feadxus/ z_image_turbo:bf16

647 1 month ago

tools thinking
... /
model
cb79c31c85f0 · 12GB
    Metadata
  • general.architecture
    lumina2
  • general.file_type
    F16
    • Tensor
  • cap_embedder.0.weight
    F32
    [2560]
  • cap_embedder.1.bias
    F32
    [3840]
  • cap_embedder.1.weight
    F16
    [2560, 3840]
  • cap_pad_token
    F32
    [3840, 1]
  • context_refiner.0.attention.k_norm.weight
    F32
    [128]
  • context_refiner.0.attention.out.weight
    F16
    [3840, 3840]
  • context_refiner.0.attention.q_norm.weight
    F32
    [128]
  • context_refiner.0.attention.qkv.weight
    BF16
    [3840, 11520]
  • context_refiner.0.attention_norm1.weight
    F32
    [3840]
  • context_refiner.0.attention_norm2.weight
    F32
    [3840]
  • context_refiner.0.feed_forward.w1.weight
    F16
    [3840, 10240]
  • context_refiner.0.feed_forward.w2.weight
    F16
    [10240, 3840]
  • context_refiner.0.feed_forward.w3.weight
    F16
    [3840, 10240]
  • context_refiner.0.ffn_norm1.weight
    F32
    [3840]
  • context_refiner.0.ffn_norm2.weight
    F32
    [3840]
  • context_refiner.1.attention.k_norm.weight
    F32
    [128]
  • context_refiner.1.attention.out.weight
    F16
    [3840, 3840]
  • context_refiner.1.attention.q_norm.weight
    F32
    [128]
  • context_refiner.1.attention.qkv.weight
    BF16
    [3840, 11520]
  • context_refiner.1.attention_norm1.weight
    F32
    [3840]
  • context_refiner.1.attention_norm2.weight
    F32
    [3840]
  • context_refiner.1.feed_forward.w1.weight
    F16
    [3840, 10240]
  • context_refiner.1.feed_forward.w2.weight
    F16
    [10240, 3840]
  • context_refiner.1.feed_forward.w3.weight
    F16
    [3840, 10240]
  • context_refiner.1.ffn_norm1.weight
    F32
    [3840]
  • context_refiner.1.ffn_norm2.weight
    F32
    [3840]
  • final_layer.adaLN_modulation.1.bias
    F32
    [3840]
  • final_layer.adaLN_modulation.1.weight
    F16
    [256, 3840]
  • final_layer.linear.bias
    F32
    [64]
  • final_layer.linear.weight
    F16
    [3840, 64]
  • layers.0.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.0.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.0.attention.k_norm.weight
    F32
    [128]
  • layers.0.attention.out.weight
    F16
    [3840, 3840]
  • layers.0.attention.q_norm.weight
    F32
    [128]
  • layers.0.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.0.attention_norm1.weight
    F32
    [3840]
  • layers.0.attention_norm2.weight
    F32
    [3840]
  • layers.0.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.0.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.0.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.0.ffn_norm1.weight
    F32
    [3840]
  • layers.0.ffn_norm2.weight
    F32
    [3840]
  • layers.1.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.1.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.1.attention.k_norm.weight
    F32
    [128]
  • layers.1.attention.out.weight
    F16
    [3840, 3840]
  • layers.1.attention.q_norm.weight
    F32
    [128]
  • layers.1.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.1.attention_norm1.weight
    F32
    [3840]
  • layers.1.attention_norm2.weight
    F32
    [3840]
  • layers.1.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.1.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.1.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.1.ffn_norm1.weight
    F32
    [3840]
  • layers.1.ffn_norm2.weight
    F32
    [3840]
  • layers.2.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.2.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.2.attention.k_norm.weight
    F32
    [128]
  • layers.2.attention.out.weight
    F16
    [3840, 3840]
  • layers.2.attention.q_norm.weight
    F32
    [128]
  • layers.2.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.2.attention_norm1.weight
    F32
    [3840]
  • layers.2.attention_norm2.weight
    F32
    [3840]
  • layers.2.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.2.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.2.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.2.ffn_norm1.weight
    F32
    [3840]
  • layers.2.ffn_norm2.weight
    F32
    [3840]
  • layers.3.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.3.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.3.attention.k_norm.weight
    F32
    [128]
  • layers.3.attention.out.weight
    F16
    [3840, 3840]
  • layers.3.attention.q_norm.weight
    F32
    [128]
  • layers.3.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.3.attention_norm1.weight
    F32
    [3840]
  • layers.3.attention_norm2.weight
    F32
    [3840]
  • layers.3.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.3.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.3.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.3.ffn_norm1.weight
    F32
    [3840]
  • layers.3.ffn_norm2.weight
    F32
    [3840]
  • layers.4.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.4.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.4.attention.k_norm.weight
    F32
    [128]
  • layers.4.attention.out.weight
    F16
    [3840, 3840]
  • layers.4.attention.q_norm.weight
