{
  "noisy_encoder.patch_embed.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.patch_embed.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 225728,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "patch_embed.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      0,
      262144
    ],
    "dtype": "float32",
    "shape": [
      1024,
      16,
      1,
      2,
      2
    ],
    "zip_record": "vlm_uniddt_512/data/0",
    "source_zip_crc32": "60c39b59"
  },
  "noisy_encoder.patch_embed.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.patch_embed.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 488000,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "patch_embed.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      262144,
      266240
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/1",
    "source_zip_crc32": "886fd828"
  },
  "noisy_encoder.blocks.0.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 492224,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      266240,
      270336
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/10",
    "source_zip_crc32": "0a636b95"
  },
  "noisy_encoder.blocks.6.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 496448,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.6.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      270336,
      294912
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/100",
    "source_zip_crc32": "b2e045b8"
  },
  "noisy_encoder.blocks.7.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 198764736,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      294912,
      299008
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/101",
    "source_zip_crc32": "224c62da"
  },
  "noisy_encoder.blocks.7.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 293186112,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      299008,
      303104
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/102",
    "source_zip_crc32": "a5fa3d84"
  },
  "noisy_encoder.blocks.7.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 453692352,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      303104,
      307200
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/103",
    "source_zip_crc32": "471973c1"
  },
  "noisy_encoder.blocks.7.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 651940160,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      307200,
      311296
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/104",
    "source_zip_crc32": "56846be2"
  },
  "noisy_encoder.blocks.7.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 746361536,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      311296,
      12894208
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/105",
    "source_zip_crc32": "c958a6a9"
  },
  "noisy_encoder.blocks.7.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 902519104,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      12894208,
      12906496
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/106",
    "source_zip_crc32": "315efc5f"
  },
  "noisy_encoder.blocks.7.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 926140608,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      12906496,
      17100800
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/107",
    "source_zip_crc32": "5f9391d3"
  },
  "noisy_encoder.blocks.7.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 930335040,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      17100800,
      17104896
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/108",
    "source_zip_crc32": "ea7e2975"
  },
  "noisy_encoder.blocks.7.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 930339264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      17104896,
      33882112
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/109",
    "source_zip_crc32": "68f399bb"
  },
  "noisy_encoder.blocks.0.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 947116608,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      33882112,
      50659328
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/11",
    "source_zip_crc32": "64b7740c"
  },
  "noisy_encoder.blocks.7.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 963893952,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      50659328,
      50675712
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/110",
    "source_zip_crc32": "350b1792"
  },
  "noisy_encoder.blocks.7.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 963910464,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      50675712,
      67452928
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/111",
    "source_zip_crc32": "d81a09a9"
  },
  "noisy_encoder.blocks.7.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 980687808,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      67452928,
      67457024
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/112",
    "source_zip_crc32": "b70a4d6f"
  },
  "noisy_encoder.blocks.7.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 980692032,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      67457024,
      92622848
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/113",
    "source_zip_crc32": "7e6f43ef"
  },
  "noisy_encoder.blocks.7.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.7.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1005857984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.7.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      92622848,
      92647424
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/114",
    "source_zip_crc32": "d2fe0820"
  },
  "noisy_encoder.blocks.8.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1005882688,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      92647424,
      92651520
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/115",
    "source_zip_crc32": "e576c3ae"
  },
  "noisy_encoder.blocks.8.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1005886912,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      92651520,
      92655616
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/116",
    "source_zip_crc32": "37f36d87"
  },
  "noisy_encoder.blocks.8.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1005891136,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      92655616,
      92659712
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/117",
    "source_zip_crc32": "f32262a8"
  },
  "noisy_encoder.blocks.8.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1005895360,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      92659712,
      92663808
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/118",
    "source_zip_crc32": "8209c874"
  },
  "noisy_encoder.blocks.8.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1005899584,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      92663808,
      105246720
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/119",
    "source_zip_crc32": "26baafcb"
  },
  "noisy_encoder.blocks.0.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1018482624,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      105246720,
      105263104
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/12",
    "source_zip_crc32": "6832a3a6"
  },
  "noisy_encoder.blocks.8.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1018499136,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      105263104,
      105275392
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/120",
    "source_zip_crc32": "dee98016"
  },
  "noisy_encoder.blocks.8.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1018511552,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      105275392,
      109469696
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/121",
    "source_zip_crc32": "18673ddd"
  },
  "noisy_encoder.blocks.8.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1022705984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      109469696,
      109473792
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/122",
    "source_zip_crc32": "5c93e976"
  },
  "noisy_encoder.blocks.8.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1022710208,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      109473792,
      126251008
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/123",
    "source_zip_crc32": "af61d0ea"
  },
  "noisy_encoder.blocks.8.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1039487552,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      126251008,
      126267392
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/124",
    "source_zip_crc32": "5ab2788b"
  },
  "noisy_encoder.blocks.8.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1039504064,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      126267392,
      143044608
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/125",
    "source_zip_crc32": "b742c118"
  },
  "noisy_encoder.blocks.8.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1056281408,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      143044608,
      143048704
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/126",
    "source_zip_crc32": "52021cf6"
  },
  "noisy_encoder.blocks.8.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1056285632,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      143048704,
      168214528
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/127",
    "source_zip_crc32": "a2c054fc"
  },
