24 2 weeks ago

tools thinking
509a10dae02b · 4.0GB
    Metadata
  • general.architecture
    granitehybrid
  • general.file_type
    Q4_K_M
  • granitehybrid.attention.head_count
    12
  • granitehybrid.attention.head_count_kv
    [0, 0, 0, 0, 0, ...]
  • granitehybrid.attention.layer_norm_rms_epsilon
    1e-05
  • granitehybrid.attention.scale
    0.0078125
  • granitehybrid.block_count
    40
  • granitehybrid.context_length
    1048576
  • granitehybrid.embedding_length
    1536
  • granitehybrid.embedding_scale
    12
  • granitehybrid.expert_count
    62
  • granitehybrid.expert_shared_feed_forward_length
    1024
  • granitehybrid.expert_used_count
    6
  • granitehybrid.feed_forward_length
    512
  • granitehybrid.logit_scale
    6
  • granitehybrid.residual_scale
    0.22
  • granitehybrid.rope.dimension_count
    128
  • granitehybrid.rope.freq_base
    10000
  • granitehybrid.rope.scaling.finetuned
    false
  • granitehybrid.ssm.conv_kernel
    4
  • granitehybrid.ssm.group_count
    1
  • granitehybrid.ssm.inner_size
    3072
  • granitehybrid.ssm.state_size
    128
  • granitehybrid.ssm.time_step_rank
    48
  • granitehybrid.vocab_size
    49160
  • tokenizer.ggml.bos_token_id
    0
  • tokenizer.ggml.eos_token_id
    0
  • tokenizer.ggml.merges
    [Ġ Ġ, ĠĠ ĠĠ, ĠĠĠĠ ĠĠĠĠ, ĠĠ Ġ, e r, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    0
  • tokenizer.ggml.pre
    refact
  • tokenizer.ggml.token_type
    [3, 3, 3, 3, 3, ...]
  • tokenizer.ggml.tokens
    [<|end_of_text|>, <fim_prefix>, <fim_middle>, <fim_suffix>, <fim_pad>, ...]
  • tokenizer.ggml.unknown_token_id
    0
  • Tensor
  • token_embd.weight
    Q6_K
    [1536, 49160]
  • blk.0
  • blk.0.attn_norm.weight
    F32
    [1536]
  • blk.0.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.0.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.0.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.0.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.0.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.0.ffn_norm.weight
    F32
    [1536]
  • blk.0.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.0.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.0.ssm_a
    F32
    [1, 48]
  • blk.0.ssm_conv1d.bias
    F32
    [3328]
  • blk.0.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.0.ssm_d
    F32
    [1, 48]
  • blk.0.ssm_dt.bias
    F32
    [48]
  • blk.0.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.0.ssm_norm.weight
    F32
    [3072]
  • blk.0.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.1
  • blk.1.attn_norm.weight
    F32
    [1536]
  • blk.1.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.1.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.1.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.1.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.1.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.1.ffn_norm.weight
    F32
    [1536]
  • blk.1.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.1.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.1.ssm_a
    F32
    [1, 48]
  • blk.1.ssm_conv1d.bias
    F32
    [3328]
  • blk.1.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.1.ssm_d
    F32
    [1, 48]
  • blk.1.ssm_dt.bias
    F32
    [48]
  • blk.1.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.1.ssm_norm.weight
    F32
    [3072]
  • blk.1.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.2
  • blk.2.attn_norm.weight
    F32
    [1536]
  • blk.2.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.2.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.2.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.2.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.2.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.2.ffn_norm.weight
    F32
    [1536]
  • blk.2.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.2.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.2.ssm_a
    F32
    [1, 48]
  • blk.2.ssm_conv1d.bias
    F32
    [3328]
  • blk.2.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.2.ssm_d
    F32
    [1, 48]
  • blk.2.ssm_dt.bias
    F32
    [48]
  • blk.2.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.2.ssm_norm.weight
    F32
    [3072]
  • blk.2.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.3
  • blk.3.attn_norm.weight
    F32
    [1536]
  • blk.3.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.3.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.3.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.3.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.3.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.3.ffn_norm.weight
    F32
    [1536]
  • blk.3.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.3.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.3.ssm_a
    F32
    [1, 48]
  • blk.3.ssm_conv1d.bias
    F32
    [3328]
  • blk.3.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.3.ssm_d
    F32
    [1, 48]
  • blk.3.ssm_dt.bias
    F32
    [48]
  • blk.3.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.3.ssm_norm.weight
    F32
    [3072]
  • blk.3.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.4
  • blk.4.attn_norm.weight
    F32
    [1536]
  • blk.4.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.4.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.4.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.4.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.4.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.4.ffn_norm.weight
    F32
    [1536]
  • blk.4.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.4.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.4.ssm_a
    F32
    [1, 48]
  • blk.4.ssm_conv1d.bias
    F32
    [3328]
  • blk.4.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.4.ssm_d
    F32
    [1, 48]
  • blk.4.ssm_dt.bias
    F32
    [48]
  • blk.4.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.4.ssm_norm.weight
    F32
    [3072]
  • blk.4.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.5
  • blk.5.attn_k.weight
