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.43.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.43.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.43.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.43.ffn_norm.weight
    F32
    [6144]
  • blk.43.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.43.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.43.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.43.indexer.k_norm.bias
    F32
    [128]
  • blk.43.indexer.k_norm.weight
    F32
    [128]
  • blk.43.indexer.proj.weight
    F32
    [6144, 32]
  • blk.43.indexer.q_norm.bias
    F32
    [6144]
  • blk.43.indexer.q_norm.weight
    F32
    [6144]
  • blk.43.indexer.ssmax.bias
    F32
    [32]
  • blk.43.indexer.ssmax.weight
    F32
    [32]
  • blk.44
  • blk.44.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.44.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.44.attn_kv_a_norm.weight
    F32
    [512]
  • blk.44.attn_norm.weight
    F32
    [6144]
  • blk.44.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.44.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.44.attn_ssmax.bias
    F32
    [64]
  • blk.44.attn_ssmax.weight
    F32
    [64]
  • blk.44.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.44.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.44.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.44.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.44.ffn_norm.weight
    F32
    [6144]
  • blk.44.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.44.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.44.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.44.indexer.k_norm.bias
    F32
    [128]
  • blk.44.indexer.k_norm.weight
    F32
    [128]
  • blk.44.indexer.proj.weight
    F32
    [6144, 32]
  • blk.44.indexer.q_norm.bias
    F32
    [6144]
  • blk.44.indexer.q_norm.weight
    F32
    [6144]
  • blk.44.indexer.ssmax.bias
    F32
    [32]
  • blk.44.indexer.ssmax.weight
    F32
    [32]
  • blk.45
  • blk.45.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.45.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.45.attn_kv_a_norm.weight
    F32
    [512]
  • blk.45.attn_norm.weight
    F32
    [6144]
  • blk.45.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.45.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.45.attn_ssmax.bias
    F32
    [64]
  • blk.45.attn_ssmax.weight
    F32
    [64]
  • blk.45.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.45.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.45.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.45.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.45.ffn_norm.weight
    F32
    [6144]
  • blk.45.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.46
  • blk.46.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.46.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.46.attn_kv_a_norm.weight
    F32
    [512]
  • blk.46.attn_norm.weight
    F32
    [6144]
  • blk.46.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.46.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.46.attn_ssmax.bias
    F32
    [64]
  • blk.46.attn_ssmax.weight
    F32
    [64]
  • blk.46.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.46.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.46.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.46.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.46.ffn_norm.weight
    F32
    [6144]
  • blk.46.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.47
  • blk.47.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.47.attn_norm.weight
    F32
    [6144]
  • blk.47.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.47.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.47.attn_ssmax.bias
    F32
    [64]
  • blk.47.attn_ssmax.weight
    F32
    [64]
  • blk.47.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.47.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.47.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.47.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.47.ffn_norm.weight
    F32
    [6144]
  • blk.47.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.47.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.47.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.47.indexer.k_norm.bias
    F32
    [128]
  • blk.47.indexer.k_norm.weight
    F32
    [128]
  • blk.47.indexer.proj.weight
    F32
    [6144, 32]
  • blk.47.indexer.q_norm.bias
    F32
    [6144]
  • blk.47.indexer.q_norm.weight
    F32
    [6144]
  • blk.47.indexer.ssmax.bias
    F32
    [32]
  • blk.47.indexer.ssmax.weight
    F32
    [32]
  • blk.48
  • blk.48.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.48.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.48.attn_kv_a_norm.weight
    F32
    [512]
  • blk.48.attn_norm.weight
    F32
    [6144]
  • blk.48.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.48.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.48.attn_ssmax.bias
    F32
    [64]
  • blk.48.attn_ssmax.weight
    F32
    [64]
  • blk.48.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.48.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.48.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.48.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.48.ffn_norm.weight
    F32
    [6144]
  • blk.48.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.48.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.48.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.48.indexer.k_norm.bias
    F32
    [128]
  • blk.48.indexer.k_norm.weight