  "noisy_encoder.blocks.8.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.8.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1081451584,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.8.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      168214528,
      168239104
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/128",
    "source_zip_crc32": "f9d8f1b5"
  },
  "noisy_encoder.blocks.9.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1081476288,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      168239104,
      168243200
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/129",
    "source_zip_crc32": "f68e5571"
  },
  "noisy_encoder.blocks.0.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1081480512,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      168243200,
      185020416
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/13",
    "source_zip_crc32": "5b969e3a"
  },
  "noisy_encoder.blocks.9.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1098257856,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      185020416,
      185024512
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/130",
    "source_zip_crc32": "9788e813"
  },
  "noisy_encoder.blocks.9.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1098262080,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      185024512,
      185028608
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/131",
    "source_zip_crc32": "be18fa70"
  },
  "noisy_encoder.blocks.9.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1098266304,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      185028608,
      185032704
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/132",
    "source_zip_crc32": "06de77a4"
  },
  "noisy_encoder.blocks.9.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1098270528,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      185032704,
      197615616
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/133",
    "source_zip_crc32": "5d17d77b"
  },
  "noisy_encoder.blocks.9.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1110853568,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      197615616,
      197627904
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/134",
    "source_zip_crc32": "03e04413"
  },
  "noisy_encoder.blocks.9.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1110865984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      197627904,
      201822208
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/135",
    "source_zip_crc32": "6ce310fe"
  },
  "noisy_encoder.blocks.9.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1115060416,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      201822208,
      201826304
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/136",
    "source_zip_crc32": "be0e7c61"
  },
  "noisy_encoder.blocks.9.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1115064640,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      201826304,
      218603520
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/137",
    "source_zip_crc32": "236736e1"
  },
  "noisy_encoder.blocks.9.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1131841984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      218603520,
      218619904
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/138",
    "source_zip_crc32": "4dda5b61"
  },
  "noisy_encoder.blocks.9.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1131858496,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      218619904,
      235397120
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/139",
    "source_zip_crc32": "643f3926"
  },
  "noisy_encoder.blocks.0.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1148635840,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      235397120,
      235401216
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/14",
    "source_zip_crc32": "18c5542b"
  },
  "noisy_encoder.blocks.9.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1148640064,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      235401216,
      235405312
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/140",
    "source_zip_crc32": "fbe281e5"
  },
  "noisy_encoder.blocks.9.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1148644288,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      235405312,
      260571136
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/141",
    "source_zip_crc32": "44eb9fdb"
  },
  "noisy_encoder.blocks.9.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.9.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1173810240,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.9.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      260571136,
      260595712
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/142",
    "source_zip_crc32": "b22b2fe1"
  },
  "noisy_encoder.blocks.10.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1173834944,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      260595712,
      260599808
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/143",
    "source_zip_crc32": "5f6debe6"
  },
  "noisy_encoder.blocks.10.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1173839168,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      260599808,
      260603904
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/144",
    "source_zip_crc32": "86e9d2df"
  },
  "noisy_encoder.blocks.10.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1173843392,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      260603904,
      260608000
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/145",
    "source_zip_crc32": "4d533404"
  },
  "noisy_encoder.blocks.10.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1173847616,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      260608000,
      260612096
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/146",
    "source_zip_crc32": "4542397b"
  },
  "noisy_encoder.blocks.10.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1173851840,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      260612096,
      273195008
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/147",
    "source_zip_crc32": "591b7516"
  },
  "noisy_encoder.blocks.10.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1186434880,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      273195008,
      273207296
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/148",
    "source_zip_crc32": "d8c613e0"
  },
  "noisy_encoder.blocks.10.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1186447296,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      273207296,
      277401600
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/149",
    "source_zip_crc32": "137440b6"
  },
  "noisy_encoder.blocks.0.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1190641728,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      277401600,
      302567424
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/15",
    "source_zip_crc32": "073b76ce"
  },
  "noisy_encoder.blocks.10.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1215807680,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      302567424,
      302571520
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/150",
    "source_zip_crc32": "2b6b16b0"
  },
  "noisy_encoder.blocks.10.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1215811904,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      302571520,
      319348736
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/151",
    "source_zip_crc32": "568006eb"
  },
  "noisy_encoder.blocks.10.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1232589248,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      319348736,
      319365120
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/152",
    "source_zip_crc32": "4c978e9c"
  },
  "noisy_encoder.blocks.10.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1232605760,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      319365120,
      336142336
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/153",
    "source_zip_crc32": "b9879cd1"
  },
  "noisy_encoder.blocks.10.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1249383104,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      336142336,
      336146432
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/154",
    "source_zip_crc32": "ed659741"
  },
  "noisy_encoder.blocks.10.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1249387328,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      336146432,
      361312256
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/155",
    "source_zip_crc32": "c7cde447"
  },
  "noisy_encoder.blocks.10.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.10.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274553280,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.10.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      361312256,
      361336832
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/156",
    "source_zip_crc32": "2fde7dcb"
  },
  "noisy_encoder.blocks.11.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274577984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      361336832,
      361340928
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/157",
    "source_zip_crc32": "aaa3266a"
  },
