qwen3.5:9b

174K just now

Qwen 3.5 is a family of open-source multimodal models that delivers exceptional utility and performance.

vision tools thinking cloud 0.8b 2b 4b 9b 27b 35b 122b
dec52a44569a · 6.6GB
    Metadata
  • general.architecture
    qwen35
  • general.file_type
    Q4_K_M
  • qwen35.attention.head_count
    16
  • qwen35.attention.head_count_kv
    [0, 0, 0, 4, 0, ...]
  • qwen35.attention.key_length
    256
  • qwen35.attention.layer_norm_rms_epsilon
    1e-06
  • qwen35.attention.value_length
    256
  • qwen35.block_count
    32
  • qwen35.context_length
    262144
  • qwen35.embedding_length
    4096
  • qwen35.feed_forward_length
    12288
  • qwen35.full_attention_interval
    4
  • qwen35.image_token_id
    248056
  • qwen35.mrope_sections
    [11, 11, 10]
  • qwen35.rope.dimension_count
    64
  • qwen35.rope.dimension_sections
    [11, 11, 10]
  • qwen35.rope.freq_base
    1e+07
  • qwen35.rope.mrope_interleaved
    true
  • qwen35.rope.mrope_section
    [11, 11, 10]
  • qwen35.ssm.conv_kernel
    4
  • qwen35.ssm.group_count
    16
  • qwen35.ssm.inner_size
    4096
  • qwen35.ssm.state_size
    128
  • qwen35.ssm.time_step_rank
    32
  • qwen35.ssm.v_head_reordered
    true
  • qwen35.vision.attention.head_count
    16
  • qwen35.vision.block_count
    27
  • qwen35.vision.deepstack_visual_indexes
    []
  • qwen35.vision.embedding_length
    1152
  • qwen35.vision.image_mean
    [0.5, 0.5, 0.5]
  • qwen35.vision.image_std
    [0.5, 0.5, 0.5]
  • qwen35.vision.longest_edge
    16777216
  • qwen35.vision.num_channels
    3
  • qwen35.vision.patch_size
    16
  • qwen35.vision.shortest_edge
    65536
  • qwen35.vision.spatial_merge_size
    2
  • qwen35.vision.temporal_patch_size
    2
  • qwen35.vision_end_token_id
    248054
  • qwen35.vision_start_token_id
    248053
  • tokenizer.ggml.add_eos_token
    false
  • tokenizer.ggml.add_padding_token
    false
  • tokenizer.ggml.eos_token_id
    248046
  • tokenizer.ggml.merges
    [Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
  • tokenizer.ggml.model
    gpt2
  • tokenizer.ggml.padding_token_id
    248044
  • tokenizer.ggml.pre
    qwen35
  • tokenizer.ggml.scores
    [0, 1, 2, 3, 4, ...]
  • tokenizer.ggml.token_type
    [1, 1, 1, 1, 1, ...]
  • tokenizer.ggml.tokens
    [!, ", #, $, %, ...]
  • Tensor
  • token_embd.weight
    Q4_K
    [4096, 248320]
  • blk.0
  • blk.0.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.0.attn_norm.weight
    F32
    [4096]
  • blk.0.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.0.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.0.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.0.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.0.post_attention_norm.weight
    F32
    [4096]
  • blk.0.ssm_a
    F32
    [32]
  • blk.0.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.0.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.0.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.0.ssm_dt
    F32
    [32]
  • blk.0.ssm_norm.weight
    F32
    [128]
  • blk.0.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.1
  • blk.1.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.1.attn_norm.weight
    F32
    [4096]
  • blk.1.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.1.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.1.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.1.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.1.post_attention_norm.weight
    F32
    [4096]
  • blk.1.ssm_a
    F32
    [32]
  • blk.1.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.1.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.1.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.1.ssm_dt
    F32
    [32]
  • blk.1.ssm_norm.weight
    F32
    [128]
  • blk.1.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.2
  • blk.2.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.2.attn_norm.weight
    F32
    [4096]
  • blk.2.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.2.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.2.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.2.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.2.post_attention_norm.weight
    F32
    [4096]
  • blk.2.ssm_a
    F32
    [32]
  • blk.2.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.2.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.2.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.2.ssm_dt
    F32
    [32]
  • blk.2.ssm_norm.weight
    F32
    [128]
  • blk.2.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.3
  • blk.3.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.3.attn_k_norm.weight
    F32
    [256]
  • blk.3.attn_norm.weight
    F32
    [4096]
  • blk.3.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.3.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.3.attn_q_norm.weight
    F32
    [256]
  • blk.3.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.3.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.3.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.3.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.3.post_attention_norm.weight
    F32
    [4096]
  • blk.4
  • blk.4.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.4.attn_norm.weight
    F32
    [4096]
  • blk.4.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.4.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.4.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.4.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.4.post_attention_norm.weight
    F32
    [4096]
  • blk.4.ssm_a
    F32
    [32]
  • blk.4.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.4.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.4.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.4.ssm_dt
    F32
    [32]
  • blk.4.ssm_norm.weight
    F32
    [128]
  • blk.4.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.5
  • blk.5.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.5.attn_norm.weight
    F32
    [4096]