    Q4_K
    [1536, 512]
  • blk.5.attn_norm.weight
    F32
    [1536]
  • blk.5.attn_output.weight
    Q4_K
    [1536, 1536]
  • blk.5.attn_q.weight
    Q4_K
    [1536, 1536]
  • blk.5.attn_v.weight
    Q4_K
    [1536, 512]
  • blk.5.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.5.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.5.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.5.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.5.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.5.ffn_norm.weight
    F32
    [1536]
  • blk.5.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.5.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.6
  • blk.6.attn_norm.weight
    F32
    [1536]
  • blk.6.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.6.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.6.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.6.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.6.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.6.ffn_norm.weight
    F32
    [1536]
  • blk.6.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.6.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.6.ssm_a
    F32
    [1, 48]
  • blk.6.ssm_conv1d.bias
    F32
    [3328]
  • blk.6.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.6.ssm_d
    F32
    [1, 48]
  • blk.6.ssm_dt.bias
    F32
    [48]
  • blk.6.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.6.ssm_norm.weight
    F32
    [3072]
  • blk.6.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.7
  • blk.7.attn_norm.weight
    F32
    [1536]
  • blk.7.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.7.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.7.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.7.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.7.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.7.ffn_norm.weight
    F32
    [1536]
  • blk.7.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.7.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.7.ssm_a
    F32
    [1, 48]
  • blk.7.ssm_conv1d.bias
    F32
    [3328]
  • blk.7.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.7.ssm_d
    F32
    [1, 48]
  • blk.7.ssm_dt.bias
    F32
    [48]
  • blk.7.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.7.ssm_norm.weight
    F32
    [3072]
  • blk.7.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.8
  • blk.8.attn_norm.weight
    F32
    [1536]
  • blk.8.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.8.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.8.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.8.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.8.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.8.ffn_norm.weight
    F32
    [1536]
  • blk.8.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.8.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.8.ssm_a
    F32
    [1, 48]
  • blk.8.ssm_conv1d.bias
    F32
    [3328]
  • blk.8.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.8.ssm_d
    F32
    [1, 48]
  • blk.8.ssm_dt.bias
    F32
    [48]
  • blk.8.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.8.ssm_norm.weight
    F32
    [3072]
  • blk.8.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.9
  • blk.9.attn_norm.weight
    F32
    [1536]
  • blk.9.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.9.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.9.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.9.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.9.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.9.ffn_norm.weight
    F32
    [1536]
  • blk.9.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.9.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.9.ssm_a
    F32
    [1, 48]
  • blk.9.ssm_conv1d.bias
    F32
    [3328]
  • blk.9.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.9.ssm_d
    F32
    [1, 48]
  • blk.9.ssm_dt.bias
    F32
    [48]
  • blk.9.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.9.ssm_norm.weight
    F32
    [3072]
  • blk.9.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.10
  • blk.10.attn_norm.weight
    F32
    [1536]
  • blk.10.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.10.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.10.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.10.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.10.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.10.ffn_norm.weight
    F32
    [1536]
  • blk.10.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.10.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.10.ssm_a
    F32
    [1, 48]
  • blk.10.ssm_conv1d.bias
    F32
    [3328]
  • blk.10.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.10.ssm_d
    F32
    [1, 48]
  • blk.10.ssm_dt.bias
    F32
    [48]
  • blk.10.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.10.ssm_norm.weight
    F32
    [3072]
  • blk.10.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.11
  • blk.11.attn_norm.weight
    F32
    [1536]
  • blk.11.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.11.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.11.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.11.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.11.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.11.ffn_norm.weight
    F32
    [1536]
  • blk.11.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.11.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.11.ssm_a
    F32
    [1, 48]
  • blk.11.ssm_conv1d.bias
    F32
    [3328]
  • blk.11.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.11.ssm_d
    F32
    [1, 48]
  • blk.11.ssm_dt.bias
    F32
    [48]
  • blk.11.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.11.ssm_norm.weight
    F32
    [3072]
  • blk.11.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.12
  • blk.12.attn_norm.weight
    F32
    [1536]
