220 1 week ago

New version GLM-4.6

579108167db6 · 216GB
    Metadata
  • general.architecture
    glm4moe
  • general.file_type
    Q4_K_M
  • glm4moe.attention.head_count
    96
  • glm4moe.attention.head_count_kv
    8
  • glm4moe.attention.key_length
    128
  • glm4moe.attention.layer_norm_rms_epsilon
    1e-05
  • glm4moe.attention.value_length
    128
  • glm4moe.block_count
    93
  • glm4moe.context_length
    202752
  • glm4moe.embedding_length
    5120
  • glm4moe.expert_count
    160
  • glm4moe.expert_feed_forward_length
    1536
  • glm4moe.expert_gating_func
    2
  • glm4moe.expert_shared_count
    1
  • glm4moe.expert_used_count
    8
  • glm4moe.expert_weights_norm
    true
  • glm4moe.expert_weights_scale
    2.5
  • glm4moe.feed_forward_length
    12288
  • glm4moe.leading_dense_block_count
    3
  • glm4moe.nextn_predict_layers
    1
  • glm4moe.rope.dimension_count
    64
  • glm4moe.rope.freq_base
    1e+06
  • tokenizer.ggml.bos_token_id
    151331
  • tokenizer.ggml.eom_token_id
    151338
  • tokenizer.ggml.eos_token_id
    151329
  • tokenizer.ggml.eot_token_id
    151336
  • tokenizer.ggml.merges
    [Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, i n, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    151330
  • tokenizer.ggml.pre
    glm4
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
  • tokenizer.ggml.unknown_token_id
    151329
  • quantize.imatrix.chunks_count
    51
  • quantize.imatrix.dataset
    unsloth_calibration_GLM-4.6.txt
  • quantize.imatrix.entries_count
    1000
  • quantize.imatrix.file
    GLM-4.6-GGUF/imatrix_unsloth.gguf
  • split.count
    0
  • split.no
    0
  • split.tensors.count
    1759
  • Tensor
  • token_embd.weight
    Q4_K
    [5120, 151552]
  • blk.0
  • blk.0.attn_k.bias
    F32
    [1024]
  • blk.0.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.0.attn_k_norm.weight
    F32
    [128]
  • blk.0.attn_norm.weight
    F32
    [5120]
  • blk.0.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.0.attn_q.bias
    F32
    [12288]
  • blk.0.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.0.attn_q_norm.weight
    F32
    [128]
  • blk.0.attn_v.bias
    F32
    [1024]
  • blk.0.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.0.ffn_down.weight
    Q6_K
    [12288, 5120]
  • blk.0.ffn_gate.weight
    Q4_K
    [5120, 12288]
  • blk.0.ffn_up.weight
    Q4_K
    [5120, 12288]
  • blk.0.post_attention_norm.weight
    F32
    [5120]
  • blk.1
  • blk.1.attn_k.bias
    F32
    [1024]
  • blk.1.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.1.attn_k_norm.weight
    F32
    [128]
  • blk.1.attn_norm.weight
    F32
    [5120]
  • blk.1.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.1.attn_q.bias
    F32
    [12288]
  • blk.1.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.1.attn_q_norm.weight
    F32
    [128]
  • blk.1.attn_v.bias
    F32
    [1024]
  • blk.1.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.1.ffn_down.weight
    Q6_K
    [12288, 5120]
  • blk.1.ffn_gate.weight
    Q4_K
    [5120, 12288]
  • blk.1.ffn_up.weight
    Q4_K
    [5120, 12288]
  • blk.1.post_attention_norm.weight
    F32
    [5120]
  • blk.2
  • blk.2.attn_k.bias
    F32
    [1024]
  • blk.2.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.2.attn_k_norm.weight
    F32
    [128]
  • blk.2.attn_norm.weight
    F32
    [5120]
  • blk.2.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.2.attn_q.bias
    F32
    [12288]
  • blk.2.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.2.attn_q_norm.weight
    F32
    [128]
  • blk.2.attn_v.bias
    F32
    [1024]
  • blk.2.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.2.ffn_down.weight
    Q6_K
    [12288, 5120]
  • blk.2.ffn_gate.weight
    Q4_K
    [5120, 12288]
  • blk.2.ffn_up.weight
    Q4_K
    [5120, 12288]
  • blk.2.post_attention_norm.weight
    F32
    [5120]
  • blk.3
  • blk.3.attn_k.bias
    F32
    [1024]
  • blk.3.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.3.attn_k_norm.weight
    F32
    [128]
  • blk.3.attn_norm.weight
    F32
    [5120]
  • blk.3.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.3.attn_q.bias
    F32
    [12288]
  • blk.3.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.3.attn_q_norm.weight
    F32
    [128]
  • blk.3.attn_v.bias
    F32
    [1024]
  • blk.3.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.3.exp_probs_b.bias
    F32
    [160]
  • blk.3.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.3.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.3.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.3.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.3.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.3.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.3.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.3.post_attention_norm.weight
    F32
    [5120]
  • blk.4
  • blk.4.attn_k.bias
    F32
    [1024]
  • blk.4.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.4.attn_k_norm.weight
    F32
    [128]
  • blk.4.attn_norm.weight
    F32
    [5120]
  • blk.4.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.4.attn_q.bias
    F32
    [12288]
  • blk.4.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.4.attn_q_norm.weight
    F32
    [128]
  • blk.4.attn_v.bias
    F32
    [1024]
  • blk.4.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.4.exp_probs_b.bias
    F32
    [160]
  • blk.4.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.4.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.4.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.4.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.4.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.4.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.4.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.4.post_attention_norm.weight
    F32
    [5120]
  • blk.5
  • blk.5.attn_k.bias
    F32
    [1024]
  • blk.5.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.5.attn_k_norm.weight
    F32
    [128]
  • blk.5.attn_norm.weight
    F32
    [5120]
  • blk.5.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.5.attn_q.bias
    F32
    [12288]
  • blk.5.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.5.attn_q_norm.weight
    F32
    [128]
  • blk.5.attn_v.bias
    F32
    [1024]
  • blk.5.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.5.exp_probs_b.bias
    F32
    [160]
  • blk.5.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.5.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.5.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.5.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.5.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.5.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.5.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.5.post_attention_norm.weight
    F32
    [5120]
  • blk.6
  • blk.6.attn_k.bias
    F32
    [1024]
  • blk.6.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.6.attn_k_norm.weight
    F32
    [128]
  • blk.6.attn_norm.weight
    F32
    [5120]
  • blk.6.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.6.attn_q.bias
    F32
    [12288]
  • blk.6.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.6.attn_q_norm.weight
    F32
    [128]
  • blk.6.attn_v.bias
    F32
    [1024]
  • blk.6.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.6.exp_probs_b.bias
    F32
    [160]
  • blk.6.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.6.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.6.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.6.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.6.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.6.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.6.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.6.post_attention_norm.weight
    F32
    [5120]
  • blk.7
  • blk.7.attn_k.bias
    F32
    [1024]
  • blk.7.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.7.attn_k_norm.weight
    F32
    [128]
  • blk.7.attn_norm.weight
    F32
    [5120]
  • blk.7.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.7.attn_q.bias
    F32
    [12288]
  • blk.7.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.7.attn_q_norm.weight
    F32
    [128]
  • blk.7.attn_v.bias
    F32
    [1024]
  • blk.7.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.7.exp_probs_b.bias
    F32
    [160]
  • blk.7.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.7.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.7.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.7.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.7.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.7.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.7.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.7.post_attention_norm.weight
    F32
    [5120]
  • blk.8
  • blk.8.attn_k.bias
    F32
    [1024]
  • blk.8.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.8.attn_k_norm.weight
    F32
    [128]
  • blk.8.attn_norm.weight
    F32
    [5120]
  • blk.8.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.8.attn_q.bias
    F32
    [12288]
  • blk.8.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.8.attn_q_norm.weight
    F32
    [128]
  • blk.8.attn_v.bias
    F32
    [1024]
  • blk.8.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.8.exp_probs_b.bias
    F32
    [160]
  • blk.8.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.8.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.8.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.8.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.8.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.8.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.8.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.8.post_attention_norm.weight
    F32
    [5120]
  • blk.9
  • blk.9.attn_k.bias
    F32
    [1024]
  • blk.9.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.9.attn_k_norm.weight
    F32
    [128]
  • blk.9.attn_norm.weight
    F32
    [5120]
  • blk.9.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.9.attn_q.bias
    F32
    [12288]
  • blk.9.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.9.attn_q_norm.weight
    F32
    [128]
  • blk.9.attn_v.bias
    F32
    [1024]
  • blk.9.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.9.exp_probs_b.bias
    F32
    [160]
  • blk.9.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.9.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.9.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.9.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.9.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.9.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.9.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.9.post_attention_norm.weight
    F32
    [5120]
  • blk.10
  • blk.10.attn_k.bias
    F32
    [1024]
  • blk.10.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.10.attn_k_norm.weight
    F32
    [128]
  • blk.10.attn_norm.weight
    F32
    [5120]
  • blk.10.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.10.attn_q.bias
    F32
    [12288]
  • blk.10.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.10.attn_q_norm.weight
    F32
    [128]
  • blk.10.attn_v.bias
    F32
    [1024]
  • blk.10.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.10.exp_probs_b.bias
    F32
    [160]
  • blk.10.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.10.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.10.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.10.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.10.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.10.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.10.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.10.post_attention_norm.weight
    F32
    [5120]
  • blk.11
  • blk.11.attn_k.bias
    F32