  • blk.5.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.5.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.5.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.5.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.5.post_attention_norm.weight
    F32
    [4096]
  • blk.5.ssm_a
    F32
    [32]
  • blk.5.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.5.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.5.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.5.ssm_dt
    F32
    [32]
  • blk.5.ssm_norm.weight
    F32
    [128]
  • blk.5.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.6
  • blk.6.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.6.attn_norm.weight
    F32
    [4096]
  • blk.6.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.6.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.6.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.6.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.6.post_attention_norm.weight
    F32
    [4096]
  • blk.6.ssm_a
    F32
    [32]
  • blk.6.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.6.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.6.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.6.ssm_dt
    F32
    [32]
  • blk.6.ssm_norm.weight
    F32
    [128]
  • blk.6.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.7
  • blk.7.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.7.attn_k_norm.weight
    F32
    [256]
  • blk.7.attn_norm.weight
    F32
    [4096]
  • blk.7.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.7.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.7.attn_q_norm.weight
    F32
    [256]
  • blk.7.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.7.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.7.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.7.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.7.post_attention_norm.weight
    F32
    [4096]
  • blk.8
  • blk.8.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.8.attn_norm.weight
    F32
    [4096]
  • blk.8.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.8.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.8.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.8.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.8.post_attention_norm.weight
    F32
    [4096]
  • blk.8.ssm_a
    F32
    [32]
  • blk.8.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.8.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.8.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.8.ssm_dt
    F32
    [32]
  • blk.8.ssm_norm.weight
    F32
    [128]
  • blk.8.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.9
  • blk.9.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.9.attn_norm.weight
    F32
    [4096]
  • blk.9.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.9.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.9.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.9.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.9.post_attention_norm.weight
    F32
    [4096]
  • blk.9.ssm_a
    F32
    [32]
  • blk.9.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.9.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.9.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.9.ssm_dt
    F32
    [32]
  • blk.9.ssm_norm.weight
    F32
    [128]
  • blk.9.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.10
  • blk.10.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.10.attn_norm.weight
    F32
    [4096]
  • blk.10.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.10.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.10.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.10.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.10.post_attention_norm.weight
    F32
    [4096]
  • blk.10.ssm_a
    F32
    [32]
  • blk.10.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.10.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.10.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.10.ssm_dt
    F32
    [32]
  • blk.10.ssm_norm.weight
    F32
    [128]
  • blk.10.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.11
  • blk.11.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.11.attn_k_norm.weight
    F32
    [256]
  • blk.11.attn_norm.weight
    F32
    [4096]
  • blk.11.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.11.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.11.attn_q_norm.weight
    F32
    [256]
  • blk.11.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.11.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.11.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.11.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.11.post_attention_norm.weight
    F32
    [4096]
  • blk.12
  • blk.12.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.12.attn_norm.weight
    F32
    [4096]
  • blk.12.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.12.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.12.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.12.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.12.post_attention_norm.weight
    F32
    [4096]
  • blk.12.ssm_a
    F32
    [32]
  • blk.12.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.12.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.12.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.12.ssm_dt
    F32
    [32]
  • blk.12.ssm_norm.weight
    F32
    [128]
  • blk.12.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.13
  • blk.13.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.13.attn_norm.weight
    F32
    [4096]
  • blk.13.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.13.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.13.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.13.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.13.post_attention_norm.weight
    F32
    [4096]
  • blk.13.ssm_a
    F32
    [32]
  • blk.13.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.13.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.13.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.13.ssm_dt
