Willem27/
Qwen3-VL-8B-Thinking:latest

593 1 month ago

... /
model
f104d6d62c51 · 8.7GB
    Metadata
  • general.architecture
    Qwen3-VL-8B-Thinking
  • attention_bias
    false
  • bos_token_id
    151643
  • eos_token_id
    151645
  • head_dim
    128
  • hidden_act
    silu
  • hidden_size
    4096
  • intermediate_size
    12288
  • max_position_embeddings
    262144
  • num_attention_heads
    32
  • num_hidden_layers
    36
  • num_key_value_heads
    8
  • rms_norm_eps
    1e-06
  • rope_theta
    5e+06
  • tensor_count
    399
  • tie_word_embeddings
    false
  • version
    3
  • vocab_size
    151936
    • Tensor
  • embed_tokens.weight
    Q8_0
    [4096, 151936]
  • layers.0.input_layernorm.weight
    F32
    [4096]
  • layers.0.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.0.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.0.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.0.post_attention_layernorm.weight
    F32
    [4096]
  • layers.0.self_attn.k_norm.weight
    F32
    [128]
  • layers.0.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.0.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.0.self_attn.q_norm.weight
    F32
    [128]
  • layers.0.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.0.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.1.input_layernorm.weight
    F32
    [4096]
  • layers.1.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.1.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.1.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.1.post_attention_layernorm.weight
    F32
    [4096]
  • layers.1.self_attn.k_norm.weight
    F32
    [128]
  • layers.1.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.1.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.1.self_attn.q_norm.weight
    F32
    [128]
  • layers.1.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.1.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.2.input_layernorm.weight
    F32
    [4096]
  • layers.2.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.2.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.2.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.2.post_attention_layernorm.weight
    F32
    [4096]
  • layers.2.self_attn.k_norm.weight
    F32
    [128]
  • layers.2.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.2.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.2.self_attn.q_norm.weight
    F32
    [128]
  • layers.2.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.2.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.3.input_layernorm.weight
    F32
    [4096]
  • layers.3.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.3.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.3.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.3.post_attention_layernorm.weight
    F32
    [4096]
  • layers.3.self_attn.k_norm.weight
    F32
    [128]
  • layers.3.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.3.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.3.self_attn.q_norm.weight
    F32
    [128]
  • layers.3.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.3.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.4.input_layernorm.weight
    F32
    [4096]
  • layers.4.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.4.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.4.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.4.post_attention_layernorm.weight
    F32
    [4096]
  • layers.4.self_attn.k_norm.weight
    F32
    [128]
  • layers.4.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.4.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.4.self_attn.q_norm.weight
    F32
    [128]
  • layers.4.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.4.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.5.input_layernorm.weight
    F32
    [4096]
  • layers.5.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.5.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.5.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.5.post_attention_layernorm.weight
    F32
    [4096]
  • layers.5.self_attn.k_norm.weight
    F32
    [128]
  • layers.5.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.5.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.5.self_attn.q_norm.weight
    F32
    [128]
  • layers.5.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.5.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.6.input_layernorm.weight
    F32
    [4096]
  • layers.6.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.6.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.6.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.6.post_attention_layernorm.weight
    F32
    [4096]
  • layers.6.self_attn.k_norm.weight
    F32
    [128]
  • layers.6.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.6.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.6.self_attn.q_norm.weight
    F32
    [128]
  • layers.6.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.6.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.7.input_layernorm.weight
    F32
    [4096]
  • layers.7.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.7.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.7.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.7.post_attention_layernorm.weight
    F32
    [4096]
  • layers.7.self_attn.k_norm.weight
    F32
    [128]
  • layers.7.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.7.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.7.self_attn.q_norm.weight
    F32
    [128]
  • layers.7.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.7.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.8.input_layernorm.weight
    F32
    [4096]
  • layers.8.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.8.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.8.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.8.post_attention_layernorm.weight
    F32
    [4096]
  • layers.8.self_attn.k_norm.weight
    F32
    [128]
  • layers.8.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.8.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.8.self_attn.q_norm.weight
    F32
    [128]
  • layers.8.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.8.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.9.input_layernorm.weight
    F32
    [4096]
  • layers.9.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.9.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.9.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.9.post_attention_layernorm.weight
    F32
    [4096]
  • layers.9.self_attn.k_norm.weight
    F32
    [128]
  • layers.9.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.9.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.9.self_attn.q_norm.weight
    F32
    [128]
  • layers.9.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.9.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.10.input_layernorm.weight
    F32
    [4096]
  • layers.10.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.10.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.10.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.10.post_attention_layernorm.weight
    F32
    [4096]
  • layers.10.self_attn.k_norm.weight
    F32
    [128]
  • layers.10.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.10.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.10.self_attn.q_norm.weight
    F32
    [128]
  • layers.10.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.10.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.11.input_layernorm.weight
    F32
    [4096]
  • layers.11.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.11.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.11.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.11.post_attention_layernorm.weight
    F32
    [4096]