  • blk.12.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.12.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.12.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.12.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.12.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.12.ffn_norm.weight
    F32
    [1536]
  • blk.12.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.12.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.12.ssm_a
    F32
    [1, 48]
  • blk.12.ssm_conv1d.bias
    F32
    [3328]
  • blk.12.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.12.ssm_d
    F32
    [1, 48]
  • blk.12.ssm_dt.bias
    F32
    [48]
  • blk.12.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.12.ssm_norm.weight
    F32
    [3072]
  • blk.12.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.13
  • blk.13.attn_norm.weight
    F32
    [1536]
  • blk.13.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.13.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.13.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.13.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.13.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.13.ffn_norm.weight
    F32
    [1536]
  • blk.13.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.13.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.13.ssm_a
    F32
    [1, 48]
  • blk.13.ssm_conv1d.bias
    F32
    [3328]
  • blk.13.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.13.ssm_d
    F32
    [1, 48]
  • blk.13.ssm_dt.bias
    F32
    [48]
  • blk.13.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.13.ssm_norm.weight
    F32
    [3072]
  • blk.13.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.14
  • blk.14.attn_norm.weight
    F32
    [1536]
  • blk.14.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.14.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.14.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.14.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.14.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.14.ffn_norm.weight
    F32
    [1536]
  • blk.14.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.14.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.14.ssm_a
    F32
    [1, 48]
  • blk.14.ssm_conv1d.bias
    F32
    [3328]
  • blk.14.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.14.ssm_d
    F32
    [1, 48]
  • blk.14.ssm_dt.bias
    F32
    [48]
  • blk.14.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.14.ssm_norm.weight
    F32
    [3072]
  • blk.14.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.15
  • blk.15.attn_k.weight
    Q4_K
    [1536, 512]
  • blk.15.attn_norm.weight
    F32
    [1536]
  • blk.15.attn_output.weight
    Q4_K
    [1536, 1536]
  • blk.15.attn_q.weight
    Q4_K
    [1536, 1536]
  • blk.15.attn_v.weight
    Q4_K
    [1536, 512]
  • blk.15.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.15.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.15.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.15.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.15.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.15.ffn_norm.weight
    F32
    [1536]
  • blk.15.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.15.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.16
  • blk.16.attn_norm.weight
    F32
    [1536]
  • blk.16.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.16.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.16.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.16.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.16.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.16.ffn_norm.weight
    F32
    [1536]
  • blk.16.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.16.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.16.ssm_a
    F32
    [1, 48]
  • blk.16.ssm_conv1d.bias
    F32
    [3328]
  • blk.16.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.16.ssm_d
    F32
    [1, 48]
  • blk.16.ssm_dt.bias
    F32
    [48]
  • blk.16.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.16.ssm_norm.weight
    F32
    [3072]
  • blk.16.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.17
  • blk.17.attn_norm.weight
    F32
    [1536]
  • blk.17.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.17.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.17.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.17.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.17.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.17.ffn_norm.weight
    F32
    [1536]
  • blk.17.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.17.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.17.ssm_a
    F32
    [1, 48]
  • blk.17.ssm_conv1d.bias
    F32
    [3328]
  • blk.17.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.17.ssm_d
    F32
    [1, 48]
  • blk.17.ssm_dt.bias
    F32
    [48]
  • blk.17.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.17.ssm_norm.weight
    F32
    [3072]
  • blk.17.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.18
  • blk.18.attn_norm.weight
    F32
    [1536]
  • blk.18.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.18.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.18.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.18.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.18.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.18.ffn_norm.weight
    F32
    [1536]
  • blk.18.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.18.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.18.ssm_a
    F32
    [1, 48]
  • blk.18.ssm_conv1d.bias
    F32
    [3328]
  • blk.18.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.18.ssm_d
    F32
    [1, 48]
  • blk.18.ssm_dt.bias
    F32
    [48]
  • blk.18.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.18.ssm_norm.weight
    F32
    [3072]
  • blk.18.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.19
  • blk.19.attn_norm.weight
    F32
    [1536]