  "noisy_encoder.blocks.11.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274582208,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      361340928,
      361345024
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/158",
    "source_zip_crc32": "1ef3bcbf"
  },
  "noisy_encoder.blocks.11.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274586432,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      361345024,
      361349120
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/159",
    "source_zip_crc32": "30eaf9e4"
  },
  "noisy_encoder.blocks.0.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274590656,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      361349120,
      361373696
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/16",
    "source_zip_crc32": "b114c012"
  },
  "noisy_encoder.blocks.11.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274615360,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      361373696,
      361377792
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/160",
    "source_zip_crc32": "0feed595"
  },
  "noisy_encoder.blocks.11.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1274619584,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      361377792,
      373960704
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/161",
    "source_zip_crc32": "b9fb5be4"
  },
  "noisy_encoder.blocks.11.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1287202624,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      373960704,
      373972992
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/162",
    "source_zip_crc32": "80f9cfb9"
  },
  "noisy_encoder.blocks.11.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1287215040,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      373972992,
      378167296
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/163",
    "source_zip_crc32": "1594e82b"
  },
  "noisy_encoder.blocks.11.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1291409472,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      378167296,
      378171392
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/164",
    "source_zip_crc32": "bb4274ab"
  },
  "noisy_encoder.blocks.11.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1291413696,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      378171392,
      394948608
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/165",
    "source_zip_crc32": "e374e37a"
  },
  "noisy_encoder.blocks.11.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1308191040,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      394948608,
      394964992
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/166",
    "source_zip_crc32": "6b0219fe"
  },
  "noisy_encoder.blocks.11.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1308207552,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      394964992,
      411742208
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/167",
    "source_zip_crc32": "b04006e2"
  },
  "noisy_encoder.blocks.11.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1324984896,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      411742208,
      411746304
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/168",
    "source_zip_crc32": "239b845c"
  },
  "noisy_encoder.blocks.11.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1324989120,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      411746304,
      436912128
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/169",
    "source_zip_crc32": "72d86ff9"
  },
  "noisy_encoder.blocks.1.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350155072,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      436912128,
      436916224
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/17",
    "source_zip_crc32": "8a12b9e8"
  },
  "noisy_encoder.blocks.11.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.11.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350159296,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.11.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      436916224,
      436940800
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/170",
    "source_zip_crc32": "585fecc7"
  },
  "noisy_encoder.blocks.12.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350184000,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      436940800,
      436944896
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/171",
    "source_zip_crc32": "47c5eab2"
  },
  "noisy_encoder.blocks.12.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350188224,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      436944896,
      436948992
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/172",
    "source_zip_crc32": "f5f14fb0"
  },
  "noisy_encoder.blocks.12.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350192448,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      436948992,
      436953088
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/173",
    "source_zip_crc32": "ba2b8640"
  },
  "noisy_encoder.blocks.12.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350196672,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      436953088,
      436957184
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/174",
    "source_zip_crc32": "6f12e1db"
  },
  "noisy_encoder.blocks.12.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1350200896,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      436957184,
      449540096
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/175",
    "source_zip_crc32": "4ccb3dd9"
  },
  "noisy_encoder.blocks.12.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1362783936,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      449540096,
      449552384
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/176",
    "source_zip_crc32": "cbd0e1c0"
  },
  "noisy_encoder.blocks.12.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1362796352,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      449552384,
      453746688
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/177",
    "source_zip_crc32": "1c18e2a2"
  },
  "noisy_encoder.blocks.12.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1366990784,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      453746688,
      453750784
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/178",
    "source_zip_crc32": "235f8ab4"
  },
  "noisy_encoder.blocks.12.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1366995008,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      453750784,
      470528000
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/179",
    "source_zip_crc32": "f943afdd"
  },
  "noisy_encoder.blocks.1.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1383772352,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      470528000,
      470532096
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/18",
    "source_zip_crc32": "6e39ddc6"
  },
  "noisy_encoder.blocks.12.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1383776576,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      470532096,
      470548480
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/180",
    "source_zip_crc32": "948fff92"
  },
  "noisy_encoder.blocks.12.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1383793088,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      470548480,
      487325696
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/181",
    "source_zip_crc32": "ba601719"
  },
  "noisy_encoder.blocks.12.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1400570432,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      487325696,
      487329792
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/182",
    "source_zip_crc32": "9171e6de"
  },
  "noisy_encoder.blocks.12.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1400574656,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      487329792,
      512495616
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/183",
    "source_zip_crc32": "d2f10ef5"
  },
  "noisy_encoder.blocks.12.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.12.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1425740608,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.12.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      512495616,
      512520192
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/184",
    "source_zip_crc32": "44e8ece6"
  },
  "noisy_encoder.blocks.13.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1425765312,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      512520192,
      512524288
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/185",
    "source_zip_crc32": "ea3f75cb"
  },
  "noisy_encoder.blocks.13.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1425769536,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      512524288,
      512528384
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/186",
    "source_zip_crc32": "c75e5f07"
  },
  "noisy_encoder.blocks.13.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1425773760,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      512528384,
      512532480
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/187",
    "source_zip_crc32": "b5bbba88"
  },
  "noisy_encoder.blocks.13.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1425777984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      512532480,
      512536576
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/188",
    "source_zip_crc32": "82a922da"