    F32
    [128]
  • layers.4.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.4.attention_norm1.weight
    F32
    [3840]
  • layers.4.attention_norm2.weight
    F32
    [3840]
  • layers.4.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.4.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.4.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.4.ffn_norm1.weight
    F32
    [3840]
  • layers.4.ffn_norm2.weight
    F32
    [3840]
  • layers.5.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.5.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.5.attention.k_norm.weight
    F32
    [128]
  • layers.5.attention.out.weight
    F16
    [3840, 3840]
  • layers.5.attention.q_norm.weight
    F32
    [128]
  • layers.5.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.5.attention_norm1.weight
    F32
    [3840]
  • layers.5.attention_norm2.weight
    F32
    [3840]
  • layers.5.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.5.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.5.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.5.ffn_norm1.weight
    F32
    [3840]
  • layers.5.ffn_norm2.weight
    F32
    [3840]
  • layers.6.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.6.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.6.attention.k_norm.weight
    F32
    [128]
  • layers.6.attention.out.weight
    F16
    [3840, 3840]
  • layers.6.attention.q_norm.weight
    F32
    [128]
  • layers.6.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.6.attention_norm1.weight
    F32
    [3840]
  • layers.6.attention_norm2.weight
    F32
    [3840]
  • layers.6.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.6.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.6.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.6.ffn_norm1.weight
    F32
    [3840]
  • layers.6.ffn_norm2.weight
    F32
    [3840]
  • layers.7.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.7.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.7.attention.k_norm.weight
    F32
    [128]
  • layers.7.attention.out.weight
    F16
    [3840, 3840]
  • layers.7.attention.q_norm.weight
    F32
    [128]
  • layers.7.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.7.attention_norm1.weight
    F32
    [3840]
  • layers.7.attention_norm2.weight
    F32
    [3840]
  • layers.7.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.7.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.7.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.7.ffn_norm1.weight
    F32
    [3840]
  • layers.7.ffn_norm2.weight
    F32
    [3840]
  • layers.8.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.8.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.8.attention.k_norm.weight
    F32
    [128]
  • layers.8.attention.out.weight
    F16
    [3840, 3840]
  • layers.8.attention.q_norm.weight
    F32
    [128]
  • layers.8.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.8.attention_norm1.weight
    F32
    [3840]
  • layers.8.attention_norm2.weight
    F32
    [3840]
  • layers.8.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.8.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.8.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.8.ffn_norm1.weight
    F32
    [3840]
  • layers.8.ffn_norm2.weight
    F32
    [3840]
  • layers.9.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.9.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.9.attention.k_norm.weight
    F32
    [128]
  • layers.9.attention.out.weight
    F16
    [3840, 3840]
  • layers.9.attention.q_norm.weight
    F32
    [128]
  • layers.9.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.9.attention_norm1.weight
    F32
    [3840]
  • layers.9.attention_norm2.weight
    F32
    [3840]
  • layers.9.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.9.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.9.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.9.ffn_norm1.weight
    F32
    [3840]
  • layers.9.ffn_norm2.weight
    F32
    [3840]
  • layers.10.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.10.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.10.attention.k_norm.weight
    F32
    [128]
  • layers.10.attention.out.weight
    F16
    [3840, 3840]
  • layers.10.attention.q_norm.weight
    F32
    [128]
  • layers.10.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.10.attention_norm1.weight
    F32
    [3840]
  • layers.10.attention_norm2.weight
    F32
    [3840]
  • layers.10.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.10.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.10.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.10.ffn_norm1.weight
    F32
    [3840]
  • layers.10.ffn_norm2.weight
    F32
    [3840]
  • layers.11.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.11.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.11.attention.k_norm.weight
    F32
    [128]
  • layers.11.attention.out.weight
    F16
    [3840, 3840]
  • layers.11.attention.q_norm.weight
    F32
    [128]
  • layers.11.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.11.attention_norm1.weight
    F32