    F32
    [128]
  • blk.48.indexer.proj.weight
    F32
    [6144, 32]
  • blk.48.indexer.q_norm.bias
    F32
    [6144]
  • blk.48.indexer.q_norm.weight
    F32
    [6144]
  • blk.48.indexer.ssmax.bias
    F32
    [32]
  • blk.48.indexer.ssmax.weight
    F32
    [32]
  • blk.49
  • blk.49.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.49.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.49.attn_kv_a_norm.weight
    F32
    [512]
  • blk.49.attn_norm.weight
    F32
    [6144]
  • blk.49.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.49.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.49.attn_ssmax.bias
    F32
    [64]
  • blk.49.attn_ssmax.weight
    F32
    [64]
  • blk.49.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.49.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.49.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.49.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.49.ffn_norm.weight
    F32
    [6144]
  • blk.49.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.50
  • blk.50.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.50.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.50.attn_kv_a_norm.weight
    F32
    [512]
  • blk.50.attn_norm.weight
    F32
    [6144]
  • blk.50.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.50.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.50.attn_ssmax.bias
    F32
    [64]
  • blk.50.attn_ssmax.weight
    F32
    [64]
  • blk.50.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.50.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.50.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.50.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.50.ffn_norm.weight
    F32
    [6144]
  • blk.50.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.51
  • blk.51.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.51.attn_norm.weight
    F32
    [6144]
  • blk.51.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.51.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.51.attn_ssmax.bias
    F32
    [64]
  • blk.51.attn_ssmax.weight
    F32
    [64]
  • blk.51.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.51.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.51.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.51.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.51.ffn_norm.weight
    F32
    [6144]
  • blk.51.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.51.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.51.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.51.indexer.k_norm.bias
    F32
    [128]
  • blk.51.indexer.k_norm.weight
    F32
    [128]
  • blk.51.indexer.proj.weight
    F32
    [6144, 32]
  • blk.51.indexer.q_norm.bias
    F32
    [6144]
  • blk.51.indexer.q_norm.weight
    F32
    [6144]
  • blk.51.indexer.ssmax.bias
    F32
    [32]
  • blk.51.indexer.ssmax.weight
    F32
    [32]
  • blk.52
  • blk.52.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.52.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.52.attn_kv_a_norm.weight
    F32
    [512]
  • blk.52.attn_norm.weight
    F32
    [6144]
  • blk.52.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.52.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.52.attn_ssmax.bias
    F32
    [64]
  • blk.52.attn_ssmax.weight
    F32
    [64]
  • blk.52.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.52.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.52.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.52.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.52.ffn_norm.weight
    F32
    [6144]
  • blk.52.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.52.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.52.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.52.indexer.k_norm.bias
    F32
    [128]
  • blk.52.indexer.k_norm.weight
    F32
    [128]
  • blk.52.indexer.proj.weight
    F32
    [6144, 32]
  • blk.52.indexer.q_norm.bias
    F32
    [6144]
  • blk.52.indexer.q_norm.weight
    F32
    [6144]
  • blk.52.indexer.ssmax.bias
    F32
    [32]
  • blk.52.indexer.ssmax.weight
    F32
    [32]
  • blk.53
  • blk.53.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.53.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.53.attn_kv_a_norm.weight
    F32
    [512]
  • blk.53.attn_norm.weight
    F32
    [6144]
  • blk.53.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.53.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.53.attn_ssmax.bias
    F32
    [64]
  • blk.53.attn_ssmax.weight
    F32
    [64]
  • blk.53.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.53.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.53.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.53.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.53.ffn_norm.weight
    F32
    [6144]
  • blk.53.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.54
  • blk.54.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.54.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.54.attn_kv_a_norm.weight
    F32
    [512]
  • blk.54.attn_norm.weight
    F32
    [6144]
  • blk.54.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.54.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.54.attn_ssmax.bias
    F32
    [64]
  • blk.54.attn_ssmax.weight
    F32
    [64]
  • blk.54.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.54.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.54.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.54.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.54.ffn_norm.weight