    F32
    [32]
  • blk.13.ssm_norm.weight
    F32
    [128]
  • blk.13.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.14
  • blk.14.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.14.attn_norm.weight
    F32
    [4096]
  • blk.14.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.14.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.14.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.14.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.14.post_attention_norm.weight
    F32
    [4096]
  • blk.14.ssm_a
    F32
    [32]
  • blk.14.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.14.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.14.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.14.ssm_dt
    F32
    [32]
  • blk.14.ssm_norm.weight
    F32
    [128]
  • blk.14.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.15
  • blk.15.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.15.attn_k_norm.weight
    F32
    [256]
  • blk.15.attn_norm.weight
    F32
    [4096]
  • blk.15.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.15.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.15.attn_q_norm.weight
    F32
    [256]
  • blk.15.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.15.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.15.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.15.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.15.post_attention_norm.weight
    F32
    [4096]
  • blk.16
  • blk.16.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.16.attn_norm.weight
    F32
    [4096]
  • blk.16.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.16.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.16.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.16.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.16.post_attention_norm.weight
    F32
    [4096]
  • blk.16.ssm_a
    F32
    [32]
  • blk.16.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.16.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.16.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.16.ssm_dt
    F32
    [32]
  • blk.16.ssm_norm.weight
    F32
    [128]
  • blk.16.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.17
  • blk.17.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.17.attn_norm.weight
    F32
    [4096]
  • blk.17.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.17.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.17.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.17.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.17.post_attention_norm.weight
    F32
    [4096]
  • blk.17.ssm_a
    F32
    [32]
  • blk.17.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.17.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.17.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.17.ssm_dt
    F32
    [32]
  • blk.17.ssm_norm.weight
    F32
    [128]
  • blk.17.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.18
  • blk.18.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.18.attn_norm.weight
    F32
    [4096]
  • blk.18.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.18.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.18.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.18.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.18.post_attention_norm.weight
    F32
    [4096]
  • blk.18.ssm_a
    F32
    [32]
  • blk.18.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.18.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.18.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.18.ssm_dt
    F32
    [32]
  • blk.18.ssm_norm.weight
    F32
    [128]
  • blk.18.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.19
  • blk.19.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.19.attn_k_norm.weight
    F32
    [256]
  • blk.19.attn_norm.weight
    F32
    [4096]
  • blk.19.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.19.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.19.attn_q_norm.weight
    F32
    [256]
  • blk.19.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.19.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.19.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.19.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.19.post_attention_norm.weight
    F32
    [4096]
  • blk.20
  • blk.20.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.20.attn_norm.weight
    F32
    [4096]
  • blk.20.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.20.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.20.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.20.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.20.post_attention_norm.weight
    F32
    [4096]
  • blk.20.ssm_a
    F32
    [32]
  • blk.20.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.20.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.20.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.20.ssm_dt
    F32
    [32]
  • blk.20.ssm_norm.weight
    F32
    [128]
  • blk.20.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.21
  • blk.21.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.21.attn_norm.weight
    F32
    [4096]
  • blk.21.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.21.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.21.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.21.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.21.post_attention_norm.weight
    F32
    [4096]
  • blk.21.ssm_a
    F32
    [32]
  • blk.21.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.21.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.21.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.21.ssm_dt
    F32
    [32]
  • blk.21.ssm_norm.weight
    F32
    [128]
  • blk.21.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.22
  • blk.22.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.22.attn_norm.weight
    F32
    [4096]