    [1024]
  • blk.11.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.11.attn_k_norm.weight
    F32
    [128]
  • blk.11.attn_norm.weight
    F32
    [5120]
  • blk.11.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.11.attn_q.bias
    F32
    [12288]
  • blk.11.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.11.attn_q_norm.weight
    F32
    [128]
  • blk.11.attn_v.bias
    F32
    [1024]
  • blk.11.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.11.exp_probs_b.bias
    F32
    [160]
  • blk.11.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.11.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.11.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.11.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.11.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.11.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.11.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.11.post_attention_norm.weight
    F32
    [5120]
  • blk.12
  • blk.12.attn_k.bias
    F32
    [1024]
  • blk.12.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.12.attn_k_norm.weight
    F32
    [128]
  • blk.12.attn_norm.weight
    F32
    [5120]
  • blk.12.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.12.attn_q.bias
    F32
    [12288]
  • blk.12.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.12.attn_q_norm.weight
    F32
    [128]
  • blk.12.attn_v.bias
    F32
    [1024]
  • blk.12.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.12.exp_probs_b.bias
    F32
    [160]
  • blk.12.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.12.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.12.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.12.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.12.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.12.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.12.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.12.post_attention_norm.weight
    F32
    [5120]
  • blk.13
  • blk.13.attn_k.bias
    F32
    [1024]
  • blk.13.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.13.attn_k_norm.weight
    F32
    [128]
  • blk.13.attn_norm.weight
    F32
    [5120]
  • blk.13.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.13.attn_q.bias
    F32
    [12288]
  • blk.13.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.13.attn_q_norm.weight
    F32
    [128]
  • blk.13.attn_v.bias
    F32
    [1024]
  • blk.13.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.13.exp_probs_b.bias
    F32
    [160]
  • blk.13.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.13.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.13.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.13.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.13.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.13.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.13.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.13.post_attention_norm.weight
    F32
    [5120]
  • blk.14
  • blk.14.attn_k.bias
    F32
    [1024]
  • blk.14.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.14.attn_k_norm.weight
    F32
    [128]
  • blk.14.attn_norm.weight
    F32
    [5120]
  • blk.14.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.14.attn_q.bias
    F32
    [12288]
  • blk.14.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.14.attn_q_norm.weight
    F32
    [128]
  • blk.14.attn_v.bias
    F32
    [1024]
  • blk.14.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.14.exp_probs_b.bias
    F32
    [160]
  • blk.14.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.14.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.14.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.14.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.14.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.14.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.14.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.14.post_attention_norm.weight
    F32
    [5120]
  • blk.15
  • blk.15.attn_k.bias
    F32
    [1024]
  • blk.15.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.15.attn_k_norm.weight
    F32
    [128]
  • blk.15.attn_norm.weight
    F32
    [5120]
  • blk.15.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.15.attn_q.bias
    F32
    [12288]
  • blk.15.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.15.attn_q_norm.weight
    F32
    [128]
  • blk.15.attn_v.bias
    F32
    [1024]
  • blk.15.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.15.exp_probs_b.bias
    F32
    [160]
  • blk.15.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.15.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.15.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.15.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.15.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.15.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.15.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.15.post_attention_norm.weight
    F32
    [5120]
  • blk.16
  • blk.16.attn_k.bias
    F32
    [1024]
  • blk.16.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.16.attn_k_norm.weight
    F32
    [128]
  • blk.16.attn_norm.weight
    F32
    [5120]
  • blk.16.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.16.attn_q.bias
    F32
    [12288]
  • blk.16.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.16.attn_q_norm.weight
    F32
    [128]
  • blk.16.attn_v.bias
    F32
    [1024]
  • blk.16.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.16.exp_probs_b.bias
    F32
    [160]
  • blk.16.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.16.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.16.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.16.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.16.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.16.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.16.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.16.post_attention_norm.weight
    F32
    [5120]
  • blk.17
  • blk.17.attn_k.bias
    F32
    [1024]
  • blk.17.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.17.attn_k_norm.weight
    F32
    [128]
  • blk.17.attn_norm.weight
    F32
    [5120]
  • blk.17.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.17.attn_q.bias
    F32
    [12288]
  • blk.17.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.17.attn_q_norm.weight
    F32
    [128]
  • blk.17.attn_v.bias
    F32
    [1024]
  • blk.17.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.17.exp_probs_b.bias
    F32
    [160]
  • blk.17.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.17.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.17.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.17.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.17.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.17.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.17.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.17.post_attention_norm.weight
    F32
    [5120]
  • blk.18
  • blk.18.attn_k.bias
    F32
    [1024]
  • blk.18.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.18.attn_k_norm.weight
    F32
    [128]
  • blk.18.attn_norm.weight
    F32
    [5120]
  • blk.18.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.18.attn_q.bias
    F32
    [12288]
  • blk.18.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.18.attn_q_norm.weight
    F32
    [128]
  • blk.18.attn_v.bias
    F32
    [1024]
  • blk.18.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.18.exp_probs_b.bias
    F32
    [160]
  • blk.18.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.18.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.18.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.18.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.18.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.18.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.18.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.18.post_attention_norm.weight
    F32
    [5120]
  • blk.19
  • blk.19.attn_k.bias
    F32
    [1024]
  • blk.19.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.19.attn_k_norm.weight
    F32
    [128]
  • blk.19.attn_norm.weight
    F32
    [5120]
  • blk.19.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.19.attn_q.bias
    F32
    [12288]
  • blk.19.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.19.attn_q_norm.weight
    F32
    [128]
  • blk.19.attn_v.bias
    F32
    [1024]
  • blk.19.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.19.exp_probs_b.bias
    F32
    [160]
  • blk.19.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.19.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.19.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.19.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.19.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.19.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.19.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.19.post_attention_norm.weight
    F32
    [5120]
  • blk.20
  • blk.20.attn_k.bias
    F32
    [1024]
  • blk.20.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.20.attn_k_norm.weight
    F32
    [128]
  • blk.20.attn_norm.weight
    F32
    [5120]
  • blk.20.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.20.attn_q.bias
    F32
    [12288]
  • blk.20.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.20.attn_q_norm.weight
    F32
    [128]
  • blk.20.attn_v.bias
    F32
    [1024]
  • blk.20.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.20.exp_probs_b.bias
    F32
    [160]
  • blk.20.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.20.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.20.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.20.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.20.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.20.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.20.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.20.post_attention_norm.weight
    F32
    [5120]
  • blk.21
  • blk.21.attn_k.bias
    F32
    [1024]
  • blk.21.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.21.attn_k_norm.weight
    F32
    [128]
  • blk.21.attn_norm.weight
    F32
    [5120]
  • blk.21.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.21.attn_q.bias
    F32
    [12288]
  • blk.21.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.21.attn_q_norm.weight
    F32
    [128]
  • blk.21.attn_v.bias
    F32
    [1024]
  • blk.21.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.21.exp_probs_b.bias
    F32
    [160]
  • blk.21.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.21.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.21.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.21.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.21.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.21.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.21.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.21.post_attention_norm.weight
    F32
    [5120]
  • blk.22
  • blk.22.attn_k.bias
    F32
    [1024]
  • blk.22.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.22.attn_k_norm.weight
    F32
    [128]
  • blk.22.attn_norm.weight
    F32
    [5120]
  • blk.22.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.22.attn_q.bias
    F32
    [12288]
  • blk.22.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.22.attn_q_norm.weight
    F32
    [128]
  • blk.22.attn_v.bias
    F32
    [1024]
  • blk.22.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.22.exp_probs_b.bias
    F32
    [160]
  • blk.22.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.22.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.22.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.22.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.22.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.22.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.22.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.22.post_attention_norm.weight