  • blk.19.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.19.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.19.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.19.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.19.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.19.ffn_norm.weight
    F32
    [1536]
  • blk.19.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.19.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.19.ssm_a
    F32
    [1, 48]
  • blk.19.ssm_conv1d.bias
    F32
    [3328]
  • blk.19.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.19.ssm_d
    F32
    [1, 48]
  • blk.19.ssm_dt.bias
    F32
    [48]
  • blk.19.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.19.ssm_norm.weight
    F32
    [3072]
  • blk.19.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.20
  • blk.20.attn_norm.weight
    F32
    [1536]
  • blk.20.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.20.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.20.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.20.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.20.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.20.ffn_norm.weight
    F32
    [1536]
  • blk.20.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.20.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.20.ssm_a
    F32
    [1, 48]
  • blk.20.ssm_conv1d.bias
    F32
    [3328]
  • blk.20.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.20.ssm_d
    F32
    [1, 48]
  • blk.20.ssm_dt.bias
    F32
    [48]
  • blk.20.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.20.ssm_norm.weight
    F32
    [3072]
  • blk.20.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.21
  • blk.21.attn_norm.weight
    F32
    [1536]
  • blk.21.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.21.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.21.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.21.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.21.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.21.ffn_norm.weight
    F32
    [1536]
  • blk.21.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.21.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.21.ssm_a
    F32
    [1, 48]
  • blk.21.ssm_conv1d.bias
    F32
    [3328]
  • blk.21.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.21.ssm_d
    F32
    [1, 48]
  • blk.21.ssm_dt.bias
    F32
    [48]
  • blk.21.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.21.ssm_norm.weight
    F32
    [3072]
  • blk.21.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.22
  • blk.22.attn_norm.weight
    F32
    [1536]
  • blk.22.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.22.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.22.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.22.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.22.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.22.ffn_norm.weight
    F32
    [1536]
  • blk.22.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.22.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.22.ssm_a
    F32
    [1, 48]
  • blk.22.ssm_conv1d.bias
    F32
    [3328]
  • blk.22.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.22.ssm_d
    F32
    [1, 48]
  • blk.22.ssm_dt.bias
    F32
    [48]
  • blk.22.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.22.ssm_norm.weight
    F32
    [3072]
  • blk.22.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.23
  • blk.23.attn_norm.weight
    F32
    [1536]
  • blk.23.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.23.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.23.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.23.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.23.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.23.ffn_norm.weight
    F32
    [1536]
  • blk.23.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.23.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.23.ssm_a
    F32
    [1, 48]
  • blk.23.ssm_conv1d.bias
    F32
    [3328]
  • blk.23.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.23.ssm_d
    F32
    [1, 48]
  • blk.23.ssm_dt.bias
    F32
    [48]
  • blk.23.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.23.ssm_norm.weight
    F32
    [3072]
  • blk.23.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.24
  • blk.24.attn_norm.weight
    F32
    [1536]
  • blk.24.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.24.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.24.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.24.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.24.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.24.ffn_norm.weight
    F32
    [1536]
  • blk.24.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.24.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.24.ssm_a
    F32
    [1, 48]
  • blk.24.ssm_conv1d.bias
    F32
    [3328]
  • blk.24.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.24.ssm_d
    F32
    [1, 48]
  • blk.24.ssm_dt.bias
    F32
    [48]
  • blk.24.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.24.ssm_norm.weight
    F32
    [3072]
  • blk.24.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.25
  • blk.25.attn_k.weight
    Q4_K
    [1536, 512]
  • blk.25.attn_norm.weight
    F32
    [1536]
  • blk.25.attn_output.weight
    Q4_K
    [1536, 1536]
  • blk.25.attn_q.weight
    Q4_K
    [1536, 1536]
  • blk.25.attn_v.weight
    Q6_K
    [1536, 512]
  • blk.25.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.25.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.25.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.25.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.25.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.25.ffn_norm.weight
    F32
    [1536]
  • blk.25.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.25.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.26
  • blk.26.attn_norm.weight
    F32
    [1536]