  • blk.22.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.22.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.22.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.22.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.22.post_attention_norm.weight
    F32
    [4096]
  • blk.22.ssm_a
    F32
    [32]
  • blk.22.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.22.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.22.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.22.ssm_dt
    F32
    [32]
  • blk.22.ssm_norm.weight
    F32
    [128]
  • blk.22.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.23
  • blk.23.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.23.attn_k_norm.weight
    F32
    [256]
  • blk.23.attn_norm.weight
    F32
    [4096]
  • blk.23.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.23.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.23.attn_q_norm.weight
    F32
    [256]
  • blk.23.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.23.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.23.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.23.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.23.post_attention_norm.weight
    F32
    [4096]
  • blk.24
  • blk.24.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.24.attn_norm.weight
    F32
    [4096]
  • blk.24.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.24.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.24.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.24.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.24.post_attention_norm.weight
    F32
    [4096]
  • blk.24.ssm_a
    F32
    [32]
  • blk.24.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.24.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.24.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.24.ssm_dt
    F32
    [32]
  • blk.24.ssm_norm.weight
    F32
    [128]
  • blk.24.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.25
  • blk.25.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.25.attn_norm.weight
    F32
    [4096]
  • blk.25.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.25.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.25.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.25.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.25.post_attention_norm.weight
    F32
    [4096]
  • blk.25.ssm_a
    F32
    [32]
  • blk.25.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.25.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.25.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.25.ssm_dt
    F32
    [32]
  • blk.25.ssm_norm.weight
    F32
    [128]
  • blk.25.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.26
  • blk.26.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.26.attn_norm.weight
    F32
    [4096]
  • blk.26.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.26.ffn_down.weight
    Q4_K
    [12288, 4096]
  • blk.26.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.26.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.26.post_attention_norm.weight
    F32
    [4096]
  • blk.26.ssm_a
    F32
    [32]
  • blk.26.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.26.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.26.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.26.ssm_dt
    F32
    [32]
  • blk.26.ssm_norm.weight
    F32
    [128]
  • blk.26.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.27
  • blk.27.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.27.attn_k_norm.weight
    F32
    [256]
  • blk.27.attn_norm.weight
    F32
    [4096]
  • blk.27.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.27.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.27.attn_q_norm.weight
    F32
    [256]
  • blk.27.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.27.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.27.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.27.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.27.post_attention_norm.weight
    F32
    [4096]
  • blk.28
  • blk.28.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.28.attn_norm.weight
    F32
    [4096]
  • blk.28.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.28.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.28.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.28.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.28.post_attention_norm.weight
    F32
    [4096]
  • blk.28.ssm_a
    F32
    [32]
  • blk.28.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.28.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.28.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.28.ssm_dt
    F32
    [32]
  • blk.28.ssm_norm.weight
    F32
    [128]
  • blk.28.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.29
  • blk.29.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.29.attn_norm.weight
    F32
    [4096]
  • blk.29.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.29.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.29.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.29.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.29.post_attention_norm.weight
    F32
    [4096]
  • blk.29.ssm_a
    F32
    [32]
  • blk.29.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.29.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.29.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.29.ssm_dt
    F32
    [32]
  • blk.29.ssm_norm.weight
    F32
    [128]
  • blk.29.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.30
  • blk.30.attn_gate.weight
    Q4_K
    [4096, 4096]
  • blk.30.attn_norm.weight
    F32
    [4096]
  • blk.30.attn_qkv.weight
    Q4_K
    [4096, 8192]
  • blk.30.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.30.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.30.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.30.post_attention_norm.weight
    F32
    [4096]
  • blk.30.ssm_a
    F32
    [32]
  • blk.30.ssm_alpha.weight
    Q4_K
    [4096, 32]
  • blk.30.ssm_beta.weight
    Q4_K
    [4096, 32]
  • blk.30.ssm_conv1d.weight