  • layers.11.self_attn.k_norm.weight
    F32
    [128]
  • layers.11.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.11.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.11.self_attn.q_norm.weight
    F32
    [128]
  • layers.11.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.11.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.12.input_layernorm.weight
    F32
    [4096]
  • layers.12.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.12.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.12.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.12.post_attention_layernorm.weight
    F32
    [4096]
  • layers.12.self_attn.k_norm.weight
    F32
    [128]
  • layers.12.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.12.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.12.self_attn.q_norm.weight
    F32
    [128]
  • layers.12.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.12.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.13.input_layernorm.weight
    F32
    [4096]
  • layers.13.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.13.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.13.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.13.post_attention_layernorm.weight
    F32
    [4096]
  • layers.13.self_attn.k_norm.weight
    F32
    [128]
  • layers.13.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.13.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.13.self_attn.q_norm.weight
    F32
    [128]
  • layers.13.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.13.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.14.input_layernorm.weight
    F32
    [4096]
  • layers.14.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.14.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.14.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.14.post_attention_layernorm.weight
    F32
    [4096]
  • layers.14.self_attn.k_norm.weight
    F32
    [128]
  • layers.14.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.14.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.14.self_attn.q_norm.weight
    F32
    [128]
  • layers.14.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.14.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.15.input_layernorm.weight
    F32
    [4096]
  • layers.15.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.15.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.15.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.15.post_attention_layernorm.weight
    F32
    [4096]
  • layers.15.self_attn.k_norm.weight
    F32
    [128]
  • layers.15.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.15.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.15.self_attn.q_norm.weight
    F32
    [128]
  • layers.15.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.15.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.16.input_layernorm.weight
    F32
    [4096]
  • layers.16.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.16.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.16.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.16.post_attention_layernorm.weight
    F32
    [4096]
  • layers.16.self_attn.k_norm.weight
    F32
    [128]
  • layers.16.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.16.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.16.self_attn.q_norm.weight
    F32
    [128]
  • layers.16.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.16.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.17.input_layernorm.weight
    F32
    [4096]
  • layers.17.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.17.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.17.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.17.post_attention_layernorm.weight
    F32
    [4096]
  • layers.17.self_attn.k_norm.weight
    F32
    [128]
  • layers.17.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.17.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.17.self_attn.q_norm.weight
    F32
    [128]
  • layers.17.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.17.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.18.input_layernorm.weight
    F32
    [4096]
  • layers.18.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.18.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.18.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.18.post_attention_layernorm.weight
    F32
    [4096]
  • layers.18.self_attn.k_norm.weight
    F32
    [128]
  • layers.18.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.18.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.18.self_attn.q_norm.weight
    F32
    [128]
  • layers.18.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.18.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.19.input_layernorm.weight
    F32
    [4096]
  • layers.19.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.19.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.19.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.19.post_attention_layernorm.weight
    F32
    [4096]
  • layers.19.self_attn.k_norm.weight
    F32
    [128]
  • layers.19.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.19.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.19.self_attn.q_norm.weight
    F32
    [128]
  • layers.19.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.19.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.20.input_layernorm.weight
    F32
    [4096]
  • layers.20.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.20.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.20.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.20.post_attention_layernorm.weight
    F32
    [4096]
  • layers.20.self_attn.k_norm.weight
    F32
    [128]
  • layers.20.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.20.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.20.self_attn.q_norm.weight
    F32
    [128]
  • layers.20.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.20.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.21.input_layernorm.weight
    F32
    [4096]
  • layers.21.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.21.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.21.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.21.post_attention_layernorm.weight
    F32
    [4096]
  • layers.21.self_attn.k_norm.weight
    F32
    [128]
  • layers.21.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.21.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.21.self_attn.q_norm.weight
    F32
    [128]
  • layers.21.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.21.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.22.input_layernorm.weight
    F32
    [4096]
  • layers.22.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.22.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.22.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.22.post_attention_layernorm.weight
    F32
    [4096]
  • layers.22.self_attn.k_norm.weight
    F32
    [128]
  • layers.22.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.22.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.22.self_attn.q_norm.weight
    F32
    [128]
  • layers.22.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.22.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.23.input_layernorm.weight
    F32
    [4096]
  • layers.23.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.23.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.23.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.23.post_attention_layernorm.weight
    F32
    [4096]
  • layers.23.self_attn.k_norm.weight
    F32
    [128]
  • layers.23.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.23.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.23.self_attn.q_norm.weight
    F32
    [128]