  • blk.26.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.26.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.26.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.26.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.26.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.26.ffn_norm.weight
    F32
    [1536]
  • blk.26.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.26.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.26.ssm_a
    F32
    [1, 48]
  • blk.26.ssm_conv1d.bias
    F32
    [3328]
  • blk.26.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.26.ssm_d
    F32
    [1, 48]
  • blk.26.ssm_dt.bias
    F32
    [48]
  • blk.26.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.26.ssm_norm.weight
    F32
    [3072]
  • blk.26.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.27
  • blk.27.attn_norm.weight
    F32
    [1536]
  • blk.27.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.27.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.27.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.27.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.27.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.27.ffn_norm.weight
    F32
    [1536]
  • blk.27.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.27.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.27.ssm_a
    F32
    [1, 48]
  • blk.27.ssm_conv1d.bias
    F32
    [3328]
  • blk.27.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.27.ssm_d
    F32
    [1, 48]
  • blk.27.ssm_dt.bias
    F32
    [48]
  • blk.27.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.27.ssm_norm.weight
    F32
    [3072]
  • blk.27.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.28
  • blk.28.attn_norm.weight
    F32
    [1536]
  • blk.28.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.28.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.28.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.28.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.28.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.28.ffn_norm.weight
    F32
    [1536]
  • blk.28.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.28.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.28.ssm_a
    F32
    [1, 48]
  • blk.28.ssm_conv1d.bias
    F32
    [3328]
  • blk.28.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.28.ssm_d
    F32
    [1, 48]
  • blk.28.ssm_dt.bias
    F32
    [48]
  • blk.28.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.28.ssm_norm.weight
    F32
    [3072]
  • blk.28.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.29
  • blk.29.attn_norm.weight
    F32
    [1536]
  • blk.29.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.29.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.29.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.29.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.29.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.29.ffn_norm.weight
    F32
    [1536]
  • blk.29.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.29.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.29.ssm_a
    F32
    [1, 48]
  • blk.29.ssm_conv1d.bias
    F32
    [3328]
  • blk.29.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.29.ssm_d
    F32
    [1, 48]
  • blk.29.ssm_dt.bias
    F32
    [48]
  • blk.29.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.29.ssm_norm.weight
    F32
    [3072]
  • blk.29.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.30
  • blk.30.attn_norm.weight
    F32
    [1536]
  • blk.30.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.30.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.30.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.30.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.30.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.30.ffn_norm.weight
    F32
    [1536]
  • blk.30.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.30.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.30.ssm_a
    F32
    [1, 48]
  • blk.30.ssm_conv1d.bias
    F32
    [3328]
  • blk.30.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.30.ssm_d
    F32
    [1, 48]
  • blk.30.ssm_dt.bias
    F32
    [48]
  • blk.30.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.30.ssm_norm.weight
    F32
    [3072]
  • blk.30.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.31
  • blk.31.attn_norm.weight
    F32
    [1536]
  • blk.31.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.31.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.31.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.31.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.31.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.31.ffn_norm.weight
    F32
    [1536]
  • blk.31.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.31.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.31.ssm_a
    F32
    [1, 48]
  • blk.31.ssm_conv1d.bias
    F32
    [3328]
  • blk.31.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.31.ssm_d
    F32
    [1, 48]
  • blk.31.ssm_dt.bias
    F32
    [48]
  • blk.31.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.31.ssm_norm.weight
    F32
    [3072]
  • blk.31.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.32
  • blk.32.attn_norm.weight
    F32
    [1536]
  • blk.32.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.32.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.32.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.32.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.32.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.32.ffn_norm.weight
    F32
    [1536]
  • blk.32.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.32.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.32.ssm_a
    F32
    [1, 48]
  • blk.32.ssm_conv1d.bias
    F32
    [3328]
  • blk.32.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.32.ssm_d
    F32
    [1, 48]
  • blk.32.ssm_dt.bias
    F32
    [48]
  • blk.32.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.32.ssm_norm.weight
    F32
    [3072]
  • blk.32.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.33
  • blk.33.attn_norm.weight