    F32
    [4, 8192]
  • blk.30.ssm_dt
    F32
    [32]
  • blk.30.ssm_norm.weight
    F32
    [128]
  • blk.30.ssm_out.weight
    Q4_K
    [4096, 4096]
  • blk.31
  • blk.31.attn_k.weight
    Q4_K
    [4096, 1024]
  • blk.31.attn_k_norm.weight
    F32
    [256]
  • blk.31.attn_norm.weight
    F32
    [4096]
  • blk.31.attn_output.weight
    Q4_K
    [4096, 4096]
  • blk.31.attn_q.weight
    Q4_K
    [4096, 8192]
  • blk.31.attn_q_norm.weight
    F32
    [256]
  • blk.31.attn_v.weight
    Q6_K
    [4096, 1024]
  • blk.31.ffn_down.weight
    Q6_K
    [12288, 4096]
  • blk.31.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • blk.31.ffn_up.weight
    Q4_K
    [4096, 12288]
  • blk.31.post_attention_norm.weight
    F32
    [4096]
  • mtp.fc.weight
    Q4_K
    [8192, 4096]
  • mtp.layers.0.attn_k.weight
    Q4_K
    [4096, 1024]
  • mtp.layers.0.attn_k_norm.weight
    F32
    [256]
  • mtp.layers.0.attn_norm.weight
    F32
    [4096]
  • mtp.layers.0.attn_output.weight
    Q4_K
    [4096, 4096]
  • mtp.layers.0.attn_q.weight
    Q4_K
    [4096, 8192]
  • mtp.layers.0.attn_q_norm.weight
    F32
    [256]
  • mtp.layers.0.attn_v.weight
    Q6_K
    [4096, 1024]
  • mtp.layers.0.ffn_down.weight
    Q6_K
    [12288, 4096]
  • mtp.layers.0.ffn_gate.weight
    Q4_K
    [4096, 12288]
  • mtp.layers.0.ffn_up.weight
    Q4_K
    [4096, 12288]
  • mtp.layers.0.post_attention_norm.weight
    F32
    [4096]
  • mtp.norm.weight
    F32
    [4096]
  • mtp.pre_fc_norm_embedding.weight
    F32
    [4096]
  • mtp.pre_fc_norm_hidden.weight
    F32
    [4096]
  • output.weight
    Q6_K
    [4096, 248320]
  • v.blk.0
  • v.blk.0.attn_k.bias
    F32
    [1152]
  • v.blk.0.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.0.attn_out.bias
    F32
    [1152]
  • v.blk.0.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.0.attn_q.bias
    F32
    [1152]
  • v.blk.0.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.0.attn_v.bias
    F32
    [1152]
  • v.blk.0.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.0.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.0.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.0.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.0.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.0.norm1.bias
    F32
    [1152]
  • v.blk.0.norm1.weight
    F32
    [1152]
  • v.blk.0.norm2.bias
    F32
    [1152]
  • v.blk.0.norm2.weight
    F32
    [1152]
  • v.blk.1
  • v.blk.1.attn_k.bias
    F32
    [1152]
  • v.blk.1.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.1.attn_out.bias
    F32
    [1152]
  • v.blk.1.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.1.attn_q.bias
    F32
    [1152]
  • v.blk.1.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.1.attn_v.bias
    F32
    [1152]
  • v.blk.1.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.1.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.1.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.1.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.1.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.1.norm1.bias
    F32
    [1152]
  • v.blk.1.norm1.weight
    F32
    [1152]
  • v.blk.1.norm2.bias
    F32
    [1152]
  • v.blk.1.norm2.weight
    F32
    [1152]
  • v.blk.2
  • v.blk.2.attn_k.bias
    F32
    [1152]
  • v.blk.2.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.2.attn_out.bias
    F32
    [1152]
  • v.blk.2.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.2.attn_q.bias
    F32
    [1152]
  • v.blk.2.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.2.attn_v.bias
    F32
    [1152]
  • v.blk.2.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.2.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.2.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.2.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.2.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.2.norm1.bias
    F32
    [1152]
  • v.blk.2.norm1.weight
    F32
    [1152]
  • v.blk.2.norm2.bias
    F32
    [1152]
  • v.blk.2.norm2.weight
    F32
    [1152]
  • v.blk.3
  • v.blk.3.attn_k.bias
    F32
    [1152]
  • v.blk.3.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.3.attn_out.bias
    F32
    [1152]
  • v.blk.3.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.3.attn_q.bias
    F32
    [1152]
  • v.blk.3.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.3.attn_v.bias
    F32
    [1152]
  • v.blk.3.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.3.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.3.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.3.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.3.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.3.norm1.bias
    F32
    [1152]
  • v.blk.3.norm1.weight
    F32
    [1152]
  • v.blk.3.norm2.bias
    F32
    [1152]
  • v.blk.3.norm2.weight
    F32
    [1152]
  • v.blk.4
  • v.blk.4.attn_k.bias
    F32
    [1152]
  • v.blk.4.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.4.attn_out.bias
    F32
    [1152]
  • v.blk.4.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.4.attn_q.bias
    F32
    [1152]
  • v.blk.4.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.4.attn_v.bias
    F32
    [1152]
  • v.blk.4.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.4.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.4.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.4.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.4.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.4.norm1.bias
    F32
    [1152]
  • v.blk.4.norm1.weight
    F32
    [1152]
  • v.blk.4.norm2.bias
    F32
    [1152]
  • v.blk.4.norm2.weight
    F32
    [1152]
  • v.blk.5
  • v.blk.5.attn_k.bias
    F32
    [1152]
  • v.blk.5.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.5.attn_out.bias
    F32
    [1152]
  • v.blk.5.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.5.attn_q.bias
    F32
    [1152]
  • v.blk.5.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.5.attn_v.bias
    F32
    [1152]