    [3840]
  • layers.11.attention_norm2.weight
    F32
    [3840]
  • layers.11.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.11.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.11.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.11.ffn_norm1.weight
    F32
    [3840]
  • layers.11.ffn_norm2.weight
    F32
    [3840]
  • layers.12.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.12.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.12.attention.k_norm.weight
    F32
    [128]
  • layers.12.attention.out.weight
    F16
    [3840, 3840]
  • layers.12.attention.q_norm.weight
    F32
    [128]
  • layers.12.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.12.attention_norm1.weight
    F32
    [3840]
  • layers.12.attention_norm2.weight
    F32
    [3840]
  • layers.12.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.12.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.12.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.12.ffn_norm1.weight
    F32
    [3840]
  • layers.12.ffn_norm2.weight
    F32
    [3840]
  • layers.13.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.13.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.13.attention.k_norm.weight
    F32
    [128]
  • layers.13.attention.out.weight
    F16
    [3840, 3840]
  • layers.13.attention.q_norm.weight
    F32
    [128]
  • layers.13.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.13.attention_norm1.weight
    F32
    [3840]
  • layers.13.attention_norm2.weight
    F32
    [3840]
  • layers.13.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.13.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.13.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.13.ffn_norm1.weight
    F32
    [3840]
  • layers.13.ffn_norm2.weight
    F32
    [3840]
  • layers.14.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.14.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.14.attention.k_norm.weight
    F32
    [128]
  • layers.14.attention.out.weight
    F16
    [3840, 3840]
  • layers.14.attention.q_norm.weight
    F32
    [128]
  • layers.14.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.14.attention_norm1.weight
    F32
    [3840]
  • layers.14.attention_norm2.weight
    F32
    [3840]
  • layers.14.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.14.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.14.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.14.ffn_norm1.weight
    F32
    [3840]
  • layers.14.ffn_norm2.weight
    F32
    [3840]
  • layers.15.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.15.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.15.attention.k_norm.weight
    F32
    [128]
  • layers.15.attention.out.weight
    F16
    [3840, 3840]
  • layers.15.attention.q_norm.weight
    F32
    [128]
  • layers.15.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.15.attention_norm1.weight
    F32
    [3840]
  • layers.15.attention_norm2.weight
    F32
    [3840]
  • layers.15.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.15.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.15.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.15.ffn_norm1.weight
    F32
    [3840]
  • layers.15.ffn_norm2.weight
    F32
    [3840]
  • layers.16.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.16.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.16.attention.k_norm.weight
    F32
    [128]
  • layers.16.attention.out.weight
    F16
    [3840, 3840]
  • layers.16.attention.q_norm.weight
    F32
    [128]
  • layers.16.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.16.attention_norm1.weight
    F32
    [3840]
  • layers.16.attention_norm2.weight
    F32
    [3840]
  • layers.16.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.16.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.16.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.16.ffn_norm1.weight
    F32
    [3840]
  • layers.16.ffn_norm2.weight
    F32
    [3840]
  • layers.17.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.17.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.17.attention.k_norm.weight
    F32
    [128]
  • layers.17.attention.out.weight
    F16
    [3840, 3840]
  • layers.17.attention.q_norm.weight
    F32
    [128]
  • layers.17.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.17.attention_norm1.weight
    F32
    [3840]
  • layers.17.attention_norm2.weight
    F32
    [3840]
  • layers.17.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.17.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.17.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.17.ffn_norm1.weight
    F32
    [3840]
  • layers.17.ffn_norm2.weight
    F32
    [3840]
  • layers.18.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.18.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.18.attention.k_norm.weight
    F32
    [128]
  • layers.18.attention.out.weight
    F16
    [3840, 3840]
  • layers.18.attention.q_norm.weight
    F32
    [128]
  • layers.18.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.18.attention_norm1.weight
    F32
    [3840]
  • layers.18.attention_norm2.weight
    F32