  },
  "noisy_encoder.blocks.13.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1425782208,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      512536576,
      525119488
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/189",
    "source_zip_crc32": "fa6533af"
  },
  "noisy_encoder.blocks.1.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1438365248,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      525119488,
      525123584
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/19",
    "source_zip_crc32": "fed33dcd"
  },
  "noisy_encoder.blocks.13.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1438369472,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      525123584,
      525135872
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/190",
    "source_zip_crc32": "7ed81d37"
  },
  "noisy_encoder.blocks.13.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1438381888,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      525135872,
      529330176
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/191",
    "source_zip_crc32": "104df1a6"
  },
  "noisy_encoder.blocks.13.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1442576320,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      529330176,
      529334272
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/192",
    "source_zip_crc32": "8cf00766"
  },
  "noisy_encoder.blocks.13.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1442580544,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      529334272,
      546111488
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/193",
    "source_zip_crc32": "a7c13509"
  },
  "noisy_encoder.blocks.13.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1459357888,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      546111488,
      546127872
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/194",
    "source_zip_crc32": "afcfe122"
  },
  "noisy_encoder.blocks.13.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1459374400,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      546127872,
      562905088
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/195",
    "source_zip_crc32": "032251b7"
  },
  "noisy_encoder.blocks.13.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1476151744,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      562905088,
      562909184
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/196",
    "source_zip_crc32": "06599795"
  },
  "noisy_encoder.blocks.13.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1476155968,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      562909184,
      588075008
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/197",
    "source_zip_crc32": "ff6d7059"
  },
  "noisy_encoder.blocks.13.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.13.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1501321920,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.13.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      588075008,
      588099584
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/198",
    "source_zip_crc32": "07d94ae2"
  },
  "noisy_encoder.blocks.14.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1501346624,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      588099584,
      588103680
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/199",
    "source_zip_crc32": "de9e7dfd"
  },
  "noisy_encoder.pos_embed.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.pos_embed.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1501350848,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "pos_embed.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      588103680,
      597540864
    ],
    "dtype": "float32",
    "shape": [
      2304,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/2",
    "source_zip_crc32": "fef33bba"
  },
  "noisy_encoder.blocks.1.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1510788160,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      597540864,
      597544960
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/20",
    "source_zip_crc32": "ced6efad"
  },
  "noisy_encoder.blocks.14.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1510792384,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      597544960,
      597549056
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/200",
    "source_zip_crc32": "a386fff7"
  },
  "noisy_encoder.blocks.14.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1510796608,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      597549056,
      597553152
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/201",
    "source_zip_crc32": "12406a3c"
  },
  "noisy_encoder.blocks.14.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1510800832,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      597553152,
      597557248
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/202",
    "source_zip_crc32": "7c867e0e"
  },
  "noisy_encoder.blocks.14.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1510805056,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      597557248,
      610140160
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/203",
    "source_zip_crc32": "beaf6460"
  },
  "noisy_encoder.blocks.14.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1523388096,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      610140160,
      610152448
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/204",
    "source_zip_crc32": "c15fa229"
  },
  "noisy_encoder.blocks.14.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1523400512,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      610152448,
      614346752
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/205",
    "source_zip_crc32": "6d886219"
  },
  "noisy_encoder.blocks.14.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1527594944,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      614346752,
      614350848
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/206",
    "source_zip_crc32": "cf3c017f"
  },
  "noisy_encoder.blocks.14.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1527599168,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      614350848,
      631128064
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/207",
    "source_zip_crc32": "34661ed2"
  },
  "noisy_encoder.blocks.14.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1544376512,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      631128064,
      631144448
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/208",
    "source_zip_crc32": "9ecf6b15"
  },
  "noisy_encoder.blocks.14.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1544393024,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      631144448,
      647921664
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/209",
    "source_zip_crc32": "ea53d34a"
  },
  "noisy_encoder.blocks.1.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1561170368,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      647921664,
      660504576
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/21",
    "source_zip_crc32": "6020664b"
  },
  "noisy_encoder.blocks.14.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1573753408,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      660504576,
      660508672
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/210",
    "source_zip_crc32": "53f4773b"
  },
  "noisy_encoder.blocks.14.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1573757632,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      660508672,
      685674496
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/211",
    "source_zip_crc32": "6595b920"
  },
  "noisy_encoder.blocks.14.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.14.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1598923584,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.14.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      685674496,
      685699072
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/212",
    "source_zip_crc32": "32a244e9"
  },
  "noisy_encoder.blocks.15.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1598948288,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      685699072,
      685703168
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/213",
    "source_zip_crc32": "9b932c19"
  },
  "noisy_encoder.blocks.15.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1598952512,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      685703168,
      685707264
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/214",
    "source_zip_crc32": "0bcbf2b1"
  },
  "noisy_encoder.blocks.15.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1598956736,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      685707264,
      685711360
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/215",
    "source_zip_crc32": "1f176c72"
  },
  "noisy_encoder.blocks.15.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1598960960,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      685711360,
      685715456
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/216",
    "source_zip_crc32": "85899ea6"
  },
  "noisy_encoder.blocks.15.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1598965184,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      685715456,
      698298368
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/217",
