588 1 week ago

PLaMo Translation Model is a specialized large-scale language model developed by Preferred Networks for translation tasks.

f57e514b8015 · 3.8GB
    Metadata
  • general.architecture
    plamo2
  • general.file_type
    IQ3_XXS
  • plamo2.attention.head_count
    [0, 32, 0, 32, 0, ...]
  • plamo2.attention.head_count_kv
    [0, 4, 0, 4, 0, ...]
  • plamo2.attention.key_length
    128
  • plamo2.attention.layer_norm_rms_epsilon
    1e-06
  • plamo2.attention.value_length
    128
  • plamo2.block_count
    32
  • plamo2.context_length
    10485760
  • plamo2.embedding_length
    4096
  • plamo2.feed_forward_length
    16384
  • plamo2.rope.freq_base
    1e+06
  • plamo2.ssm.conv_kernel
    4
  • plamo2.ssm.group_count
    0
  • plamo2.ssm.inner_size
    8192
  • plamo2.ssm.state_size
    64
  • plamo2.ssm.time_step_rank
    64
  • plamo2.vocab_size
    100032
  • tokenizer.ggml.add_space_prefix
    false
  • tokenizer.ggml.bos_token_id
    1
  • tokenizer.ggml.eos_token_id
    1
  • tokenizer.ggml.eot_token_id
    4
  • tokenizer.ggml.model
    plamo2
  • tokenizer.ggml.padding_token_id
    3
  • tokenizer.ggml.pre
    default
  • tokenizer.ggml.scores
    [0, 0, 0, 0, 0, ...]
  • tokenizer.ggml.token_type
    [2, 3, 3, 3, 3, ...]
  • tokenizer.ggml.tokens
    [<|plamo:unk|>, <|plamo:bos|>, <|plamo:eos|>, <|plamo:pad|>, <|plamo:op|>, ...]
  • tokenizer.ggml.unknown_token_id
    0
  • quantize.imatrix.chunks_count
    154
  • quantize.imatrix.dataset
    /mnt/shared/users/mitmul/imatrix-dataset-for-japanese-llm/c4_en_ja_imatrix.txt
  • quantize.imatrix.entries_count
    160
  • quantize.imatrix.file
    ../ollama/tmp/plamo-2-translate-auto-imatrix.gguf
  • Tensor
  • token_embd.weight
    IQ3_S
    [4096, 100032]
  • blk.0
  • blk.0.attn_norm.weight
    F32
    [4096]
  • blk.0.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.0.ffn_norm.weight
    F32
    [4096]
  • blk.0.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.0.post_attention_norm
    F32
    [4096]
  • blk.0.post_ffw_norm
    F32
    [4096]
  • blk.0.ssm_a
    F32
    [64]
  • blk.0.ssm_b_norm
    F32
    [64]
  • blk.0.ssm_c_norm
    F32
    [64]
  • blk.0.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.0.ssm_d
    F32
    [64]
  • blk.0.ssm_dt.bias
    F32
    [64]
  • blk.0.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.0.ssm_dt_norm
    F32
    [256]
  • blk.0.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.0.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.0.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.1
  • blk.1.attn_k_norm.weight
    F32
    [128, 4]
  • blk.1.attn_norm.weight
    F32
    [4096]
  • blk.1.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.1.attn_q_norm.weight
    F32
    [128, 32]
  • blk.1.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.1.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.1.ffn_norm.weight
    F32
    [4096]
  • blk.1.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.1.post_attention_norm
    F32
    [4096]
  • blk.1.post_ffw_norm
    F32
    [4096]
  • blk.2
  • blk.2.attn_norm.weight
    F32
    [4096]
  • blk.2.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.2.ffn_norm.weight
    F32
    [4096]
  • blk.2.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.2.post_attention_norm
    F32
    [4096]
  • blk.2.post_ffw_norm
    F32
    [4096]
  • blk.2.ssm_a
    F32
    [64]
  • blk.2.ssm_b_norm
    F32
    [64]
  • blk.2.ssm_c_norm
    F32
    [64]
  • blk.2.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.2.ssm_d
    F32
    [64]
  • blk.2.ssm_dt.bias
    F32
    [64]
  • blk.2.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.2.ssm_dt_norm
    F32
    [256]
  • blk.2.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.2.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.2.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.3
  • blk.3.attn_k_norm.weight
    F32
    [128, 4]
  • blk.3.attn_norm.weight
    F32
    [4096]
  • blk.3.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.3.attn_q_norm.weight
    F32
    [128, 32]
  • blk.3.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.3.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.3.ffn_norm.weight
    F32
    [4096]
  • blk.3.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.3.post_attention_norm
    F32
    [4096]
  • blk.3.post_ffw_norm
    F32
    [4096]
  • blk.4
  • blk.4.attn_norm.weight
    F32
    [4096]
  • blk.4.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.4.ffn_norm.weight
    F32
    [4096]
  • blk.4.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.4.post_attention_norm
    F32
    [4096]