    F32
    [5120]
  • blk.23
  • blk.23.attn_k.bias
    F32
    [1024]
  • blk.23.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.23.attn_k_norm.weight
    F32
    [128]
  • blk.23.attn_norm.weight
    F32
    [5120]
  • blk.23.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.23.attn_q.bias
    F32
    [12288]
  • blk.23.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.23.attn_q_norm.weight
    F32
    [128]
  • blk.23.attn_v.bias
    F32
    [1024]
  • blk.23.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.23.exp_probs_b.bias
    F32
    [160]
  • blk.23.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.23.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.23.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.23.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.23.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.23.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.23.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.23.post_attention_norm.weight
    F32
    [5120]
  • blk.24
  • blk.24.attn_k.bias
    F32
    [1024]
  • blk.24.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.24.attn_k_norm.weight
    F32
    [128]
  • blk.24.attn_norm.weight
    F32
    [5120]
  • blk.24.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.24.attn_q.bias
    F32
    [12288]
  • blk.24.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.24.attn_q_norm.weight
    F32
    [128]
  • blk.24.attn_v.bias
    F32
    [1024]
  • blk.24.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.24.exp_probs_b.bias
    F32
    [160]
  • blk.24.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.24.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.24.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.24.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.24.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.24.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.24.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.24.post_attention_norm.weight
    F32
    [5120]
  • blk.25
  • blk.25.attn_k.bias
    F32
    [1024]
  • blk.25.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.25.attn_k_norm.weight
    F32
    [128]
  • blk.25.attn_norm.weight
    F32
    [5120]
  • blk.25.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.25.attn_q.bias
    F32
    [12288]
  • blk.25.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.25.attn_q_norm.weight
    F32
    [128]
  • blk.25.attn_v.bias
    F32
    [1024]
  • blk.25.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.25.exp_probs_b.bias
    F32
    [160]
  • blk.25.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.25.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.25.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.25.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.25.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.25.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.25.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.25.post_attention_norm.weight
    F32
    [5120]
  • blk.26
  • blk.26.attn_k.bias
    F32
    [1024]
  • blk.26.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.26.attn_k_norm.weight
    F32
    [128]
  • blk.26.attn_norm.weight
    F32
    [5120]
  • blk.26.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.26.attn_q.bias
    F32
    [12288]
  • blk.26.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.26.attn_q_norm.weight
    F32
    [128]
  • blk.26.attn_v.bias
    F32
    [1024]
  • blk.26.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.26.exp_probs_b.bias
    F32
    [160]
  • blk.26.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.26.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.26.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.26.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.26.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.26.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.26.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.26.post_attention_norm.weight
    F32
    [5120]
  • blk.27
  • blk.27.attn_k.bias
    F32
    [1024]
  • blk.27.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.27.attn_k_norm.weight
    F32
    [128]
  • blk.27.attn_norm.weight
    F32
    [5120]
  • blk.27.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.27.attn_q.bias
    F32
    [12288]
  • blk.27.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.27.attn_q_norm.weight
    F32
    [128]
  • blk.27.attn_v.bias
    F32
    [1024]
  • blk.27.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.27.exp_probs_b.bias
    F32
    [160]
  • blk.27.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.27.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.27.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.27.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.27.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.27.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.27.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.27.post_attention_norm.weight
    F32
    [5120]
  • blk.28
  • blk.28.attn_k.bias
    F32
    [1024]
  • blk.28.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.28.attn_k_norm.weight
    F32
    [128]
  • blk.28.attn_norm.weight
    F32
    [5120]
  • blk.28.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.28.attn_q.bias
    F32
    [12288]
  • blk.28.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.28.attn_q_norm.weight
    F32
    [128]
  • blk.28.attn_v.bias
    F32
    [1024]
  • blk.28.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.28.exp_probs_b.bias
    F32
    [160]
  • blk.28.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.28.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.28.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.28.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.28.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.28.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.28.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.28.post_attention_norm.weight
    F32
    [5120]
  • blk.29
  • blk.29.attn_k.bias
    F32
    [1024]
  • blk.29.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.29.attn_k_norm.weight
    F32
    [128]
  • blk.29.attn_norm.weight
    F32
    [5120]
  • blk.29.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.29.attn_q.bias
    F32
    [12288]
  • blk.29.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.29.attn_q_norm.weight
    F32
    [128]
  • blk.29.attn_v.bias
    F32
    [1024]
  • blk.29.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.29.exp_probs_b.bias
    F32
    [160]
  • blk.29.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.29.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.29.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.29.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.29.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.29.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.29.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.29.post_attention_norm.weight
    F32
    [5120]
  • blk.30
  • blk.30.attn_k.bias
    F32
    [1024]
  • blk.30.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.30.attn_k_norm.weight
    F32
    [128]
  • blk.30.attn_norm.weight
    F32
    [5120]
  • blk.30.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.30.attn_q.bias
    F32
    [12288]
  • blk.30.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.30.attn_q_norm.weight
    F32
    [128]
  • blk.30.attn_v.bias
    F32
    [1024]
  • blk.30.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.30.exp_probs_b.bias
    F32
    [160]
  • blk.30.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.30.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.30.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.30.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.30.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.30.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.30.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.30.post_attention_norm.weight
    F32
    [5120]
  • blk.31
  • blk.31.attn_k.bias
    F32
    [1024]
  • blk.31.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.31.attn_k_norm.weight
    F32
    [128]
  • blk.31.attn_norm.weight
    F32
    [5120]
  • blk.31.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.31.attn_q.bias
    F32
    [12288]
  • blk.31.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.31.attn_q_norm.weight
    F32
    [128]
  • blk.31.attn_v.bias
    F32
    [1024]
  • blk.31.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.31.exp_probs_b.bias
    F32
    [160]
  • blk.31.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.31.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.31.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.31.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.31.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.31.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.31.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.31.post_attention_norm.weight
    F32
    [5120]
  • blk.32
  • blk.32.attn_k.bias
    F32
    [1024]
  • blk.32.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.32.attn_k_norm.weight
    F32
    [128]
  • blk.32.attn_norm.weight
    F32
    [5120]
  • blk.32.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.32.attn_q.bias
    F32
    [12288]
  • blk.32.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.32.attn_q_norm.weight
    F32
    [128]
  • blk.32.attn_v.bias
    F32
    [1024]
  • blk.32.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.32.exp_probs_b.bias
    F32
    [160]
  • blk.32.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.32.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.32.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.32.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.32.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.32.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.32.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.32.post_attention_norm.weight
    F32
    [5120]
  • blk.33
  • blk.33.attn_k.bias
    F32
    [1024]
  • blk.33.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.33.attn_k_norm.weight
    F32
    [128]
  • blk.33.attn_norm.weight
    F32
    [5120]
  • blk.33.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.33.attn_q.bias
    F32
    [12288]
  • blk.33.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.33.attn_q_norm.weight
    F32
    [128]
  • blk.33.attn_v.bias
    F32
    [1024]
  • blk.33.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.33.exp_probs_b.bias
    F32
    [160]
  • blk.33.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.33.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.33.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.33.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.33.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.33.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.33.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.33.post_attention_norm.weight
    F32
    [5120]
  • blk.34
  • blk.34.attn_k.bias
    F32
    [1024]
  • blk.34.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.34.attn_k_norm.weight
    F32
    [128]
  • blk.34.attn_norm.weight
    F32
    [5120]
  • blk.34.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.34.attn_q.bias
    F32
    [12288]
  • blk.34.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.34.attn_q_norm.weight
    F32
    [128]
  • blk.34.attn_v.bias
    F32
    [1024]
  • blk.34.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.34.exp_probs_b.bias
    F32
    [160]