    F32
    [6144]
  • blk.54.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.55
  • blk.55.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.55.attn_norm.weight
    F32
    [6144]
  • blk.55.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.55.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.55.attn_ssmax.bias
    F32
    [64]
  • blk.55.attn_ssmax.weight
    F32
    [64]
  • blk.55.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.55.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.55.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.55.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.55.ffn_norm.weight
    F32
    [6144]
  • blk.55.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.55.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.55.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.55.indexer.k_norm.bias
    F32
    [128]
  • blk.55.indexer.k_norm.weight
    F32
    [128]
  • blk.55.indexer.proj.weight
    F32
    [6144, 32]
  • blk.55.indexer.q_norm.bias
    F32
    [6144]
  • blk.55.indexer.q_norm.weight
    F32
    [6144]
  • blk.55.indexer.ssmax.bias
    F32
    [32]
  • blk.55.indexer.ssmax.weight
    F32
    [32]
  • blk.56
  • blk.56.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.56.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.56.attn_kv_a_norm.weight
    F32
    [512]
  • blk.56.attn_norm.weight
    F32
    [6144]
  • blk.56.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.56.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.56.attn_ssmax.bias
    F32
    [64]
  • blk.56.attn_ssmax.weight
    F32
    [64]
  • blk.56.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.56.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.56.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.56.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.56.ffn_norm.weight
    F32
    [6144]
  • blk.56.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.56.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.56.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.56.indexer.k_norm.bias
    F32
    [128]
  • blk.56.indexer.k_norm.weight
    F32
    [128]
  • blk.56.indexer.proj.weight
    F32
    [6144, 32]
  • blk.56.indexer.q_norm.bias
    F32
    [6144]
  • blk.56.indexer.q_norm.weight
    F32
    [6144]
  • blk.56.indexer.ssmax.bias
    F32
    [32]
  • blk.56.indexer.ssmax.weight
    F32
    [32]
  • blk.57
  • blk.57.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.57.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.57.attn_kv_a_norm.weight
    F32
    [512]
  • blk.57.attn_norm.weight
    F32
    [6144]
  • blk.57.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.57.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.57.attn_ssmax.bias
    F32
    [64]
  • blk.57.attn_ssmax.weight
    F32
    [64]
  • blk.57.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.57.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.57.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.57.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.57.ffn_norm.weight
    F32
    [6144]
  • blk.57.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.58
  • blk.58.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.58.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.58.attn_kv_a_norm.weight
    F32
    [512]
  • blk.58.attn_norm.weight
    F32
    [6144]
  • blk.58.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.58.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.58.attn_ssmax.bias
    F32
    [64]
  • blk.58.attn_ssmax.weight
    F32
    [64]
  • blk.58.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.58.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.58.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.58.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.58.ffn_norm.weight
    F32
    [6144]
  • blk.58.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.59
  • blk.59.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.59.attn_norm.weight
    F32
    [6144]
  • blk.59.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.59.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.59.attn_ssmax.bias
    F32
    [64]
  • blk.59.attn_ssmax.weight
    F32
    [64]
  • blk.59.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.59.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.59.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.59.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.59.ffn_norm.weight
    F32
    [6144]
  • blk.59.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.59.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.59.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.59.indexer.k_norm.bias
    F32
    [128]
  • blk.59.indexer.k_norm.weight
    F32
    [128]
  • blk.59.indexer.proj.weight
    F32
    [6144, 32]
  • blk.59.indexer.q_norm.bias
    F32
    [6144]
  • blk.59.indexer.q_norm.weight
    F32
    [6144]
  • blk.59.indexer.ssmax.bias
    F32
    [32]
  • blk.59.indexer.ssmax.weight
    F32
    [32]
  • blk.60
  • blk.60.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.60.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.60.attn_kv_a_norm.weight
    F32
    [512]
  • blk.60.attn_norm.weight
    F32
    [6144]
  • blk.60.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.60.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.60.attn_ssmax.bias
    F32
    [64]
  • blk.60.attn_ssmax.weight
    F32
    [64]