  • v.blk.5.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.5.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.5.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.5.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.5.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.5.norm1.bias
    F32
    [1152]
  • v.blk.5.norm1.weight
    F32
    [1152]
  • v.blk.5.norm2.bias
    F32
    [1152]
  • v.blk.5.norm2.weight
    F32
    [1152]
  • v.blk.6
  • v.blk.6.attn_k.bias
    F32
    [1152]
  • v.blk.6.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.6.attn_out.bias
    F32
    [1152]
  • v.blk.6.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.6.attn_q.bias
    F32
    [1152]
  • v.blk.6.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.6.attn_v.bias
    F32
    [1152]
  • v.blk.6.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.6.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.6.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.6.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.6.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.6.norm1.bias
    F32
    [1152]
  • v.blk.6.norm1.weight
    F32
    [1152]
  • v.blk.6.norm2.bias
    F32
    [1152]
  • v.blk.6.norm2.weight
    F32
    [1152]
  • v.blk.7
  • v.blk.7.attn_k.bias
    F32
    [1152]
  • v.blk.7.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.7.attn_out.bias
    F32
    [1152]
  • v.blk.7.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.7.attn_q.bias
    F32
    [1152]
  • v.blk.7.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.7.attn_v.bias
    F32
    [1152]
  • v.blk.7.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.7.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.7.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.7.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.7.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.7.norm1.bias
    F32
    [1152]
  • v.blk.7.norm1.weight
    F32
    [1152]
  • v.blk.7.norm2.bias
    F32
    [1152]
  • v.blk.7.norm2.weight
    F32
    [1152]
  • v.blk.8
  • v.blk.8.attn_k.bias
    F32
    [1152]
  • v.blk.8.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.8.attn_out.bias
    F32
    [1152]
  • v.blk.8.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.8.attn_q.bias
    F32
    [1152]
  • v.blk.8.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.8.attn_v.bias
    F32
    [1152]
  • v.blk.8.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.8.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.8.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.8.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.8.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.8.norm1.bias
    F32
    [1152]
  • v.blk.8.norm1.weight
    F32
    [1152]
  • v.blk.8.norm2.bias
    F32
    [1152]
  • v.blk.8.norm2.weight
    F32
    [1152]
  • v.blk.9
  • v.blk.9.attn_k.bias
    F32
    [1152]
  • v.blk.9.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.9.attn_out.bias
    F32
    [1152]
  • v.blk.9.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.9.attn_q.bias
    F32
    [1152]
  • v.blk.9.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.9.attn_v.bias
    F32
    [1152]
  • v.blk.9.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.9.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.9.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.9.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.9.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.9.norm1.bias
    F32
    [1152]
  • v.blk.9.norm1.weight
    F32
    [1152]
  • v.blk.9.norm2.bias
    F32
    [1152]
  • v.blk.9.norm2.weight
    F32
    [1152]
  • v.blk.10
  • v.blk.10.attn_k.bias
    F32
    [1152]
  • v.blk.10.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.10.attn_out.bias
    F32
    [1152]
  • v.blk.10.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.10.attn_q.bias
    F32
    [1152]
  • v.blk.10.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.10.attn_v.bias
    F32
    [1152]
  • v.blk.10.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.10.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.10.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.10.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.10.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.10.norm1.bias
    F32
    [1152]
  • v.blk.10.norm1.weight
    F32
    [1152]
  • v.blk.10.norm2.bias
    F32
    [1152]
  • v.blk.10.norm2.weight
    F32
    [1152]
  • v.blk.11
  • v.blk.11.attn_k.bias
    F32
    [1152]
  • v.blk.11.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.11.attn_out.bias
    F32
    [1152]
  • v.blk.11.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.11.attn_q.bias
    F32
    [1152]
  • v.blk.11.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.11.attn_v.bias
    F32
    [1152]
  • v.blk.11.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.11.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.11.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.11.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.11.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.11.norm1.bias
    F32
    [1152]
  • v.blk.11.norm1.weight
    F32
    [1152]
  • v.blk.11.norm2.bias
    F32
    [1152]
  • v.blk.11.norm2.weight
    F32
    [1152]
  • v.blk.12
  • v.blk.12.attn_k.bias
    F32
    [1152]
  • v.blk.12.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.12.attn_out.bias
    F32
    [1152]
  • v.blk.12.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.12.attn_q.bias
    F32
    [1152]
  • v.blk.12.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.12.attn_v.bias
    F32
    [1152]
  • v.blk.12.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.12.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.12.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.12.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.12.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.12.norm1.bias
    F32
    [1152]
  • v.blk.12.norm1.weight
    F32
    [1152]