  • blk.4.post_ffw_norm
    F32
    [4096]
  • blk.4.ssm_a
    F32
    [64]
  • blk.4.ssm_b_norm
    F32
    [64]
  • blk.4.ssm_c_norm
    F32
    [64]
  • blk.4.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.4.ssm_d
    F32
    [64]
  • blk.4.ssm_dt.bias
    F32
    [64]
  • blk.4.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.4.ssm_dt_norm
    F32
    [256]
  • blk.4.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.4.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.4.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.5
  • blk.5.attn_k_norm.weight
    F32
    [128, 4]
  • blk.5.attn_norm.weight
    F32
    [4096]
  • blk.5.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.5.attn_q_norm.weight
    F32
    [128, 32]
  • blk.5.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.5.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.5.ffn_norm.weight
    F32
    [4096]
  • blk.5.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.5.post_attention_norm
    F32
    [4096]
  • blk.5.post_ffw_norm
    F32
    [4096]
  • blk.6
  • blk.6.attn_norm.weight
    F32
    [4096]
  • blk.6.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.6.ffn_norm.weight
    F32
    [4096]
  • blk.6.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.6.post_attention_norm
    F32
    [4096]
  • blk.6.post_ffw_norm
    F32
    [4096]
  • blk.6.ssm_a
    F32
    [64]
  • blk.6.ssm_b_norm
    F32
    [64]
  • blk.6.ssm_c_norm
    F32
    [64]
  • blk.6.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.6.ssm_d
    F32
    [64]
  • blk.6.ssm_dt.bias
    F32
    [64]
  • blk.6.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.6.ssm_dt_norm
    F32
    [256]
  • blk.6.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.6.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.6.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.7
  • blk.7.attn_k_norm.weight
    F32
    [128, 4]
  • blk.7.attn_norm.weight
    F32
    [4096]
  • blk.7.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.7.attn_q_norm.weight
    F32
    [128, 32]
  • blk.7.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.7.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.7.ffn_norm.weight
    F32
    [4096]
  • blk.7.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.7.post_attention_norm
    F32
    [4096]
  • blk.7.post_ffw_norm
    F32
    [4096]
  • blk.8
  • blk.8.attn_norm.weight
    F32
    [4096]
  • blk.8.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.8.ffn_norm.weight
    F32
    [4096]
  • blk.8.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.8.post_attention_norm
    F32
    [4096]
  • blk.8.post_ffw_norm
    F32
    [4096]
  • blk.8.ssm_a
    F32
    [64]
  • blk.8.ssm_b_norm
    F32
    [64]
  • blk.8.ssm_c_norm
    F32
    [64]
  • blk.8.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.8.ssm_d
    F32
    [64]
  • blk.8.ssm_dt.bias
    F32
    [64]
  • blk.8.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.8.ssm_dt_norm
    F32
    [256]
  • blk.8.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.8.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.8.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.9
  • blk.9.attn_k_norm.weight
    F32
    [128, 4]
  • blk.9.attn_norm.weight
    F32
    [4096]
  • blk.9.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.9.attn_q_norm.weight
    F32
    [128, 32]
  • blk.9.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.9.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.9.ffn_norm.weight
    F32
    [4096]
  • blk.9.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.9.post_attention_norm
    F32
    [4096]
  • blk.9.post_ffw_norm
    F32
    [4096]
  • blk.10
  • blk.10.attn_norm.weight
    F32
    [4096]
  • blk.10.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.10.ffn_norm.weight
    F32
    [4096]
  • blk.10.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.10.post_attention_norm
    F32
    [4096]
  • blk.10.post_ffw_norm
    F32
    [4096]
  • blk.10.ssm_a
    F32
    [64]
  • blk.10.ssm_b_norm
    F32
    [64]
  • blk.10.ssm_c_norm
    F32
    [64]
  • blk.10.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.10.ssm_d
    F32
    [64]
  • blk.10.ssm_dt.bias
    F32
    [64]
  • blk.10.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.10.ssm_dt_norm
    F32
    [256]
  • blk.10.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.10.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.10.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.11
  • blk.11.attn_k_norm.weight
    F32
    [128, 4]
  • blk.11.attn_norm.weight
    F32
    [4096]