  • blk.34.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.34.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.34.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.34.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.34.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.34.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.34.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.34.post_attention_norm.weight
    F32
    [5120]
  • blk.35
  • blk.35.attn_k.bias
    F32
    [1024]
  • blk.35.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.35.attn_k_norm.weight
    F32
    [128]
  • blk.35.attn_norm.weight
    F32
    [5120]
  • blk.35.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.35.attn_q.bias
    F32
    [12288]
  • blk.35.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.35.attn_q_norm.weight
    F32
    [128]
  • blk.35.attn_v.bias
    F32
    [1024]
  • blk.35.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.35.exp_probs_b.bias
    F32
    [160]
  • blk.35.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.35.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.35.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.35.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.35.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.35.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.35.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.35.post_attention_norm.weight
    F32
    [5120]
  • blk.36
  • blk.36.attn_k.bias
    F32
    [1024]
  • blk.36.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.36.attn_k_norm.weight
    F32
    [128]
  • blk.36.attn_norm.weight
    F32
    [5120]
  • blk.36.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.36.attn_q.bias
    F32
    [12288]
  • blk.36.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.36.attn_q_norm.weight
    F32
    [128]
  • blk.36.attn_v.bias
    F32
    [1024]
  • blk.36.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.36.exp_probs_b.bias
    F32
    [160]
  • blk.36.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.36.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.36.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.36.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.36.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.36.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.36.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.36.post_attention_norm.weight
    F32
    [5120]
  • blk.37
  • blk.37.attn_k.bias
    F32
    [1024]
  • blk.37.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.37.attn_k_norm.weight
    F32
    [128]
  • blk.37.attn_norm.weight
    F32
    [5120]
  • blk.37.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.37.attn_q.bias
    F32
    [12288]
  • blk.37.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.37.attn_q_norm.weight
    F32
    [128]
  • blk.37.attn_v.bias
    F32
    [1024]
  • blk.37.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.37.exp_probs_b.bias
    F32
    [160]
  • blk.37.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.37.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.37.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.37.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.37.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.37.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.37.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.37.post_attention_norm.weight
    F32
    [5120]
  • blk.38
  • blk.38.attn_k.bias
    F32
    [1024]
  • blk.38.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.38.attn_k_norm.weight
    F32
    [128]
  • blk.38.attn_norm.weight
    F32
    [5120]
  • blk.38.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.38.attn_q.bias
    F32
    [12288]
  • blk.38.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.38.attn_q_norm.weight
    F32
    [128]
  • blk.38.attn_v.bias
    F32
    [1024]
  • blk.38.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.38.exp_probs_b.bias
    F32
    [160]
  • blk.38.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.38.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.38.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.38.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.38.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.38.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.38.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.38.post_attention_norm.weight
    F32
    [5120]
  • blk.39
  • blk.39.attn_k.bias
    F32
    [1024]
  • blk.39.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.39.attn_k_norm.weight
    F32
    [128]
  • blk.39.attn_norm.weight
    F32
    [5120]
  • blk.39.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.39.attn_q.bias
    F32
    [12288]
  • blk.39.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.39.attn_q_norm.weight
    F32
    [128]
  • blk.39.attn_v.bias
    F32
    [1024]
  • blk.39.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.39.exp_probs_b.bias
    F32
    [160]
  • blk.39.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.39.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.39.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.39.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.39.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.39.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.39.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.39.post_attention_norm.weight
    F32
    [5120]
  • blk.40
  • blk.40.attn_k.bias
    F32
    [1024]
  • blk.40.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.40.attn_k_norm.weight
    F32
    [128]
  • blk.40.attn_norm.weight
    F32
    [5120]
  • blk.40.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.40.attn_q.bias
    F32
    [12288]
  • blk.40.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.40.attn_q_norm.weight
    F32
    [128]
  • blk.40.attn_v.bias
    F32
    [1024]
  • blk.40.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.40.exp_probs_b.bias
    F32
    [160]
  • blk.40.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.40.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.40.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.40.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.40.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.40.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.40.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.40.post_attention_norm.weight
    F32
    [5120]
  • blk.41
  • blk.41.attn_k.bias
    F32
    [1024]
  • blk.41.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.41.attn_k_norm.weight
    F32
    [128]
  • blk.41.attn_norm.weight
    F32
    [5120]
  • blk.41.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.41.attn_q.bias
    F32
    [12288]
  • blk.41.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.41.attn_q_norm.weight
    F32
    [128]
  • blk.41.attn_v.bias
    F32
    [1024]
  • blk.41.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.41.exp_probs_b.bias
    F32
    [160]
  • blk.41.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.41.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.41.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.41.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.41.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.41.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.41.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.41.post_attention_norm.weight
    F32
    [5120]
  • blk.42
  • blk.42.attn_k.bias
    F32
    [1024]
  • blk.42.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.42.attn_k_norm.weight
    F32
    [128]
  • blk.42.attn_norm.weight
    F32
    [5120]
  • blk.42.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.42.attn_q.bias
    F32
    [12288]
  • blk.42.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.42.attn_q_norm.weight
    F32
    [128]
  • blk.42.attn_v.bias
    F32
    [1024]
  • blk.42.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.42.exp_probs_b.bias
    F32
    [160]
  • blk.42.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.42.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.42.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.42.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.42.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.42.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.42.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.42.post_attention_norm.weight
    F32
    [5120]
  • blk.43
  • blk.43.attn_k.bias
    F32
    [1024]
  • blk.43.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.43.attn_k_norm.weight
    F32
    [128]
  • blk.43.attn_norm.weight
    F32
    [5120]
  • blk.43.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.43.attn_q.bias
    F32
    [12288]
  • blk.43.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.43.attn_q_norm.weight
    F32
    [128]
  • blk.43.attn_v.bias
    F32
    [1024]
  • blk.43.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.43.exp_probs_b.bias
    F32
    [160]
  • blk.43.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.43.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.43.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.43.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.43.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.43.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.43.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.43.post_attention_norm.weight
    F32
    [5120]
  • blk.44
  • blk.44.attn_k.bias
    F32
    [1024]
  • blk.44.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.44.attn_k_norm.weight
    F32
    [128]
  • blk.44.attn_norm.weight
    F32
    [5120]
  • blk.44.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.44.attn_q.bias
    F32
    [12288]
  • blk.44.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.44.attn_q_norm.weight
    F32
    [128]
  • blk.44.attn_v.bias
    F32
    [1024]
  • blk.44.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.44.exp_probs_b.bias
    F32
    [160]
  • blk.44.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.44.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.44.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.44.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.44.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.44.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.44.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.44.post_attention_norm.weight
    F32
    [5120]
  • blk.45
  • blk.45.attn_k.bias
    F32
    [1024]
  • blk.45.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.45.attn_k_norm.weight
    F32
    [128]
  • blk.45.attn_norm.weight
    F32
    [5120]
  • blk.45.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.45.attn_q.bias
    F32
    [12288]
  • blk.45.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.45.attn_q_norm.weight
    F32
    [128]
  • blk.45.attn_v.bias
    F32
    [1024]
  • blk.45.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.45.exp_probs_b.bias
    F32
    [160]
  • blk.45.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.45.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.45.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.45.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.45.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.45.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.45.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.45.post_attention_norm.weight
    F32
    [5120]
  • blk.46
  • blk.46.attn_k.bias
    F32
    [1024]
  • blk.46.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.46.attn_k_norm.weight
    F32
    [128]
  • blk.46.attn_norm.weight
    F32
    [5120]
  • blk.46.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.46.attn_q.bias
    F32
    [12288]