  • v.blk.12.norm2.bias
    F32
    [1152]
  • v.blk.12.norm2.weight
    F32
    [1152]
  • v.blk.13
  • v.blk.13.attn_k.bias
    F32
    [1152]
  • v.blk.13.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.13.attn_out.bias
    F32
    [1152]
  • v.blk.13.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.13.attn_q.bias
    F32
    [1152]
  • v.blk.13.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.13.attn_v.bias
    F32
    [1152]
  • v.blk.13.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.13.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.13.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.13.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.13.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.13.norm1.bias
    F32
    [1152]
  • v.blk.13.norm1.weight
    F32
    [1152]
  • v.blk.13.norm2.bias
    F32
    [1152]
  • v.blk.13.norm2.weight
    F32
    [1152]
  • v.blk.14
  • v.blk.14.attn_k.bias
    F32
    [1152]
  • v.blk.14.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.14.attn_out.bias
    F32
    [1152]
  • v.blk.14.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.14.attn_q.bias
    F32
    [1152]
  • v.blk.14.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.14.attn_v.bias
    F32
    [1152]
  • v.blk.14.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.14.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.14.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.14.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.14.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.14.norm1.bias
    F32
    [1152]
  • v.blk.14.norm1.weight
    F32
    [1152]
  • v.blk.14.norm2.bias
    F32
    [1152]
  • v.blk.14.norm2.weight
    F32
    [1152]
  • v.blk.15
  • v.blk.15.attn_k.bias
    F32
    [1152]
  • v.blk.15.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.15.attn_out.bias
    F32
    [1152]
  • v.blk.15.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.15.attn_q.bias
    F32
    [1152]
  • v.blk.15.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.15.attn_v.bias
    F32
    [1152]
  • v.blk.15.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.15.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.15.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.15.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.15.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.15.norm1.bias
    F32
    [1152]
  • v.blk.15.norm1.weight
    F32
    [1152]
  • v.blk.15.norm2.bias
    F32
    [1152]
  • v.blk.15.norm2.weight
    F32
    [1152]
  • v.blk.16
  • v.blk.16.attn_k.bias
    F32
    [1152]
  • v.blk.16.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.16.attn_out.bias
    F32
    [1152]
  • v.blk.16.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.16.attn_q.bias
    F32
    [1152]
  • v.blk.16.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.16.attn_v.bias
    F32
    [1152]
  • v.blk.16.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.16.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.16.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.16.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.16.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.16.norm1.bias
    F32
    [1152]
  • v.blk.16.norm1.weight
    F32
    [1152]
  • v.blk.16.norm2.bias
    F32
    [1152]
  • v.blk.16.norm2.weight
    F32
    [1152]
  • v.blk.17
  • v.blk.17.attn_k.bias
    F32
    [1152]
  • v.blk.17.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.17.attn_out.bias
    F32
    [1152]
  • v.blk.17.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.17.attn_q.bias
    F32
    [1152]
  • v.blk.17.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.17.attn_v.bias
    F32
    [1152]
  • v.blk.17.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.17.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.17.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.17.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.17.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.17.norm1.bias
    F32
    [1152]
  • v.blk.17.norm1.weight
    F32
    [1152]
  • v.blk.17.norm2.bias
    F32
    [1152]
  • v.blk.17.norm2.weight
    F32
    [1152]
  • v.blk.18
  • v.blk.18.attn_k.bias
    F32
    [1152]
  • v.blk.18.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.18.attn_out.bias
    F32
    [1152]
  • v.blk.18.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.18.attn_q.bias
    F32
    [1152]
  • v.blk.18.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.18.attn_v.bias
    F32
    [1152]
  • v.blk.18.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.18.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.18.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.18.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.18.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.18.norm1.bias
    F32
    [1152]
  • v.blk.18.norm1.weight
    F32
    [1152]
  • v.blk.18.norm2.bias
    F32
    [1152]
  • v.blk.18.norm2.weight
    F32
    [1152]
  • v.blk.19
  • v.blk.19.attn_k.bias
    F32
    [1152]
  • v.blk.19.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.19.attn_out.bias
    F32
    [1152]
  • v.blk.19.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.19.attn_q.bias
    F32
    [1152]
  • v.blk.19.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.19.attn_v.bias
    F32
    [1152]
  • v.blk.19.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.19.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.19.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.19.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.19.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.19.norm1.bias
    F32
    [1152]
  • v.blk.19.norm1.weight
    F32
    [1152]
  • v.blk.19.norm2.bias
    F32
    [1152]
  • v.blk.19.norm2.weight
    F32
    [1152]
  • v.blk.20
  • v.blk.20.attn_k.bias
    F32
    [1152]
  • v.blk.20.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.20.attn_out.bias
    F32
    [1152]