  • blk.60.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.60.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.60.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.60.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.60.ffn_norm.weight
    F32
    [6144]
  • blk.60.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.60.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.60.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.60.indexer.k_norm.bias
    F32
    [128]
  • blk.60.indexer.k_norm.weight
    F32
    [128]
  • blk.60.indexer.proj.weight
    F32
    [6144, 32]
  • blk.60.indexer.q_norm.bias
    F32
    [6144]
  • blk.60.indexer.q_norm.weight
    F32
    [6144]
  • blk.60.indexer.ssmax.bias
    F32
    [32]
  • blk.60.indexer.ssmax.weight
    F32
    [32]
  • blk.61
  • blk.61.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.61.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.61.attn_kv_a_norm.weight
    F32
    [512]
  • blk.61.attn_norm.weight
    F32
    [6144]
  • blk.61.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.61.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.61.attn_ssmax.bias
    F32
    [64]
  • blk.61.attn_ssmax.weight
    F32
    [64]
  • blk.61.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.61.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.61.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.61.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.61.ffn_norm.weight
    F32
    [6144]
  • blk.61.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.62
  • blk.62.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.62.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.62.attn_kv_a_norm.weight
    F32
    [512]
  • blk.62.attn_norm.weight
    F32
    [6144]
  • blk.62.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.62.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.62.attn_ssmax.bias
    F32
    [64]
  • blk.62.attn_ssmax.weight
    F32
    [64]
  • blk.62.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.62.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.62.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.62.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.62.ffn_norm.weight
    F32
    [6144]
  • blk.62.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.63
  • blk.63.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.63.attn_norm.weight
    F32
    [6144]
  • blk.63.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.63.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.63.attn_ssmax.bias
    F32
    [64]
  • blk.63.attn_ssmax.weight
    F32
    [64]
  • blk.63.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.63.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.63.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.63.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.63.ffn_norm.weight
    F32
    [6144]
  • blk.63.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.63.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.63.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.63.indexer.k_norm.bias
    F32
    [128]
  • blk.63.indexer.k_norm.weight
    F32
    [128]
  • blk.63.indexer.proj.weight
    F32
    [6144, 32]
  • blk.63.indexer.q_norm.bias
    F32
    [6144]
  • blk.63.indexer.q_norm.weight
    F32
    [6144]
  • blk.63.indexer.ssmax.bias
    F32
    [32]
  • blk.63.indexer.ssmax.weight
    F32
    [32]
  • blk.64
  • blk.64.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.64.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.64.attn_kv_a_norm.weight
    F32
    [512]
  • blk.64.attn_norm.weight
    F32
    [6144]
  • blk.64.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.64.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.64.attn_ssmax.bias
    F32
    [64]
  • blk.64.attn_ssmax.weight
    F32
    [64]
  • blk.64.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.64.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.64.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.64.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.64.ffn_norm.weight
    F32
    [6144]
  • blk.64.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.64.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.64.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.64.indexer.k_norm.bias
    F32
    [128]
  • blk.64.indexer.k_norm.weight
    F32
    [128]
  • blk.64.indexer.proj.weight
    F32
    [6144, 32]
  • blk.64.indexer.q_norm.bias
    F32
    [6144]
  • blk.64.indexer.q_norm.weight
    F32
    [6144]
  • blk.64.indexer.ssmax.bias
    F32
    [32]
  • blk.64.indexer.ssmax.weight
    F32
    [32]
  • blk.65
  • blk.65.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.65.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.65.attn_kv_a_norm.weight
    F32
    [512]
  • blk.65.attn_norm.weight
    F32
    [6144]
  • blk.65.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.65.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.65.attn_ssmax.bias
    F32
    [64]
  • blk.65.attn_ssmax.weight
    F32
    [64]
  • blk.65.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.65.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.65.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.65.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.65.ffn_norm.weight
    F32
    [6144]
  • blk.65.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.66
  • blk.66.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.66.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.66.attn_kv_a_norm.weight
    F32
    [512]
  • blk.66.attn_norm.weight
    F32
    [6144]