    "source_zip_crc32": "3a04f6c1"
  },
  "noisy_encoder.blocks.15.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1611548224,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      698298368,
      698310656
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/218",
    "source_zip_crc32": "6b143757"
  },
  "noisy_encoder.blocks.15.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1611560640,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      698310656,
      702504960
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/219",
    "source_zip_crc32": "9e4ca313"
  },
  "noisy_encoder.blocks.1.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1615755072,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      702504960,
      702517248
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/22",
    "source_zip_crc32": "39518e99"
  },
  "noisy_encoder.blocks.15.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1615767488,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      702517248,
      702521344
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/220",
    "source_zip_crc32": "38652d23"
  },
  "noisy_encoder.blocks.15.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1615771712,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      702521344,
      719298560
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/221",
    "source_zip_crc32": "ac392358"
  },
  "noisy_encoder.blocks.15.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1632549056,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      719298560,
      719314944
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/222",
    "source_zip_crc32": "ec27a57e"
  },
  "noisy_encoder.blocks.15.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1632565568,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      719314944,
      736092160
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/223",
    "source_zip_crc32": "4ab08be3"
  },
  "noisy_encoder.blocks.15.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1649342912,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      736092160,
      736096256
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/224",
    "source_zip_crc32": "cea160aa"
  },
  "noisy_encoder.blocks.15.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1649347136,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      736096256,
      761262080
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/225",
    "source_zip_crc32": "f31aa00c"
  },
  "noisy_encoder.blocks.15.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.15.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1674513088,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.15.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      761262080,
      761286656
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/226",
    "source_zip_crc32": "95f81055"
  },
  "noisy_encoder.blocks.16.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1674537792,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      761286656,
      761290752
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/227",
    "source_zip_crc32": "258ce9cd"
  },
  "noisy_encoder.blocks.16.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1674542016,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      761290752,
      761294848
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/228",
    "source_zip_crc32": "d1174804"
  },
  "noisy_encoder.blocks.16.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1674546240,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      761294848,
      761298944
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/229",
    "source_zip_crc32": "85741cd2"
  },
  "noisy_encoder.blocks.1.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1674550464,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      761298944,
      765493248
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/23",
    "source_zip_crc32": "ef5b760a"
  },
  "noisy_encoder.blocks.16.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1678744896,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      765493248,
      765497344
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/230",
    "source_zip_crc32": "bc063f4a"
  },
  "noisy_encoder.blocks.16.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1678749120,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      765497344,
      778080256
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/231",
    "source_zip_crc32": "17ac3581"
  },
  "noisy_encoder.blocks.16.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1691332160,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      778080256,
      778092544
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/232",
    "source_zip_crc32": "70d68d32"
  },
  "noisy_encoder.blocks.16.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1691344576,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      778092544,
      782286848
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/233",
    "source_zip_crc32": "d36d8de0"
  },
  "noisy_encoder.blocks.16.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1695539008,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      782286848,
      782290944
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/234",
    "source_zip_crc32": "64480d25"
  },
  "noisy_encoder.blocks.16.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1695543232,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      782290944,
      799068160
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/235",
    "source_zip_crc32": "c00584dd"
  },
  "noisy_encoder.blocks.16.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1712320576,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      799068160,
      799084544
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/236",
    "source_zip_crc32": "8dd71066"
  },
  "noisy_encoder.blocks.16.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1712337088,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      799084544,
      815861760
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/237",
    "source_zip_crc32": "0c9da33a"
  },
  "noisy_encoder.blocks.16.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1729114432,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      815861760,
      815865856
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/238",
    "source_zip_crc32": "3421a695"
  },
  "noisy_encoder.blocks.16.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1729118656,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      815865856,
      841031680
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/239",
    "source_zip_crc32": "2a20c044"
  },
  "noisy_encoder.blocks.1.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754284608,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      841031680,
      841035776
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/24",
    "source_zip_crc32": "1786baf5"
  },
  "noisy_encoder.blocks.16.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.16.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754288832,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.16.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      841035776,
      841060352
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/240",
    "source_zip_crc32": "adee4335"
  },
  "noisy_encoder.blocks.17.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754313536,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      841060352,
      841064448
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/241",
    "source_zip_crc32": "ea2ff4be"
  },
  "noisy_encoder.blocks.17.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754317760,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      841064448,
      841068544
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/242",
    "source_zip_crc32": "df3a52bc"
  },
  "noisy_encoder.blocks.17.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754321984,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      841068544,
      841072640
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/243",
    "source_zip_crc32": "34ba76e0"
  },
  "noisy_encoder.blocks.17.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754326208,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      841072640,
      841076736
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/244",
    "source_zip_crc32": "861b9d8d"
  },
  "noisy_encoder.blocks.17.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1754330432,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      841076736,
      853659648
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/245",
    "source_zip_crc32": "5bab767b"
  },
  "noisy_encoder.blocks.17.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1766913472,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      853659648,
      853671936
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/246",
    "source_zip_crc32": "35499907"
  },
  "noisy_encoder.blocks.17.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1766925888,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      853671936,
      857866240
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/247",
    "source_zip_crc32": "1bd49a76"
  },
  "noisy_encoder.blocks.17.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1771120320,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      857866240,