  • blk.46.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.46.attn_q_norm.weight
    F32
    [128]
  • blk.46.attn_v.bias
    F32
    [1024]
  • blk.46.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.46.exp_probs_b.bias
    F32
    [160]
  • blk.46.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.46.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.46.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.46.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.46.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.46.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.46.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.46.post_attention_norm.weight
    F32
    [5120]
  • blk.47
  • blk.47.attn_k.bias
    F32
    [1024]
  • blk.47.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.47.attn_k_norm.weight
    F32
    [128]
  • blk.47.attn_norm.weight
    F32
    [5120]
  • blk.47.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.47.attn_q.bias
    F32
    [12288]
  • blk.47.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.47.attn_q_norm.weight
    F32
    [128]
  • blk.47.attn_v.bias
    F32
    [1024]
  • blk.47.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.47.exp_probs_b.bias
    F32
    [160]
  • blk.47.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.47.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.47.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.47.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.47.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.47.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.47.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.47.post_attention_norm.weight
    F32
    [5120]
  • blk.48
  • blk.48.attn_k.bias
    F32
    [1024]
  • blk.48.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.48.attn_k_norm.weight
    F32
    [128]
  • blk.48.attn_norm.weight
    F32
    [5120]
  • blk.48.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.48.attn_q.bias
    F32
    [12288]
  • blk.48.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.48.attn_q_norm.weight
    F32
    [128]
  • blk.48.attn_v.bias
    F32
    [1024]
  • blk.48.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.48.exp_probs_b.bias
    F32
    [160]
  • blk.48.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.48.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.48.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.48.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.48.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.48.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.48.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.48.post_attention_norm.weight
    F32
    [5120]
  • blk.49
  • blk.49.attn_k.bias
    F32
    [1024]
  • blk.49.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.49.attn_k_norm.weight
    F32
    [128]
  • blk.49.attn_norm.weight
    F32
    [5120]
  • blk.49.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.49.attn_q.bias
    F32
    [12288]
  • blk.49.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.49.attn_q_norm.weight
    F32
    [128]
  • blk.49.attn_v.bias
    F32
    [1024]
  • blk.49.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.49.exp_probs_b.bias
    F32
    [160]
  • blk.49.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.49.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.49.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.49.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.49.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.49.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.49.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.49.post_attention_norm.weight
    F32
    [5120]
  • blk.50
  • blk.50.attn_k.bias
    F32
    [1024]
  • blk.50.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.50.attn_k_norm.weight
    F32
    [128]
  • blk.50.attn_norm.weight
    F32
    [5120]
  • blk.50.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.50.attn_q.bias
    F32
    [12288]
  • blk.50.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.50.attn_q_norm.weight
    F32
    [128]
  • blk.50.attn_v.bias
    F32
    [1024]
  • blk.50.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.50.exp_probs_b.bias
    F32
    [160]
  • blk.50.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.50.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.50.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.50.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.50.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.50.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.50.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.50.post_attention_norm.weight
    F32
    [5120]
  • blk.51
  • blk.51.attn_k.bias
    F32
    [1024]
  • blk.51.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.51.attn_k_norm.weight
    F32
    [128]
  • blk.51.attn_norm.weight
    F32
    [5120]
  • blk.51.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.51.attn_q.bias
    F32
    [12288]
  • blk.51.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.51.attn_q_norm.weight
    F32
    [128]
  • blk.51.attn_v.bias
    F32
    [1024]
  • blk.51.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.51.exp_probs_b.bias
    F32
    [160]
  • blk.51.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.51.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.51.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.51.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.51.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.51.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.51.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.51.post_attention_norm.weight
    F32
    [5120]
  • blk.52
  • blk.52.attn_k.bias
    F32
    [1024]
  • blk.52.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.52.attn_k_norm.weight
    F32
    [128]
  • blk.52.attn_norm.weight
    F32
    [5120]
  • blk.52.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.52.attn_q.bias
    F32
    [12288]
  • blk.52.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.52.attn_q_norm.weight
    F32
    [128]
  • blk.52.attn_v.bias
    F32
    [1024]
  • blk.52.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.52.exp_probs_b.bias
    F32
    [160]
  • blk.52.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.52.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.52.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.52.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.52.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.52.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.52.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.52.post_attention_norm.weight
    F32
    [5120]
  • blk.53
  • blk.53.attn_k.bias
    F32
    [1024]
  • blk.53.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.53.attn_k_norm.weight
    F32
    [128]
  • blk.53.attn_norm.weight
    F32
    [5120]
  • blk.53.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.53.attn_q.bias
    F32
    [12288]
  • blk.53.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.53.attn_q_norm.weight
    F32
    [128]
  • blk.53.attn_v.bias
    F32
    [1024]
  • blk.53.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.53.exp_probs_b.bias
    F32
    [160]
  • blk.53.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.53.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.53.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.53.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.53.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.53.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.53.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.53.post_attention_norm.weight
    F32
    [5120]
  • blk.54
  • blk.54.attn_k.bias
    F32
    [1024]
  • blk.54.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.54.attn_k_norm.weight
    F32
    [128]
  • blk.54.attn_norm.weight
    F32
    [5120]
  • blk.54.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.54.attn_q.bias
    F32
    [12288]
  • blk.54.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.54.attn_q_norm.weight
    F32
    [128]
  • blk.54.attn_v.bias
    F32
    [1024]
  • blk.54.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.54.exp_probs_b.bias
    F32
    [160]
  • blk.54.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.54.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.54.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.54.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.54.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.54.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.54.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.54.post_attention_norm.weight
    F32
    [5120]
  • blk.55
  • blk.55.attn_k.bias
    F32
    [1024]
  • blk.55.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.55.attn_k_norm.weight
    F32
    [128]
  • blk.55.attn_norm.weight
    F32
    [5120]
  • blk.55.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.55.attn_q.bias
    F32
    [12288]
  • blk.55.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.55.attn_q_norm.weight
    F32
    [128]
  • blk.55.attn_v.bias
    F32
    [1024]
  • blk.55.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.55.exp_probs_b.bias
    F32
    [160]
  • blk.55.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.55.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.55.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.55.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.55.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.55.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.55.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.55.post_attention_norm.weight
    F32
    [5120]
  • blk.56
  • blk.56.attn_k.bias
    F32
    [1024]
  • blk.56.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.56.attn_k_norm.weight
    F32
    [128]
  • blk.56.attn_norm.weight
    F32
    [5120]
  • blk.56.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.56.attn_q.bias
    F32
    [12288]
  • blk.56.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.56.attn_q_norm.weight
    F32
    [128]
  • blk.56.attn_v.bias
    F32
    [1024]
  • blk.56.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.56.exp_probs_b.bias
    F32
    [160]
  • blk.56.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.56.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.56.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.56.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.56.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.56.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.56.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.56.post_attention_norm.weight
    F32
    [5120]
  • blk.57
  • blk.57.attn_k.bias
    F32
    [1024]
  • blk.57.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.57.attn_k_norm.weight
    F32
    [128]
  • blk.57.attn_norm.weight
    F32
    [5120]
  • blk.57.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.57.attn_q.bias
    F32
    [12288]
  • blk.57.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.57.attn_q_norm.weight
    F32
    [128]
  • blk.57.attn_v.bias
    F32
    [1024]
  • blk.57.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.57.exp_probs_b.bias
    F32
    [160]
  • blk.57.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.57.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.57.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.57.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.57.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.57.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.57.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.57.post_attention_norm.weight
    F32
    [5120]
  • blk.58
  • blk.58.attn_k.bias
    F32
    [1024]