  • blk.66.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.66.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.66.attn_ssmax.bias
    F32
    [64]
  • blk.66.attn_ssmax.weight
    F32
    [64]
  • blk.66.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.66.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.66.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.66.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.66.ffn_norm.weight
    F32
    [6144]
  • blk.66.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.67
  • blk.67.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.67.attn_norm.weight
    F32
    [6144]
  • blk.67.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.67.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.67.attn_ssmax.bias
    F32
    [64]
  • blk.67.attn_ssmax.weight
    F32
    [64]
  • blk.67.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.67.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.67.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.67.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.67.ffn_norm.weight
    F32
    [6144]
  • blk.67.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.67.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.67.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.67.indexer.k_norm.bias
    F32
    [128]
  • blk.67.indexer.k_norm.weight
    F32
    [128]
  • blk.67.indexer.proj.weight
    F32
    [6144, 32]
  • blk.67.indexer.q_norm.bias
    F32
    [6144]
  • blk.67.indexer.q_norm.weight
    F32
    [6144]
  • blk.67.indexer.ssmax.bias
    F32
    [32]
  • blk.67.indexer.ssmax.weight
    F32
    [32]
  • blk.68
  • blk.68.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.68.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.68.attn_kv_a_norm.weight
    F32
    [512]
  • blk.68.attn_norm.weight
    F32
    [6144]
  • blk.68.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.68.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.68.attn_ssmax.bias
    F32
    [64]
  • blk.68.attn_ssmax.weight
    F32
    [64]
  • blk.68.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.68.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.68.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.68.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.68.ffn_norm.weight
    F32
    [6144]
  • blk.68.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.68.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.68.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.68.indexer.k_norm.bias
    F32
    [128]
  • blk.68.indexer.k_norm.weight
    F32
    [128]
  • blk.68.indexer.proj.weight
    F32
    [6144, 32]
  • blk.68.indexer.q_norm.bias
    F32
    [6144]
  • blk.68.indexer.q_norm.weight
    F32
    [6144]
  • blk.68.indexer.ssmax.bias
    F32
    [32]
  • blk.68.indexer.ssmax.weight
    F32
    [32]
  • blk.69
  • blk.69.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.69.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.69.attn_kv_a_norm.weight
    F32
    [512]
  • blk.69.attn_norm.weight
    F32
    [6144]
  • blk.69.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.69.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.69.attn_ssmax.bias
    F32
    [64]
  • blk.69.attn_ssmax.weight
    F32
    [64]
  • blk.69.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.69.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.69.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.69.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.69.ffn_norm.weight
    F32
    [6144]
  • blk.69.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.70
  • blk.70.attn_k_b.weight
    Q8_0
    [64, 512, 64]
  • blk.70.attn_kv_a_mqa.weight
    Q8_0
    [6144, 576]
  • blk.70.attn_kv_a_norm.weight
    F32
    [512]
  • blk.70.attn_norm.weight
    F32
    [6144]
  • blk.70.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.70.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.70.attn_ssmax.bias
    F32
    [64]
  • blk.70.attn_ssmax.weight
    F32
    [64]
  • blk.70.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.70.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.70.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.70.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.70.ffn_norm.weight
    F32
    [6144]
  • blk.70.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.71
  • blk.71.attn_k_b.weight
    Q8_0
    [128, 512, 64]
  • blk.71.attn_norm.weight
    F32
    [6144]
  • blk.71.attn_output.weight
    Q4_K
    [8192, 6144]
  • blk.71.attn_q.weight
    IQ4_XS
    [6144, 8192]
  • blk.71.attn_ssmax.bias
    F32
    [64]
  • blk.71.attn_ssmax.weight
    F32
    [64]
  • blk.71.attn_v_b.weight
    Q5_K
    [512, 128, 64]
  • blk.71.ffn_down_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.71.ffn_gate_inp.weight
    F32
    [6144, 128]
  • blk.71.ffn_gate_shexp.weight
    Q4_K
    [6144, 6144]
  • blk.71.ffn_norm.weight
    F32
    [6144]
  • blk.71.ffn_up_shexp.weight
    IQ4_XS
    [6144, 6144]
  • blk.71.indexer.attn_k.weight
    Q8_0
    [6144, 128]
  • blk.71.indexer.attn_q_b.weight
    IQ4_XS
    [6144, 4096]
  • blk.71.indexer.k_norm.bias
    F32
    [128]
  • blk.71.indexer.k_norm.weight
    F32
    [128]
  • blk.71.indexer.proj.weight
    F32
    [6144, 32]
  • blk.71.indexer.q_norm.bias
    F32
    [6144]
  • blk.71.indexer.q_norm.weight
    F32
    [6144]
  • blk.71.indexer.ssmax.bias
    F32
    [32]
  • blk.71.indexer.ssmax.weight
    F32
    [32]
  • output.weight
    Q5_K
    [6144, 204800]
  • rope_freqs.weight
    F32
    [32]
  • output_norm.weight
    F32
    [6144]