      857870336
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/248",
    "source_zip_crc32": "d4634246"
  },
  "noisy_encoder.blocks.17.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1771124544,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      857870336,
      874647552
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/249",
    "source_zip_crc32": "ae633749"
  },
  "noisy_encoder.blocks.1.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1787901888,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      874647552,
      891424768
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/25",
    "source_zip_crc32": "ac4c0e8f"
  },
  "noisy_encoder.blocks.17.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1804679232,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      891424768,
      891441152
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/250",
    "source_zip_crc32": "4103d211"
  },
  "noisy_encoder.blocks.17.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1804695744,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      891441152,
      908218368
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/251",
    "source_zip_crc32": "1262c8d8"
  },
  "noisy_encoder.blocks.17.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1821473088,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      908218368,
      908222464
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/252",
    "source_zip_crc32": "9d0733df"
  },
  "noisy_encoder.blocks.17.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1821477312,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      908222464,
      933388288
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/253",
    "source_zip_crc32": "b65883bb"
  },
  "noisy_encoder.blocks.17.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.17.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1846643264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.17.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      933388288,
      933412864
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/254",
    "source_zip_crc32": "0f7112ce"
  },
  "noisy_encoder.blocks.18.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1846667968,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      933412864,
      933416960
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/255",
    "source_zip_crc32": "6697467f"
  },
  "noisy_encoder.blocks.18.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1846672192,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      933416960,
      933421056
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/256",
    "source_zip_crc32": "7dee1404"
  },
  "noisy_encoder.blocks.18.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1846676416,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      933421056,
      933425152
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/257",
    "source_zip_crc32": "e5f6aad6"
  },
  "noisy_encoder.blocks.18.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1846680640,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      933425152,
      933429248
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/258",
    "source_zip_crc32": "a83c449a"
  },
  "noisy_encoder.blocks.18.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1846684864,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      933429248,
      946012160
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/259",
    "source_zip_crc32": "cb058299"
  },
  "noisy_encoder.blocks.1.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1859267904,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      946012160,
      946028544
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/26",
    "source_zip_crc32": "8f6a95d0"
  },
  "noisy_encoder.blocks.18.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1859284416,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      946028544,
      946040832
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/260",
    "source_zip_crc32": "a9c02249"
  },
  "noisy_encoder.blocks.18.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1859296832,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      946040832,
      950235136
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/261",
    "source_zip_crc32": "dd738c8d"
  },
  "noisy_encoder.blocks.18.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1863491264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      950235136,
      950239232
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/262",
    "source_zip_crc32": "ccbd30c7"
  },
  "noisy_encoder.blocks.18.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1863495488,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      950239232,
      967016448
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/263",
    "source_zip_crc32": "4e11271f"
  },
  "noisy_encoder.blocks.18.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1880272832,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      967016448,
      967032832
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/264",
    "source_zip_crc32": "5f602222"
  },
  "noisy_encoder.blocks.18.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1880289344,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      967032832,
      983810048
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/265",
    "source_zip_crc32": "a71a691c"
  },
  "noisy_encoder.blocks.18.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1897066688,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      983810048,
      983814144
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/266",
    "source_zip_crc32": "0568b522"
  },
  "noisy_encoder.blocks.18.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1897070912,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      983814144,
      1008979968
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/267",
    "source_zip_crc32": "82754456"
  },
  "noisy_encoder.blocks.18.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.18.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1922236864,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.18.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1008979968,
      1009004544
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/268",
    "source_zip_crc32": "a3861607"
  },
  "noisy_encoder.blocks.19.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1922261568,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1009004544,
      1009008640
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/269",
    "source_zip_crc32": "254e09d4"
  },
  "noisy_encoder.blocks.1.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1922265792,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1009008640,
      1025785856
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/27",
    "source_zip_crc32": "65a22d68"
  },
  "noisy_encoder.blocks.19.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1939043136,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1025785856,
      1025789952
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/270",
    "source_zip_crc32": "c04818c0"
  },
  "noisy_encoder.blocks.19.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1939047360,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1025789952,
      1025794048
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/271",
    "source_zip_crc32": "263e46c3"
  },
  "noisy_encoder.blocks.19.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1939051584,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1025794048,
      1025798144
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/272",
    "source_zip_crc32": "34a07360"
  },
  "noisy_encoder.blocks.19.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1939055808,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1025798144,
      1038381056
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/273",
    "source_zip_crc32": "e213ba5d"
  },
  "noisy_encoder.blocks.19.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1951638848,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1038381056,
      1038393344
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/274",
    "source_zip_crc32": "35855907"
  },
  "noisy_encoder.blocks.19.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1951651264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1038393344,
      1042587648
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/275",
    "source_zip_crc32": "0e3a1749"
  },
  "noisy_encoder.blocks.19.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1955845696,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1042587648,
      1042591744
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/276",
    "source_zip_crc32": "8b4bcabf"
  },
  "noisy_encoder.blocks.19.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1955849920,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1042591744,
      1059368960
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/277",
    "source_zip_crc32": "dc1bbc14"
  },
  "noisy_encoder.blocks.19.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1972627264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1059368960,
      1059385344
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/278",
    "source_zip_crc32": "50155731"
  },