  • blk.58.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.58.attn_k_norm.weight
    F32
    [128]
  • blk.58.attn_norm.weight
    F32
    [5120]
  • blk.58.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.58.attn_q.bias
    F32
    [12288]
  • blk.58.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.58.attn_q_norm.weight
    F32
    [128]
  • blk.58.attn_v.bias
    F32
    [1024]
  • blk.58.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.58.exp_probs_b.bias
    F32
    [160]
  • blk.58.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.58.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.58.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.58.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.58.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.58.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.58.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.58.post_attention_norm.weight
    F32
    [5120]
  • blk.59
  • blk.59.attn_k.bias
    F32
    [1024]
  • blk.59.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.59.attn_k_norm.weight
    F32
    [128]
  • blk.59.attn_norm.weight
    F32
    [5120]
  • blk.59.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.59.attn_q.bias
    F32
    [12288]
  • blk.59.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.59.attn_q_norm.weight
    F32
    [128]
  • blk.59.attn_v.bias
    F32
    [1024]
  • blk.59.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.59.exp_probs_b.bias
    F32
    [160]
  • blk.59.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.59.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.59.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.59.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.59.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.59.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.59.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.59.post_attention_norm.weight
    F32
    [5120]
  • blk.60
  • blk.60.attn_k.bias
    F32
    [1024]
  • blk.60.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.60.attn_k_norm.weight
    F32
    [128]
  • blk.60.attn_norm.weight
    F32
    [5120]
  • blk.60.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.60.attn_q.bias
    F32
    [12288]
  • blk.60.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.60.attn_q_norm.weight
    F32
    [128]
  • blk.60.attn_v.bias
    F32
    [1024]
  • blk.60.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.60.exp_probs_b.bias
    F32
    [160]
  • blk.60.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.60.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.60.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.60.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.60.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.60.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.60.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.60.post_attention_norm.weight
    F32
    [5120]
  • blk.61
  • blk.61.attn_k.bias
    F32
    [1024]
  • blk.61.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.61.attn_k_norm.weight
    F32
    [128]
  • blk.61.attn_norm.weight
    F32
    [5120]
  • blk.61.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.61.attn_q.bias
    F32
    [12288]
  • blk.61.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.61.attn_q_norm.weight
    F32
    [128]
  • blk.61.attn_v.bias
    F32
    [1024]
  • blk.61.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.61.exp_probs_b.bias
    F32
    [160]
  • blk.61.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.61.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.61.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.61.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.61.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.61.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.61.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.61.post_attention_norm.weight
    F32
    [5120]
  • blk.62
  • blk.62.attn_k.bias
    F32
    [1024]
  • blk.62.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.62.attn_k_norm.weight
    F32
    [128]
  • blk.62.attn_norm.weight
    F32
    [5120]
  • blk.62.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.62.attn_q.bias
    F32
    [12288]
  • blk.62.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.62.attn_q_norm.weight
    F32
    [128]
  • blk.62.attn_v.bias
    F32
    [1024]
  • blk.62.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.62.exp_probs_b.bias
    F32
    [160]
  • blk.62.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.62.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.62.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.62.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.62.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.62.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.62.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.62.post_attention_norm.weight
    F32
    [5120]
  • blk.63
  • blk.63.attn_k.bias
    F32
    [1024]
  • blk.63.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.63.attn_k_norm.weight
    F32
    [128]
  • blk.63.attn_norm.weight
    F32
    [5120]
  • blk.63.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.63.attn_q.bias
    F32
    [12288]
  • blk.63.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.63.attn_q_norm.weight
    F32
    [128]
  • blk.63.attn_v.bias
    F32
    [1024]
  • blk.63.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.63.exp_probs_b.bias
    F32
    [160]
  • blk.63.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.63.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.63.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.63.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.63.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.63.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.63.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.63.post_attention_norm.weight
    F32
    [5120]
  • blk.64
  • blk.64.attn_k.bias
    F32
    [1024]
  • blk.64.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.64.attn_k_norm.weight
    F32
    [128]
  • blk.64.attn_norm.weight
    F32
    [5120]
  • blk.64.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.64.attn_q.bias
    F32
    [12288]
  • blk.64.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.64.attn_q_norm.weight
    F32
    [128]
  • blk.64.attn_v.bias
    F32
    [1024]
  • blk.64.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.64.exp_probs_b.bias
    F32
    [160]
  • blk.64.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.64.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.64.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.64.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.64.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.64.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.64.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.64.post_attention_norm.weight
    F32
    [5120]
  • blk.65
  • blk.65.attn_k.bias
    F32
    [1024]
  • blk.65.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.65.attn_k_norm.weight
    F32
    [128]
  • blk.65.attn_norm.weight
    F32
    [5120]
  • blk.65.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.65.attn_q.bias
    F32
    [12288]
  • blk.65.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.65.attn_q_norm.weight
    F32
    [128]
  • blk.65.attn_v.bias
    F32
    [1024]
  • blk.65.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.65.exp_probs_b.bias
    F32
    [160]
  • blk.65.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.65.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.65.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.65.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.65.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.65.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.65.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.65.post_attention_norm.weight
    F32
    [5120]
  • blk.66
  • blk.66.attn_k.bias
    F32
    [1024]
  • blk.66.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.66.attn_k_norm.weight
    F32
    [128]
  • blk.66.attn_norm.weight
    F32
    [5120]
  • blk.66.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.66.attn_q.bias
    F32
    [12288]
  • blk.66.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.66.attn_q_norm.weight
    F32
    [128]
  • blk.66.attn_v.bias
    F32
    [1024]
  • blk.66.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.66.exp_probs_b.bias
    F32
    [160]
  • blk.66.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.66.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.66.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.66.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.66.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.66.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.66.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.66.post_attention_norm.weight
    F32
    [5120]
  • blk.67
  • blk.67.attn_k.bias
    F32
    [1024]
  • blk.67.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.67.attn_k_norm.weight
    F32
    [128]
  • blk.67.attn_norm.weight
    F32
    [5120]
  • blk.67.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.67.attn_q.bias
    F32
    [12288]
  • blk.67.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.67.attn_q_norm.weight
    F32
    [128]
  • blk.67.attn_v.bias
    F32
    [1024]
  • blk.67.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.67.exp_probs_b.bias
    F32
    [160]
  • blk.67.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.67.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.67.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.67.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.67.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.67.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.67.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.67.post_attention_norm.weight
    F32
    [5120]
  • blk.68
  • blk.68.attn_k.bias
    F32
    [1024]
  • blk.68.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.68.attn_k_norm.weight
    F32
    [128]
  • blk.68.attn_norm.weight
    F32
    [5120]
  • blk.68.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.68.attn_q.bias
    F32
    [12288]
  • blk.68.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.68.attn_q_norm.weight
    F32
    [128]
  • blk.68.attn_v.bias
    F32
    [1024]
  • blk.68.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.68.exp_probs_b.bias
    F32
    [160]
  • blk.68.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.68.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.68.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.68.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.68.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.68.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.68.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.68.post_attention_norm.weight
    F32
    [5120]
  • blk.69
  • blk.69.attn_k.bias
    F32
    [1024]
  • blk.69.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.69.attn_k_norm.weight
    F32
    [128]
  • blk.69.attn_norm.weight
    F32
    [5120]
  • blk.69.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.69.attn_q.bias
    F32
    [12288]
  • blk.69.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.69.attn_q_norm.weight
    F32
    [128]
  • blk.69.attn_v.bias
    F32
    [1024]
  • blk.69.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.69.exp_probs_b.bias
    F32
    [160]
  • blk.69.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.69.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.69.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.69.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.69.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.69.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.69.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.69.post_attention_norm.weight