  • blk.11.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.11.attn_q_norm.weight
    F32
    [128, 32]
  • blk.11.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.11.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.11.ffn_norm.weight
    F32
    [4096]
  • blk.11.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.11.post_attention_norm
    F32
    [4096]
  • blk.11.post_ffw_norm
    F32
    [4096]
  • blk.12
  • blk.12.attn_norm.weight
    F32
    [4096]
  • blk.12.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.12.ffn_norm.weight
    F32
    [4096]
  • blk.12.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.12.post_attention_norm
    F32
    [4096]
  • blk.12.post_ffw_norm
    F32
    [4096]
  • blk.12.ssm_a
    F32
    [64]
  • blk.12.ssm_b_norm
    F32
    [64]
  • blk.12.ssm_c_norm
    F32
    [64]
  • blk.12.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.12.ssm_d
    F32
    [64]
  • blk.12.ssm_dt.bias
    F32
    [64]
  • blk.12.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.12.ssm_dt_norm
    F32
    [256]
  • blk.12.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.12.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.12.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.13
  • blk.13.attn_k_norm.weight
    F32
    [128, 4]
  • blk.13.attn_norm.weight
    F32
    [4096]
  • blk.13.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.13.attn_q_norm.weight
    F32
    [128, 32]
  • blk.13.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.13.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.13.ffn_norm.weight
    F32
    [4096]
  • blk.13.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.13.post_attention_norm
    F32
    [4096]
  • blk.13.post_ffw_norm
    F32
    [4096]
  • blk.14
  • blk.14.attn_norm.weight
    F32
    [4096]
  • blk.14.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.14.ffn_norm.weight
    F32
    [4096]
  • blk.14.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.14.post_attention_norm
    F32
    [4096]
  • blk.14.post_ffw_norm
    F32
    [4096]
  • blk.14.ssm_a
    F32
    [64]
  • blk.14.ssm_b_norm
    F32
    [64]
  • blk.14.ssm_c_norm
    F32
    [64]
  • blk.14.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.14.ssm_d
    F32
    [64]
  • blk.14.ssm_dt.bias
    F32
    [64]
  • blk.14.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.14.ssm_dt_norm
    F32
    [256]
  • blk.14.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.14.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.14.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.15
  • blk.15.attn_k_norm.weight
    F32
    [128, 4]
  • blk.15.attn_norm.weight
    F32
    [4096]
  • blk.15.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.15.attn_q_norm.weight
    F32
    [128, 32]
  • blk.15.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.15.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.15.ffn_norm.weight
    F32
    [4096]
  • blk.15.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.15.post_attention_norm
    F32
    [4096]
  • blk.15.post_ffw_norm
    F32
    [4096]
  • blk.16
  • blk.16.attn_norm.weight
    F32
    [4096]
  • blk.16.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.16.ffn_norm.weight
    F32
    [4096]
  • blk.16.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.16.post_attention_norm
    F32
    [4096]
  • blk.16.post_ffw_norm
    F32
    [4096]
  • blk.16.ssm_a
    F32
    [64]
  • blk.16.ssm_b_norm
    F32
    [64]
  • blk.16.ssm_c_norm
    F32
    [64]
  • blk.16.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.16.ssm_d
    F32
    [64]
  • blk.16.ssm_dt.bias
    F32
    [64]
  • blk.16.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.16.ssm_dt_norm
    F32
    [256]
  • blk.16.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.16.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.16.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.17
  • blk.17.attn_k_norm.weight
    F32
    [128, 4]
  • blk.17.attn_norm.weight
    F32
    [4096]
  • blk.17.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.17.attn_q_norm.weight
    F32
    [128, 32]
  • blk.17.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.17.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.17.ffn_norm.weight
    F32
    [4096]
  • blk.17.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.17.post_attention_norm
    F32
    [4096]
  • blk.17.post_ffw_norm
    F32
    [4096]
  • blk.18
  • blk.18.attn_norm.weight
    F32
    [4096]
  • blk.18.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.18.ffn_norm.weight
    F32
    [4096]
  • blk.18.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.18.post_attention_norm