  "noisy_encoder.blocks.19.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1972643776,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1059385344,
      1076162560
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/279",
    "source_zip_crc32": "1089f119"
  },
  "noisy_encoder.blocks.1.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1989421120,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1076162560,
      1076166656
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/28",
    "source_zip_crc32": "bc781852"
  },
  "noisy_encoder.blocks.19.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1989425344,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1076166656,
      1076170752
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/280",
    "source_zip_crc32": "42395cf6"
  },
  "noisy_encoder.blocks.19.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 1989429568,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1076170752,
      1101336576
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/281",
    "source_zip_crc32": "7d1b6cfb"
  },
  "noisy_encoder.blocks.19.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.19.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2014595520,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.19.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1101336576,
      1101361152
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/282",
    "source_zip_crc32": "7f8255cb"
  },
  "noisy_encoder.blocks.20.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2014620224,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1101361152,
      1101365248
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/283",
    "source_zip_crc32": "8ebbb6ff"
  },
  "noisy_encoder.blocks.20.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2014624448,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1101365248,
      1101369344
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/284",
    "source_zip_crc32": "1a75d996"
  },
  "noisy_encoder.blocks.20.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2014628672,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1101369344,
      1101373440
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/285",
    "source_zip_crc32": "e843ce2b"
  },
  "noisy_encoder.blocks.20.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2014632896,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1101373440,
      1101377536
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/286",
    "source_zip_crc32": "099c356a"
  },
  "noisy_encoder.blocks.20.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2014637120,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1101377536,
      1113960448
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/287",
    "source_zip_crc32": "29eab23f"
  },
  "noisy_encoder.blocks.20.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2027220160,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1113960448,
      1113972736
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/288",
    "source_zip_crc32": "9bb65ae2"
  },
  "noisy_encoder.blocks.20.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2027232576,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1113972736,
      1118167040
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/289",
    "source_zip_crc32": "4d4f386f"
  },
  "noisy_encoder.blocks.1.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2031427008,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1118167040,
      1143332864
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/29",
    "source_zip_crc32": "db6e4032"
  },
  "noisy_encoder.blocks.20.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2056592960,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1143332864,
      1143336960
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/290",
    "source_zip_crc32": "929c1f31"
  },
  "noisy_encoder.blocks.20.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2056597184,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1143336960,
      1160114176
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/291",
    "source_zip_crc32": "cd9521c6"
  },
  "noisy_encoder.blocks.20.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2073374528,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1160114176,
      1160130560
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/292",
    "source_zip_crc32": "44d5bca8"
  },
  "noisy_encoder.blocks.20.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2073391040,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1160130560,
      1176907776
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/293",
    "source_zip_crc32": "51a520b1"
  },
  "noisy_encoder.blocks.20.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2090168384,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1176907776,
      1176911872
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/294",
    "source_zip_crc32": "0569b8d4"
  },
  "noisy_encoder.blocks.20.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2090172608,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1176911872,
      1202077696
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/295",
    "source_zip_crc32": "ff2dde25"
  },
  "noisy_encoder.blocks.20.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.20.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115338560,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.20.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202077696,
      1202102272
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/296",
    "source_zip_crc32": "9f83c801"
  },
  "noisy_encoder.blocks.21.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115363264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202102272,
      1202106368
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/297",
    "source_zip_crc32": "6c8fdbfe"
  },
  "noisy_encoder.blocks.21.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115367488,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202106368,
      1202110464
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/298",
    "source_zip_crc32": "c200e0fe"
  },
  "noisy_encoder.blocks.21.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115371712,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202110464,
      1202114560
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/299",
    "source_zip_crc32": "0e604dcf"
  },
  "noisy_encoder.blocks.0.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115375936,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.0.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202114560,
      1202118656
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/3",
    "source_zip_crc32": "0915fa71"
  },
  "noisy_encoder.blocks.1.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.1.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115380160,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.1.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202118656,
      1202143232
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/30",
    "source_zip_crc32": "95da49a3"
  },
  "noisy_encoder.blocks.21.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115404864,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202143232,
      1202147328
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/300",
    "source_zip_crc32": "d4edcc5b"
  },
  "noisy_encoder.blocks.21.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2115409088,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1202147328,
      1214730240
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/301",
    "source_zip_crc32": "bce3c0b9"
  },
  "noisy_encoder.blocks.21.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2127992128,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1214730240,
      1214742528
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/302",
    "source_zip_crc32": "05cadb69"
  },
  "noisy_encoder.blocks.21.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2128004544,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1214742528,
      1218936832
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/303",
    "source_zip_crc32": "9306891b"
  },
  "noisy_encoder.blocks.21.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2132198976,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1218936832,
      1218940928
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/304",
    "source_zip_crc32": "3c3292f7"
  },
  "noisy_encoder.blocks.21.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2132203200,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1218940928,
      1235718144
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/305",
    "source_zip_crc32": "5c8d7119"
  },
  "noisy_encoder.blocks.21.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2148980544,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1235718144,
      1235734528
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/306",
    "source_zip_crc32": "e1e0eb8b"
  },
  "noisy_encoder.blocks.21.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2148997056,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1235734528,
      1252511744
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/307",
    "source_zip_crc32": "4bb118d4"
  },
  "noisy_encoder.blocks.21.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2165774400,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1252511744,
      1252515840
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/308",