    F32
    [5120]
  • blk.70
  • blk.70.attn_k.bias
    F32
    [1024]
  • blk.70.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.70.attn_k_norm.weight
    F32
    [128]
  • blk.70.attn_norm.weight
    F32
    [5120]
  • blk.70.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.70.attn_q.bias
    F32
    [12288]
  • blk.70.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.70.attn_q_norm.weight
    F32
    [128]
  • blk.70.attn_v.bias
    F32
    [1024]
  • blk.70.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.70.exp_probs_b.bias
    F32
    [160]
  • blk.70.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.70.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.70.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.70.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.70.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.70.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.70.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.70.post_attention_norm.weight
    F32
    [5120]
  • blk.71
  • blk.71.attn_k.bias
    F32
    [1024]
  • blk.71.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.71.attn_k_norm.weight
    F32
    [128]
  • blk.71.attn_norm.weight
    F32
    [5120]
  • blk.71.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.71.attn_q.bias
    F32
    [12288]
  • blk.71.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.71.attn_q_norm.weight
    F32
    [128]
  • blk.71.attn_v.bias
    F32
    [1024]
  • blk.71.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.71.exp_probs_b.bias
    F32
    [160]
  • blk.71.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.71.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.71.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.71.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.71.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.71.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.71.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.71.post_attention_norm.weight
    F32
    [5120]
  • blk.72
  • blk.72.attn_k.bias
    F32
    [1024]
  • blk.72.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.72.attn_k_norm.weight
    F32
    [128]
  • blk.72.attn_norm.weight
    F32
    [5120]
  • blk.72.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.72.attn_q.bias
    F32
    [12288]
  • blk.72.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.72.attn_q_norm.weight
    F32
    [128]
  • blk.72.attn_v.bias
    F32
    [1024]
  • blk.72.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.72.exp_probs_b.bias
    F32
    [160]
  • blk.72.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.72.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.72.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.72.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.72.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.72.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.72.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.72.post_attention_norm.weight
    F32
    [5120]
  • blk.73
  • blk.73.attn_k.bias
    F32
    [1024]
  • blk.73.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.73.attn_k_norm.weight
    F32
    [128]
  • blk.73.attn_norm.weight
    F32
    [5120]
  • blk.73.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.73.attn_q.bias
    F32
    [12288]
  • blk.73.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.73.attn_q_norm.weight
    F32
    [128]
  • blk.73.attn_v.bias
    F32
    [1024]
  • blk.73.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.73.exp_probs_b.bias
    F32
    [160]
  • blk.73.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.73.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.73.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.73.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.73.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.73.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.73.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.73.post_attention_norm.weight
    F32
    [5120]
  • blk.74
  • blk.74.attn_k.bias
    F32
    [1024]
  • blk.74.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.74.attn_k_norm.weight
    F32
    [128]
  • blk.74.attn_norm.weight
    F32
    [5120]
  • blk.74.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.74.attn_q.bias
    F32
    [12288]
  • blk.74.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.74.attn_q_norm.weight
    F32
    [128]
  • blk.74.attn_v.bias
    F32
    [1024]
  • blk.74.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.74.exp_probs_b.bias
    F32
    [160]
  • blk.74.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.74.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.74.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.74.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.74.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.74.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.74.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.74.post_attention_norm.weight
    F32
    [5120]
  • blk.75
  • blk.75.attn_k.bias
    F32
    [1024]
  • blk.75.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.75.attn_k_norm.weight
    F32
    [128]
  • blk.75.attn_norm.weight
    F32
    [5120]
  • blk.75.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.75.attn_q.bias
    F32
    [12288]
  • blk.75.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.75.attn_q_norm.weight
    F32
    [128]
  • blk.75.attn_v.bias
    F32
    [1024]
  • blk.75.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.75.exp_probs_b.bias
    F32
    [160]
  • blk.75.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.75.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.75.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.75.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.75.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.75.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.75.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.75.post_attention_norm.weight
    F32
    [5120]
  • blk.76
  • blk.76.attn_k.bias
    F32
    [1024]
  • blk.76.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.76.attn_k_norm.weight
    F32
    [128]
  • blk.76.attn_norm.weight
    F32
    [5120]
  • blk.76.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.76.attn_q.bias
    F32
    [12288]
  • blk.76.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.76.attn_q_norm.weight
    F32
    [128]
  • blk.76.attn_v.bias
    F32
    [1024]
  • blk.76.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.76.exp_probs_b.bias
    F32
    [160]
  • blk.76.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.76.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.76.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.76.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.76.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.76.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.76.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.76.post_attention_norm.weight
    F32
    [5120]
  • blk.77
  • blk.77.attn_k.bias
    F32
    [1024]
  • blk.77.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.77.attn_k_norm.weight
    F32
    [128]
  • blk.77.attn_norm.weight
    F32
    [5120]
  • blk.77.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.77.attn_q.bias
    F32
    [12288]
  • blk.77.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.77.attn_q_norm.weight
    F32
    [128]
  • blk.77.attn_v.bias
    F32
    [1024]
  • blk.77.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.77.exp_probs_b.bias
    F32
    [160]
  • blk.77.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.77.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.77.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.77.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.77.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.77.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.77.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.77.post_attention_norm.weight
    F32
    [5120]
  • blk.78
  • blk.78.attn_k.bias
    F32
    [1024]
  • blk.78.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.78.attn_k_norm.weight
    F32
    [128]
  • blk.78.attn_norm.weight
    F32
    [5120]
  • blk.78.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.78.attn_q.bias
    F32
    [12288]
  • blk.78.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.78.attn_q_norm.weight
    F32
    [128]
  • blk.78.attn_v.bias
    F32
    [1024]
  • blk.78.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.78.exp_probs_b.bias
    F32
    [160]
  • blk.78.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.78.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.78.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.78.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.78.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.78.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.78.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.78.post_attention_norm.weight
    F32
    [5120]
  • blk.79
  • blk.79.attn_k.bias
    F32
    [1024]
  • blk.79.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.79.attn_k_norm.weight
    F32
    [128]
  • blk.79.attn_norm.weight
    F32
    [5120]
  • blk.79.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.79.attn_q.bias
    F32
    [12288]
  • blk.79.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.79.attn_q_norm.weight
    F32
    [128]
  • blk.79.attn_v.bias
    F32
    [1024]
  • blk.79.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.79.exp_probs_b.bias
    F32
    [160]
  • blk.79.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.79.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.79.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.79.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.79.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.79.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.79.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.79.post_attention_norm.weight
    F32
    [5120]
  • blk.80
  • blk.80.attn_k.bias
    F32
    [1024]
  • blk.80.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.80.attn_k_norm.weight
    F32
    [128]
  • blk.80.attn_norm.weight
    F32
    [5120]
  • blk.80.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.80.attn_q.bias
    F32
    [12288]
  • blk.80.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.80.attn_q_norm.weight
    F32
    [128]
  • blk.80.attn_v.bias
    F32
    [1024]
  • blk.80.attn_v.weight
    Q4_K
    [5120, 1024]
  • blk.80.exp_probs_b.bias
    F32
    [160]
  • blk.80.ffn_down_exps.weight
    Q4_K
    [1536, 5120, 160]
  • blk.80.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.80.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.80.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.80.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.80.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.80.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.80.post_attention_norm.weight
    F32
    [5120]
  • blk.81
  • blk.81.attn_k.bias
    F32
    [1024]
  • blk.81.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.81.attn_k_norm.weight
    F32
    [128]
  • blk.81.attn_norm.weight
    F32
    [5120]
  • blk.81.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.81.attn_q.bias
    F32
    [12288]
  • blk.81.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.81.attn_q_norm.weight
    F32
    [128]
  • blk.81.attn_v.bias
    F32
    [1024]
  • blk.81.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.81.exp_probs_b.bias
    F32
    [160]