    [3840]
  • layers.18.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.18.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.18.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.18.ffn_norm1.weight
    F32
    [3840]
  • layers.18.ffn_norm2.weight
    F32
    [3840]
  • layers.19.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.19.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.19.attention.k_norm.weight
    F32
    [128]
  • layers.19.attention.out.weight
    F16
    [3840, 3840]
  • layers.19.attention.q_norm.weight
    F32
    [128]
  • layers.19.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.19.attention_norm1.weight
    F32
    [3840]
  • layers.19.attention_norm2.weight
    F32
    [3840]
  • layers.19.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.19.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.19.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.19.ffn_norm1.weight
    F32
    [3840]
  • layers.19.ffn_norm2.weight
    F32
    [3840]
  • layers.20.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.20.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.20.attention.k_norm.weight
    F32
    [128]
  • layers.20.attention.out.weight
    F16
    [3840, 3840]
  • layers.20.attention.q_norm.weight
    F32
    [128]
  • layers.20.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.20.attention_norm1.weight
    F32
    [3840]
  • layers.20.attention_norm2.weight
    F32
    [3840]
  • layers.20.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.20.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.20.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.20.ffn_norm1.weight
    F32
    [3840]
  • layers.20.ffn_norm2.weight
    F32
    [3840]
  • layers.21.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.21.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.21.attention.k_norm.weight
    F32
    [128]
  • layers.21.attention.out.weight
    F16
    [3840, 3840]
  • layers.21.attention.q_norm.weight
    F32
    [128]
  • layers.21.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.21.attention_norm1.weight
    F32
    [3840]
  • layers.21.attention_norm2.weight
    F32
    [3840]
  • layers.21.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.21.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.21.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.21.ffn_norm1.weight
    F32
    [3840]
  • layers.21.ffn_norm2.weight
    F32
    [3840]
  • layers.22.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.22.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.22.attention.k_norm.weight
    F32
    [128]
  • layers.22.attention.out.weight
    F16
    [3840, 3840]
  • layers.22.attention.q_norm.weight
    F32
    [128]
  • layers.22.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.22.attention_norm1.weight
    F32
    [3840]
  • layers.22.attention_norm2.weight
    F32
    [3840]
  • layers.22.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.22.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.22.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.22.ffn_norm1.weight
    F32
    [3840]
  • layers.22.ffn_norm2.weight
    F32
    [3840]
  • layers.23.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.23.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.23.attention.k_norm.weight
    F32
    [128]
  • layers.23.attention.out.weight
    F16
    [3840, 3840]
  • layers.23.attention.q_norm.weight
    F32
    [128]
  • layers.23.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.23.attention_norm1.weight
    F32
    [3840]
  • layers.23.attention_norm2.weight
    F32
    [3840]
  • layers.23.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.23.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.23.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.23.ffn_norm1.weight
    F32
    [3840]
  • layers.23.ffn_norm2.weight
    F32
    [3840]
  • layers.24.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.24.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.24.attention.k_norm.weight
    F32
    [128]
  • layers.24.attention.out.weight
    F16
    [3840, 3840]
  • layers.24.attention.q_norm.weight
    F32
    [128]
  • layers.24.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.24.attention_norm1.weight
    F32
    [3840]
  • layers.24.attention_norm2.weight
    F32
    [3840]
  • layers.24.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.24.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.24.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.24.ffn_norm1.weight
    F32
    [3840]
  • layers.24.ffn_norm2.weight
    F32
    [3840]
  • layers.25.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.25.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.25.attention.k_norm.weight
    F32
    [128]
  • layers.25.attention.out.weight
    F16
    [3840, 3840]
  • layers.25.attention.q_norm.weight
    F32
    [128]
  • layers.25.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.25.attention_norm1.weight
    F32
    [3840]
  • layers.25.attention_norm2.weight
    F32
    [3840]