  • v.blk.20.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.20.attn_q.bias
    F32
    [1152]
  • v.blk.20.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.20.attn_v.bias
    F32
    [1152]
  • v.blk.20.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.20.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.20.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.20.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.20.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.20.norm1.bias
    F32
    [1152]
  • v.blk.20.norm1.weight
    F32
    [1152]
  • v.blk.20.norm2.bias
    F32
    [1152]
  • v.blk.20.norm2.weight
    F32
    [1152]
  • v.blk.21
  • v.blk.21.attn_k.bias
    F32
    [1152]
  • v.blk.21.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.21.attn_out.bias
    F32
    [1152]
  • v.blk.21.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.21.attn_q.bias
    F32
    [1152]
  • v.blk.21.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.21.attn_v.bias
    F32
    [1152]
  • v.blk.21.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.21.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.21.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.21.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.21.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.21.norm1.bias
    F32
    [1152]
  • v.blk.21.norm1.weight
    F32
    [1152]
  • v.blk.21.norm2.bias
    F32
    [1152]
  • v.blk.21.norm2.weight
    F32
    [1152]
  • v.blk.22
  • v.blk.22.attn_k.bias
    F32
    [1152]
  • v.blk.22.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.22.attn_out.bias
    F32
    [1152]
  • v.blk.22.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.22.attn_q.bias
    F32
    [1152]
  • v.blk.22.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.22.attn_v.bias
    F32
    [1152]
  • v.blk.22.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.22.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.22.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.22.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.22.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.22.norm1.bias
    F32
    [1152]
  • v.blk.22.norm1.weight
    F32
    [1152]
  • v.blk.22.norm2.bias
    F32
    [1152]
  • v.blk.22.norm2.weight
    F32
    [1152]
  • v.blk.23
  • v.blk.23.attn_k.bias
    F32
    [1152]
  • v.blk.23.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.23.attn_out.bias
    F32
    [1152]
  • v.blk.23.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.23.attn_q.bias
    F32
    [1152]
  • v.blk.23.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.23.attn_v.bias
    F32
    [1152]
  • v.blk.23.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.23.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.23.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.23.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.23.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.23.norm1.bias
    F32
    [1152]
  • v.blk.23.norm1.weight
    F32
    [1152]
  • v.blk.23.norm2.bias
    F32
    [1152]
  • v.blk.23.norm2.weight
    F32
    [1152]
  • v.blk.24
  • v.blk.24.attn_k.bias
    F32
    [1152]
  • v.blk.24.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.24.attn_out.bias
    F32
    [1152]
  • v.blk.24.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.24.attn_q.bias
    F32
    [1152]
  • v.blk.24.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.24.attn_v.bias
    F32
    [1152]
  • v.blk.24.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.24.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.24.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.24.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.24.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.24.norm1.bias
    F32
    [1152]
  • v.blk.24.norm1.weight
    F32
    [1152]
  • v.blk.24.norm2.bias
    F32
    [1152]
  • v.blk.24.norm2.weight
    F32
    [1152]
  • v.blk.25
  • v.blk.25.attn_k.bias
    F32
    [1152]
  • v.blk.25.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.25.attn_out.bias
    F32
    [1152]
  • v.blk.25.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.25.attn_q.bias
    F32
    [1152]
  • v.blk.25.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.25.attn_v.bias
    F32
    [1152]
  • v.blk.25.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.25.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.25.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.25.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.25.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.25.norm1.bias
    F32
    [1152]
  • v.blk.25.norm1.weight
    F32
    [1152]
  • v.blk.25.norm2.bias
    F32
    [1152]
  • v.blk.25.norm2.weight
    F32
    [1152]
  • v.blk.26
  • v.blk.26.attn_k.bias
    F32
    [1152]
  • v.blk.26.attn_k.weight
    F16
    [1152, 1152]
  • v.blk.26.attn_out.bias
    F32
    [1152]
  • v.blk.26.attn_out.weight
    F16
    [1152, 1152]
  • v.blk.26.attn_q.bias
    F32
    [1152]
  • v.blk.26.attn_q.weight
    F16
    [1152, 1152]
  • v.blk.26.attn_v.bias
    F32
    [1152]
  • v.blk.26.attn_v.weight
    F16
    [1152, 1152]
  • v.blk.26.mlp.linear_fc1.bias
    F32
    [4304]
  • v.blk.26.mlp.linear_fc1.weight
    F16
    [1152, 4304]
  • v.blk.26.mlp.linear_fc2.bias
    F32
    [1152]
  • v.blk.26.mlp.linear_fc2.weight
    F16
    [4304, 1152]
  • v.blk.26.norm1.bias
    F32
    [1152]
  • v.blk.26.norm1.weight
    F32
    [1152]
  • v.blk.26.norm2.bias
    F32
    [1152]
  • v.blk.26.norm2.weight
    F32
    [1152]
  • v.merger.linear_fc1.bias
    F32
    [4608]
  • v.merger.linear_fc1.weight
    F16
    [4608, 4608]
  • v.merger.linear_fc2.bias
    F32
    [4096]
  • v.merger.linear_fc2.weight
    F16
    [4608, 4096]
  • v.merger.norm.bias
    F32
    [1152]
  • v.merger.norm.weight
    F32
    [1152]
  • v.patch_embed.bias
    F32
    [1152]
  • v.patch_embed.weight
    F16
    [16, 16, 2, 3456]
  • v.pos_embed.weight
    F16
    [1152, 2304]
  • output_norm.weight
    F32
    [4096]