    F32
    [4096]
  • blk.18.post_ffw_norm
    F32
    [4096]
  • blk.18.ssm_a
    F32
    [64]
  • blk.18.ssm_b_norm
    F32
    [64]
  • blk.18.ssm_c_norm
    F32
    [64]
  • blk.18.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.18.ssm_d
    F32
    [64]
  • blk.18.ssm_dt.bias
    F32
    [64]
  • blk.18.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.18.ssm_dt_norm
    F32
    [256]
  • blk.18.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.18.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.18.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.19
  • blk.19.attn_k_norm.weight
    F32
    [128, 4]
  • blk.19.attn_norm.weight
    F32
    [4096]
  • blk.19.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.19.attn_q_norm.weight
    F32
    [128, 32]
  • blk.19.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.19.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.19.ffn_norm.weight
    F32
    [4096]
  • blk.19.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.19.post_attention_norm
    F32
    [4096]
  • blk.19.post_ffw_norm
    F32
    [4096]
  • blk.20
  • blk.20.attn_norm.weight
    F32
    [4096]
  • blk.20.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.20.ffn_norm.weight
    F32
    [4096]
  • blk.20.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.20.post_attention_norm
    F32
    [4096]
  • blk.20.post_ffw_norm
    F32
    [4096]
  • blk.20.ssm_a
    F32
    [64]
  • blk.20.ssm_b_norm
    F32
    [64]
  • blk.20.ssm_c_norm
    F32
    [64]
  • blk.20.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.20.ssm_d
    F32
    [64]
  • blk.20.ssm_dt.bias
    F32
    [64]
  • blk.20.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.20.ssm_dt_norm
    F32
    [256]
  • blk.20.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.20.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.20.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.21
  • blk.21.attn_k_norm.weight
    F32
    [128, 4]
  • blk.21.attn_norm.weight
    F32
    [4096]
  • blk.21.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.21.attn_q_norm.weight
    F32
    [128, 32]
  • blk.21.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.21.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.21.ffn_norm.weight
    F32
    [4096]
  • blk.21.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.21.post_attention_norm
    F32
    [4096]
  • blk.21.post_ffw_norm
    F32
    [4096]
  • blk.22
  • blk.22.attn_norm.weight
    F32
    [4096]
  • blk.22.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.22.ffn_norm.weight
    F32
    [4096]
  • blk.22.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.22.post_attention_norm
    F32
    [4096]
  • blk.22.post_ffw_norm
    F32
    [4096]
  • blk.22.ssm_a
    F32
    [64]
  • blk.22.ssm_b_norm
    F32
    [64]
  • blk.22.ssm_c_norm
    F32
    [64]
  • blk.22.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.22.ssm_d
    F32
    [64]
  • blk.22.ssm_dt.bias
    F32
    [64]
  • blk.22.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.22.ssm_dt_norm
    F32
    [256]
  • blk.22.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.22.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.22.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.23
  • blk.23.attn_k_norm.weight
    F32
    [128, 4]
  • blk.23.attn_norm.weight
    F32
    [4096]
  • blk.23.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.23.attn_q_norm.weight
    F32
    [128, 32]
  • blk.23.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.23.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.23.ffn_norm.weight
    F32
    [4096]
  • blk.23.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.23.post_attention_norm
    F32
    [4096]
  • blk.23.post_ffw_norm
    F32
    [4096]
  • blk.24
  • blk.24.attn_norm.weight
    F32
    [4096]
  • blk.24.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.24.ffn_norm.weight
    F32
    [4096]
  • blk.24.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.24.post_attention_norm
    F32
    [4096]
  • blk.24.post_ffw_norm
    F32
    [4096]
  • blk.24.ssm_a
    F32
    [64]
  • blk.24.ssm_b_norm
    F32
    [64]
  • blk.24.ssm_c_norm
    F32
    [64]
  • blk.24.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.24.ssm_d
    F32
    [64]
  • blk.24.ssm_dt.bias
    F32
    [64]
  • blk.24.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.24.ssm_dt_norm
    F32
    [256]
  • blk.24.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.24.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.24.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.25