  • blk.81.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.81.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.81.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.81.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.81.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.81.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.81.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.81.post_attention_norm.weight
    F32
    [5120]
  • blk.82
  • blk.82.attn_k.bias
    F32
    [1024]
  • blk.82.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.82.attn_k_norm.weight
    F32
    [128]
  • blk.82.attn_norm.weight
    F32
    [5120]
  • blk.82.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.82.attn_q.bias
    F32
    [12288]
  • blk.82.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.82.attn_q_norm.weight
    F32
    [128]
  • blk.82.attn_v.bias
    F32
    [1024]
  • blk.82.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.82.exp_probs_b.bias
    F32
    [160]
  • blk.82.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.82.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.82.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.82.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.82.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.82.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.82.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.82.post_attention_norm.weight
    F32
    [5120]
  • blk.83
  • blk.83.attn_k.bias
    F32
    [1024]
  • blk.83.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.83.attn_k_norm.weight
    F32
    [128]
  • blk.83.attn_norm.weight
    F32
    [5120]
  • blk.83.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.83.attn_q.bias
    F32
    [12288]
  • blk.83.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.83.attn_q_norm.weight
    F32
    [128]
  • blk.83.attn_v.bias
    F32
    [1024]
  • blk.83.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.83.exp_probs_b.bias
    F32
    [160]
  • blk.83.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.83.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.83.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.83.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.83.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.83.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.83.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.83.post_attention_norm.weight
    F32
    [5120]
  • blk.84
  • blk.84.attn_k.bias
    F32
    [1024]
  • blk.84.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.84.attn_k_norm.weight
    F32
    [128]
  • blk.84.attn_norm.weight
    F32
    [5120]
  • blk.84.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.84.attn_q.bias
    F32
    [12288]
  • blk.84.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.84.attn_q_norm.weight
    F32
    [128]
  • blk.84.attn_v.bias
    F32
    [1024]
  • blk.84.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.84.exp_probs_b.bias
    F32
    [160]
  • blk.84.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.84.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.84.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.84.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.84.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.84.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.84.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.84.post_attention_norm.weight
    F32
    [5120]
  • blk.85
  • blk.85.attn_k.bias
    F32
    [1024]
  • blk.85.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.85.attn_k_norm.weight
    F32
    [128]
  • blk.85.attn_norm.weight
    F32
    [5120]
  • blk.85.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.85.attn_q.bias
    F32
    [12288]
  • blk.85.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.85.attn_q_norm.weight
    F32
    [128]
  • blk.85.attn_v.bias
    F32
    [1024]
  • blk.85.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.85.exp_probs_b.bias
    F32
    [160]
  • blk.85.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.85.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.85.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.85.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.85.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.85.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.85.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.85.post_attention_norm.weight
    F32
    [5120]
  • blk.86
  • blk.86.attn_k.bias
    F32
    [1024]
  • blk.86.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.86.attn_k_norm.weight
    F32
    [128]
  • blk.86.attn_norm.weight
    F32
    [5120]
  • blk.86.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.86.attn_q.bias
    F32
    [12288]
  • blk.86.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.86.attn_q_norm.weight
    F32
    [128]
  • blk.86.attn_v.bias
    F32
    [1024]
  • blk.86.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.86.exp_probs_b.bias
    F32
    [160]
  • blk.86.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.86.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.86.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.86.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.86.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.86.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.86.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.86.post_attention_norm.weight
    F32
    [5120]
  • blk.87
  • blk.87.attn_k.bias
    F32
    [1024]
  • blk.87.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.87.attn_k_norm.weight
    F32
    [128]
  • blk.87.attn_norm.weight
    F32
    [5120]
  • blk.87.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.87.attn_q.bias
    F32
    [12288]
  • blk.87.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.87.attn_q_norm.weight
    F32
    [128]
  • blk.87.attn_v.bias
    F32
    [1024]
  • blk.87.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.87.exp_probs_b.bias
    F32
    [160]
  • blk.87.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.87.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.87.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.87.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.87.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.87.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.87.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.87.post_attention_norm.weight
    F32
    [5120]
  • blk.88
  • blk.88.attn_k.bias
    F32
    [1024]
  • blk.88.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.88.attn_k_norm.weight
    F32
    [128]
  • blk.88.attn_norm.weight
    F32
    [5120]
  • blk.88.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.88.attn_q.bias
    F32
    [12288]
  • blk.88.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.88.attn_q_norm.weight
    F32
    [128]
  • blk.88.attn_v.bias
    F32
    [1024]
  • blk.88.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.88.exp_probs_b.bias
    F32
    [160]
  • blk.88.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.88.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.88.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.88.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.88.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.88.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.88.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.88.post_attention_norm.weight
    F32
    [5120]
  • blk.89
  • blk.89.attn_k.bias
    F32
    [1024]
  • blk.89.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.89.attn_k_norm.weight
    F32
    [128]
  • blk.89.attn_norm.weight
    F32
    [5120]
  • blk.89.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.89.attn_q.bias
    F32
    [12288]
  • blk.89.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.89.attn_q_norm.weight
    F32
    [128]
  • blk.89.attn_v.bias
    F32
    [1024]
  • blk.89.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.89.exp_probs_b.bias
    F32
    [160]
  • blk.89.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.89.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.89.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.89.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.89.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.89.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.89.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.89.post_attention_norm.weight
    F32
    [5120]
  • blk.90
  • blk.90.attn_k.bias
    F32
    [1024]
  • blk.90.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.90.attn_k_norm.weight
    F32
    [128]
  • blk.90.attn_norm.weight
    F32
    [5120]
  • blk.90.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.90.attn_q.bias
    F32
    [12288]
  • blk.90.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.90.attn_q_norm.weight
    F32
    [128]
  • blk.90.attn_v.bias
    F32
    [1024]
  • blk.90.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.90.exp_probs_b.bias
    F32
    [160]
  • blk.90.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.90.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.90.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.90.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.90.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.90.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.90.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.90.post_attention_norm.weight
    F32
    [5120]
  • blk.91
  • blk.91.attn_k.bias
    F32
    [1024]
  • blk.91.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.91.attn_k_norm.weight
    F32
    [128]
  • blk.91.attn_norm.weight
    F32
    [5120]
  • blk.91.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.91.attn_q.bias
    F32
    [12288]
  • blk.91.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.91.attn_q_norm.weight
    F32
    [128]
  • blk.91.attn_v.bias
    F32
    [1024]
  • blk.91.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.91.exp_probs_b.bias
    F32
    [160]
  • blk.91.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.91.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.91.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.91.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.91.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.91.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.91.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.91.post_attention_norm.weight
    F32
    [5120]
  • blk.92
  • blk.92.attn_k.bias
    F32
    [1024]
  • blk.92.attn_k.weight
    Q4_K
    [5120, 1024]
  • blk.92.attn_k_norm.weight
    F32
    [128]
  • blk.92.attn_norm.weight
    F32
    [5120]
  • blk.92.attn_output.weight
    Q4_K
    [12288, 5120]
  • blk.92.attn_q.bias
    F32
    [12288]
  • blk.92.attn_q.weight
    Q4_K
    [5120, 12288]
  • blk.92.attn_q_norm.weight
    F32
    [128]
  • blk.92.attn_v.bias
    F32
    [1024]
  • blk.92.attn_v.weight
    Q6_K
    [5120, 1024]
  • blk.92.exp_probs_b.bias
    F32
    [160]
  • blk.92.ffn_down_exps.weight
    Q6_K
    [1536, 5120, 160]
  • blk.92.ffn_down_shexp.weight
    Q6_K
    [1536, 5120]
  • blk.92.ffn_gate_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.92.ffn_gate_inp.weight
    F32
    [5120, 160]
  • blk.92.ffn_gate_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.92.ffn_up_exps.weight
    Q4_K
    [5120, 1536, 160]
  • blk.92.ffn_up_shexp.weight
    Q5_K
    [5120, 1536]
  • blk.92.nextn.eh_proj.weight
    Q4_K
    [10240, 5120]
  • blk.92.nextn.enorm.weight
    F32
    [5120]
  • blk.92.nextn.hnorm.weight
    F32
    [5120]
  • blk.92.nextn.shared_head_norm.weight
    F32
    [5120]
  • blk.92.post_attention_norm.weight
    F32
    [5120]
  • output.weight
    Q6_K
    [5120, 151552]
  • output_norm.weight
    F32
    [5120]