  • layers.23.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.23.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.24.input_layernorm.weight
    F32
    [4096]
  • layers.24.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.24.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.24.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.24.post_attention_layernorm.weight
    F32
    [4096]
  • layers.24.self_attn.k_norm.weight
    F32
    [128]
  • layers.24.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.24.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.24.self_attn.q_norm.weight
    F32
    [128]
  • layers.24.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.24.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.25.input_layernorm.weight
    F32
    [4096]
  • layers.25.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.25.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.25.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.25.post_attention_layernorm.weight
    F32
    [4096]
  • layers.25.self_attn.k_norm.weight
    F32
    [128]
  • layers.25.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.25.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.25.self_attn.q_norm.weight
    F32
    [128]
  • layers.25.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.25.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.26.input_layernorm.weight
    F32
    [4096]
  • layers.26.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.26.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.26.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.26.post_attention_layernorm.weight
    F32
    [4096]
  • layers.26.self_attn.k_norm.weight
    F32
    [128]
  • layers.26.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.26.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.26.self_attn.q_norm.weight
    F32
    [128]
  • layers.26.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.26.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.27.input_layernorm.weight
    F32
    [4096]
  • layers.27.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.27.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.27.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.27.post_attention_layernorm.weight
    F32
    [4096]
  • layers.27.self_attn.k_norm.weight
    F32
    [128]
  • layers.27.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.27.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.27.self_attn.q_norm.weight
    F32
    [128]
  • layers.27.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.27.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.28.input_layernorm.weight
    F32
    [4096]
  • layers.28.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.28.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.28.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.28.post_attention_layernorm.weight
    F32
    [4096]
  • layers.28.self_attn.k_norm.weight
    F32
    [128]
  • layers.28.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.28.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.28.self_attn.q_norm.weight
    F32
    [128]
  • layers.28.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.28.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.29.input_layernorm.weight
    F32
    [4096]
  • layers.29.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.29.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.29.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.29.post_attention_layernorm.weight
    F32
    [4096]
  • layers.29.self_attn.k_norm.weight
    F32
    [128]
  • layers.29.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.29.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.29.self_attn.q_norm.weight
    F32
    [128]
  • layers.29.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.29.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.30.input_layernorm.weight
    F32
    [4096]
  • layers.30.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.30.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.30.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.30.post_attention_layernorm.weight
    F32
    [4096]
  • layers.30.self_attn.k_norm.weight
    F32
    [128]
  • layers.30.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.30.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.30.self_attn.q_norm.weight
    F32
    [128]
  • layers.30.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.30.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.31.input_layernorm.weight
    F32
    [4096]
  • layers.31.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.31.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.31.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.31.post_attention_layernorm.weight
    F32
    [4096]
  • layers.31.self_attn.k_norm.weight
    F32
    [128]
  • layers.31.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.31.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.31.self_attn.q_norm.weight
    F32
    [128]
  • layers.31.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.31.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.32.input_layernorm.weight
    F32
    [4096]
  • layers.32.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.32.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.32.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.32.post_attention_layernorm.weight
    F32
    [4096]
  • layers.32.self_attn.k_norm.weight
    F32
    [128]
  • layers.32.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.32.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.32.self_attn.q_norm.weight
    F32
    [128]
  • layers.32.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.32.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.33.input_layernorm.weight
    F32
    [4096]
  • layers.33.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.33.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.33.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.33.post_attention_layernorm.weight
    F32
    [4096]
  • layers.33.self_attn.k_norm.weight
    F32
    [128]
  • layers.33.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.33.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.33.self_attn.q_norm.weight
    F32
    [128]
  • layers.33.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.33.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.34.input_layernorm.weight
    F32
    [4096]
  • layers.34.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.34.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.34.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.34.post_attention_layernorm.weight
    F32
    [4096]
  • layers.34.self_attn.k_norm.weight
    F32
    [128]
  • layers.34.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.34.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.34.self_attn.q_norm.weight
    F32
    [128]
  • layers.34.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.34.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • layers.35.input_layernorm.weight
    F32
    [4096]
  • layers.35.mlp.down_proj.weight
    Q8_0
    [12288, 4096]
  • layers.35.mlp.gate_proj.weight
    Q8_0
    [4096, 12288]
  • layers.35.mlp.up_proj.weight
    Q8_0
    [4096, 12288]
  • layers.35.post_attention_layernorm.weight
    F32
    [4096]
  • layers.35.self_attn.k_norm.weight
    F32
    [128]
  • layers.35.self_attn.k_proj.weight
    Q8_0
    [4096, 1024]
  • layers.35.self_attn.o_proj.weight
    Q8_0
    [4096, 4096]
  • layers.35.self_attn.q_norm.weight
    F32
    [128]
  • layers.35.self_attn.q_proj.weight
    Q8_0
    [4096, 4096]
  • layers.35.self_attn.v_proj.weight
    Q8_0
    [4096, 1024]
  • lm_head.weight
    Q8_0
    [4096, 151936]
  • norm.weight
    F32
    [4096]