  • blk.25.attn_k_norm.weight
    F32
    [128, 4]
  • blk.25.attn_norm.weight
    F32
    [4096]
  • blk.25.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.25.attn_q_norm.weight
    F32
    [128, 32]
  • blk.25.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.25.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.25.ffn_norm.weight
    F32
    [4096]
  • blk.25.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.25.post_attention_norm
    F32
    [4096]
  • blk.25.post_ffw_norm
    F32
    [4096]
  • blk.26
  • blk.26.attn_norm.weight
    F32
    [4096]
  • blk.26.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.26.ffn_norm.weight
    F32
    [4096]
  • blk.26.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.26.post_attention_norm
    F32
    [4096]
  • blk.26.post_ffw_norm
    F32
    [4096]
  • blk.26.ssm_a
    F32
    [64]
  • blk.26.ssm_b_norm
    F32
    [64]
  • blk.26.ssm_c_norm
    F32
    [64]
  • blk.26.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.26.ssm_d
    F32
    [64]
  • blk.26.ssm_dt.bias
    F32
    [64]
  • blk.26.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.26.ssm_dt_norm
    F32
    [256]
  • blk.26.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.26.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.26.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.27
  • blk.27.attn_k_norm.weight
    F32
    [128, 4]
  • blk.27.attn_norm.weight
    F32
    [4096]
  • blk.27.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.27.attn_q_norm.weight
    F32
    [128, 32]
  • blk.27.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.27.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.27.ffn_norm.weight
    F32
    [4096]
  • blk.27.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.27.post_attention_norm
    F32
    [4096]
  • blk.27.post_ffw_norm
    F32
    [4096]
  • blk.28
  • blk.28.attn_norm.weight
    F32
    [4096]
  • blk.28.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.28.ffn_norm.weight
    F32
    [4096]
  • blk.28.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.28.post_attention_norm
    F32
    [4096]
  • blk.28.post_ffw_norm
    F32
    [4096]
  • blk.28.ssm_a
    F32
    [64]
  • blk.28.ssm_b_norm
    F32
    [64]
  • blk.28.ssm_c_norm
    F32
    [64]
  • blk.28.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.28.ssm_d
    F32
    [64]
  • blk.28.ssm_dt.bias
    F32
    [64]
  • blk.28.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.28.ssm_dt_norm
    F32
    [256]
  • blk.28.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.28.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.28.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.29
  • blk.29.attn_k_norm.weight
    F32
    [128, 4]
  • blk.29.attn_norm.weight
    F32
    [4096]
  • blk.29.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.29.attn_q_norm.weight
    F32
    [128, 32]
  • blk.29.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.29.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.29.ffn_norm.weight
    F32
    [4096]
  • blk.29.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.29.post_attention_norm
    F32
    [4096]
  • blk.29.post_ffw_norm
    F32
    [4096]
  • blk.30
  • blk.30.attn_norm.weight
    F32
    [4096]
  • blk.30.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.30.ffn_norm.weight
    F32
    [4096]
  • blk.30.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.30.post_attention_norm
    F32
    [4096]
  • blk.30.post_ffw_norm
    F32
    [4096]
  • blk.30.ssm_a
    F32
    [64]
  • blk.30.ssm_b_norm
    F32
    [64]
  • blk.30.ssm_c_norm
    F32
    [64]
  • blk.30.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.30.ssm_d
    F32
    [64]
  • blk.30.ssm_dt.bias
    F32
    [64]
  • blk.30.ssm_dt.weight
    IQ3_XXS
    [256, 64]
  • blk.30.ssm_dt_norm
    F32
    [256]
  • blk.30.ssm_in.weight
    IQ3_XXS
    [4096, 16384]
  • blk.30.ssm_out.weight
    IQ3_XXS
    [8192, 4096]
  • blk.30.ssm_x.weight
    IQ3_XXS
    [8192, 384]
  • blk.31
  • blk.31.attn_k_norm.weight
    F32
    [128, 4]
  • blk.31.attn_norm.weight
    F32
    [4096]
  • blk.31.attn_output.weight
    IQ3_S
    [4096, 4096]
  • blk.31.attn_q_norm.weight
    F32
    [128, 32]
  • blk.31.attn_qkv.weight
    IQ3_XXS
    [4096, 5120]
  • blk.31.ffn_down.weight
    IQ3_XXS
    [16384, 4096]
  • blk.31.ffn_norm.weight
    F32
    [4096]
  • blk.31.ffn_up.weight
    IQ3_XXS
    [4096, 32768]
  • blk.31.post_attention_norm
    F32
    [4096]
  • blk.31.post_ffw_norm
    F32
    [4096]
  • output.weight
    Q5_K
    [4096, 100032]
  • output_norm.weight
    F32
    [4096]