    F32
    [1536]
  • blk.33.ffn_down_exps.weight
    Q4_K
    [512, 1536, 62]
  • blk.33.ffn_down_shexp.weight
    Q4_K
    [1024, 1536]
  • blk.33.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.33.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.33.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.33.ffn_norm.weight
    F32
    [1536]
  • blk.33.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.33.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.33.ssm_a
    F32
    [1, 48]
  • blk.33.ssm_conv1d.bias
    F32
    [3328]
  • blk.33.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.33.ssm_d
    F32
    [1, 48]
  • blk.33.ssm_dt.bias
    F32
    [48]
  • blk.33.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.33.ssm_norm.weight
    F32
    [3072]
  • blk.33.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.34
  • blk.34.attn_norm.weight
    F32
    [1536]
  • blk.34.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.34.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.34.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.34.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.34.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.34.ffn_norm.weight
    F32
    [1536]
  • blk.34.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.34.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.34.ssm_a
    F32
    [1, 48]
  • blk.34.ssm_conv1d.bias
    F32
    [3328]
  • blk.34.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.34.ssm_d
    F32
    [1, 48]
  • blk.34.ssm_dt.bias
    F32
    [48]
  • blk.34.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.34.ssm_norm.weight
    F32
    [3072]
  • blk.34.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.35
  • blk.35.attn_k.weight
    Q4_K
    [1536, 512]
  • blk.35.attn_norm.weight
    F32
    [1536]
  • blk.35.attn_output.weight
    Q4_K
    [1536, 1536]
  • blk.35.attn_q.weight
    Q4_K
    [1536, 1536]
  • blk.35.attn_v.weight
    Q6_K
    [1536, 512]
  • blk.35.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.35.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.35.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.35.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.35.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.35.ffn_norm.weight
    F32
    [1536]
  • blk.35.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.35.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.36
  • blk.36.attn_norm.weight
    F32
    [1536]
  • blk.36.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.36.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.36.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.36.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.36.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.36.ffn_norm.weight
    F32
    [1536]
  • blk.36.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.36.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.36.ssm_a
    F32
    [1, 48]
  • blk.36.ssm_conv1d.bias
    F32
    [3328]
  • blk.36.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.36.ssm_d
    F32
    [1, 48]
  • blk.36.ssm_dt.bias
    F32
    [48]
  • blk.36.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.36.ssm_norm.weight
    F32
    [3072]
  • blk.36.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.37
  • blk.37.attn_norm.weight
    F32
    [1536]
  • blk.37.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.37.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.37.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.37.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.37.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.37.ffn_norm.weight
    F32
    [1536]
  • blk.37.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.37.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.37.ssm_a
    F32
    [1, 48]
  • blk.37.ssm_conv1d.bias
    F32
    [3328]
  • blk.37.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.37.ssm_d
    F32
    [1, 48]
  • blk.37.ssm_dt.bias
    F32
    [48]
  • blk.37.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.37.ssm_norm.weight
    F32
    [3072]
  • blk.37.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.38
  • blk.38.attn_norm.weight
    F32
    [1536]
  • blk.38.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.38.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.38.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.38.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.38.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.38.ffn_norm.weight
    F32
    [1536]
  • blk.38.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.38.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.38.ssm_a
    F32
    [1, 48]
  • blk.38.ssm_conv1d.bias
    F32
    [3328]
  • blk.38.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.38.ssm_d
    F32
    [1, 48]
  • blk.38.ssm_dt.bias
    F32
    [48]
  • blk.38.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.38.ssm_norm.weight
    F32
    [3072]
  • blk.38.ssm_out.weight
    Q4_K
    [3072, 1536]
  • blk.39
  • blk.39.attn_norm.weight
    F32
    [1536]
  • blk.39.ffn_down_exps.weight
    Q6_K
    [512, 1536, 62]
  • blk.39.ffn_down_shexp.weight
    Q6_K
    [1024, 1536]
  • blk.39.ffn_gate_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.39.ffn_gate_inp.weight
    F32
    [1536, 62]
  • blk.39.ffn_gate_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.39.ffn_norm.weight
    F32
    [1536]
  • blk.39.ffn_up_exps.weight
    Q4_K
    [1536, 512, 62]
  • blk.39.ffn_up_shexp.weight
    Q4_K
    [1536, 1024]
  • blk.39.ssm_a
    F32
    [1, 48]
  • blk.39.ssm_conv1d.bias
    F32
    [3328]
  • blk.39.ssm_conv1d.weight
    F32
    [4, 3328]
  • blk.39.ssm_d
    F32
    [1, 48]
  • blk.39.ssm_dt.bias
    F32
    [48]
  • blk.39.ssm_in.weight
    Q4_K
    [1536, 6448]
  • blk.39.ssm_norm.weight
    F32
    [3072]
  • blk.39.ssm_out.weight
    Q4_K
    [3072, 1536]
  • output_norm.weight
    F32
    [1536]