  • layers.25.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.25.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.25.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.25.ffn_norm1.weight
    F32
    [3840]
  • layers.25.ffn_norm2.weight
    F32
    [3840]
  • layers.26.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.26.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.26.attention.k_norm.weight
    F32
    [128]
  • layers.26.attention.out.weight
    F16
    [3840, 3840]
  • layers.26.attention.q_norm.weight
    F32
    [128]
  • layers.26.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.26.attention_norm1.weight
    F32
    [3840]
  • layers.26.attention_norm2.weight
    F32
    [3840]
  • layers.26.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.26.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.26.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.26.ffn_norm1.weight
    F32
    [3840]
  • layers.26.ffn_norm2.weight
    F32
    [3840]
  • layers.27.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.27.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.27.attention.k_norm.weight
    F32
    [128]
  • layers.27.attention.out.weight
    F16
    [3840, 3840]
  • layers.27.attention.q_norm.weight
    F32
    [128]
  • layers.27.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.27.attention_norm1.weight
    F32
    [3840]
  • layers.27.attention_norm2.weight
    F32
    [3840]
  • layers.27.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.27.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.27.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.27.ffn_norm1.weight
    F32
    [3840]
  • layers.27.ffn_norm2.weight
    F32
    [3840]
  • layers.28.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.28.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.28.attention.k_norm.weight
    F32
    [128]
  • layers.28.attention.out.weight
    F16
    [3840, 3840]
  • layers.28.attention.q_norm.weight
    F32
    [128]
  • layers.28.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.28.attention_norm1.weight
    F32
    [3840]
  • layers.28.attention_norm2.weight
    F32
    [3840]
  • layers.28.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.28.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.28.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.28.ffn_norm1.weight
    F32
    [3840]
  • layers.28.ffn_norm2.weight
    F32
    [3840]
  • layers.29.adaLN_modulation.0.bias
    F32
    [15360]
  • layers.29.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • layers.29.attention.k_norm.weight
    F32
    [128]
  • layers.29.attention.out.weight
    F16
    [3840, 3840]
  • layers.29.attention.q_norm.weight
    F32
    [128]
  • layers.29.attention.qkv.weight
    BF16
    [3840, 11520]
  • layers.29.attention_norm1.weight
    F32
    [3840]
  • layers.29.attention_norm2.weight
    F32
    [3840]
  • layers.29.feed_forward.w1.weight
    F16
    [3840, 10240]
  • layers.29.feed_forward.w2.weight
    F16
    [10240, 3840]
  • layers.29.feed_forward.w3.weight
    F16
    [3840, 10240]
  • layers.29.ffn_norm1.weight
    F32
    [3840]
  • layers.29.ffn_norm2.weight
    F32
    [3840]
  • noise_refiner.0.adaLN_modulation.0.bias
    F32
    [15360]
  • noise_refiner.0.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • noise_refiner.0.attention.k_norm.weight
    F32
    [128]
  • noise_refiner.0.attention.out.weight
    F16
    [3840, 3840]
  • noise_refiner.0.attention.q_norm.weight
    F32
    [128]
  • noise_refiner.0.attention.qkv.weight
    BF16
    [3840, 11520]
  • noise_refiner.0.attention_norm1.weight
    F32
    [3840]
  • noise_refiner.0.attention_norm2.weight
    F32
    [3840]
  • noise_refiner.0.feed_forward.w1.weight
    F16
    [3840, 10240]
  • noise_refiner.0.feed_forward.w2.weight
    F16
    [10240, 3840]
  • noise_refiner.0.feed_forward.w3.weight
    F16
    [3840, 10240]
  • noise_refiner.0.ffn_norm1.weight
    F32
    [3840]
  • noise_refiner.0.ffn_norm2.weight
    F32
    [3840]
  • noise_refiner.1.adaLN_modulation.0.bias
    F32
    [15360]
  • noise_refiner.1.adaLN_modulation.0.weight
    F16
    [256, 15360]
  • noise_refiner.1.attention.k_norm.weight
    F32
    [128]
  • noise_refiner.1.attention.out.weight
    F16
    [3840, 3840]
  • noise_refiner.1.attention.q_norm.weight
    F32
    [128]
  • noise_refiner.1.attention.qkv.weight
    BF16
    [3840, 11520]
  • noise_refiner.1.attention_norm1.weight
    F32
    [3840]
  • noise_refiner.1.attention_norm2.weight
    F32
    [3840]
  • noise_refiner.1.feed_forward.w1.weight
    F16
    [3840, 10240]
  • noise_refiner.1.feed_forward.w2.weight
    F16
    [10240, 3840]
  • noise_refiner.1.feed_forward.w3.weight
    F16
    [3840, 10240]
  • noise_refiner.1.ffn_norm1.weight
    F32
    [3840]
  • noise_refiner.1.ffn_norm2.weight
    F32
    [3840]
  • t_embedder.mlp.0.bias
    F32
    [1024]
  • t_embedder.mlp.0.weight
    F16
    [256, 1024]
  • t_embedder.mlp.2.bias
    F32
    [256]
  • t_embedder.mlp.2.weight
    F16
    [1024, 256]
  • x_embedder.bias
    F32
    [3840]
  • x_embedder.weight
    F16
    [64, 3840]
  • x_pad_token
    F32
    [3840, 1]