    "source_zip_crc32": "0ac697c3"
  },
  "noisy_encoder.blocks.21.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2165778624,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1252515840,
      1277681664
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/309",
    "source_zip_crc32": "20426ce7"
  },
  "noisy_encoder.blocks.2.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190944576,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.2.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277681664,
      1277685760
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/31",
    "source_zip_crc32": "c0c72bf1"
  },
  "noisy_encoder.blocks.21.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.21.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190948800,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.21.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277685760,
      1277710336
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/310",
    "source_zip_crc32": "bd64a186"
  },
  "noisy_encoder.blocks.22.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190973504,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277710336,
      1277714432
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/311",
    "source_zip_crc32": "c6e2e870"
  },
  "noisy_encoder.blocks.22.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190977728,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277714432,
      1277718528
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/312",
    "source_zip_crc32": "255b1ca9"
  },
  "noisy_encoder.blocks.22.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190981952,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277718528,
      1277722624
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/313",
    "source_zip_crc32": "d353cdb0"
  },
  "noisy_encoder.blocks.22.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190986176,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277722624,
      1277726720
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/314",
    "source_zip_crc32": "0f8f5ad3"
  },
  "noisy_encoder.blocks.22.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2190990400,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1277726720,
      1290309632
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/315",
    "source_zip_crc32": "c51df70d"
  },
  "noisy_encoder.blocks.22.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2203573440,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1290309632,
      1290321920
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/316",
    "source_zip_crc32": "0cdae555"
  },
  "noisy_encoder.blocks.22.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2203585856,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1290321920,
      1294516224
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/317",
    "source_zip_crc32": "02c9a102"
  },
  "noisy_encoder.blocks.22.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2207780288,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1294516224,
      1294520320
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/318",
    "source_zip_crc32": "e19e8840"
  },
  "noisy_encoder.blocks.22.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2207784512,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1294520320,
      1311297536
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/319",
    "source_zip_crc32": "392a6982"
  },
  "noisy_encoder.blocks.2.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2224561856,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.2.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1311297536,
      1311301632
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/32",
    "source_zip_crc32": "d395f3a5"
  },
  "noisy_encoder.blocks.22.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2224566080,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1311301632,
      1311318016
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/320",
    "source_zip_crc32": "9a5fd17f"
  },
  "noisy_encoder.blocks.22.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2224582592,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1311318016,
      1328095232
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/321",
    "source_zip_crc32": "341c29b5"
  },
  "noisy_encoder.blocks.22.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2241359936,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1328095232,
      1328099328
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/322",
    "source_zip_crc32": "0500bd33"
  },
  "noisy_encoder.blocks.22.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2241364160,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1328099328,
      1353265152
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/323",
    "source_zip_crc32": "c5090237"
  },
  "noisy_encoder.blocks.22.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.22.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2266530112,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.22.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1353265152,
      1353289728
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/324",
    "source_zip_crc32": "9e03068a"
  },
  "noisy_encoder.blocks.23.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2266554816,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.norm1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1353289728,
      1353293824
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/325",
    "source_zip_crc32": "c913c79b"
  },
  "noisy_encoder.blocks.23.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2266559040,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.norm1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1353293824,
      1353297920
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/326",
    "source_zip_crc32": "a3cc8527"
  },
  "noisy_encoder.blocks.23.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2266563264,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1353297920,
      1353302016
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/327",
    "source_zip_crc32": "0f692570"
  },
  "noisy_encoder.blocks.23.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2266567488,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1353302016,
      1353306112
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/328",
    "source_zip_crc32": "05ddd33b"
  },
  "noisy_encoder.blocks.23.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2266571712,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.attn.qkv.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1353306112,
      1365889024
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/329",
    "source_zip_crc32": "b892220d"
  },
  "noisy_encoder.blocks.2.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2279154752,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.2.norm2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1365889024,
      1365893120
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/33",
    "source_zip_crc32": "51a67434"
  },
  "noisy_encoder.blocks.23.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2279158976,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.attn.qkv.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1365893120,
      1365905408
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/330",
    "source_zip_crc32": "a66d19bf"
  },
  "noisy_encoder.blocks.23.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2279171392,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.attn.proj.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1365905408,
      1370099712
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/331",
    "source_zip_crc32": "4c911593"
  },
  "noisy_encoder.blocks.23.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2283365824,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.attn.proj.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1370099712,
      1370103808
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/332",
    "source_zip_crc32": "d34eeba0"
  },
  "noisy_encoder.blocks.23.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2283370048,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.mlp.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1370103808,
      1386881024
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/333",
    "source_zip_crc32": "395ab872"
  },
  "noisy_encoder.blocks.23.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2300147392,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.mlp.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1386881024,
      1386897408
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/334",
    "source_zip_crc32": "81e38e4a"
  },
  "noisy_encoder.blocks.23.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2300163904,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.mlp.linear_fc2.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1386897408,
      1403674624
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/335",
    "source_zip_crc32": "a3e5fe1b"
  },
  "noisy_encoder.blocks.23.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2316941248,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.mlp.linear_fc2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1403674624,
      1403678720
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/336",
    "source_zip_crc32": "00a1a7f1"
  },
  "noisy_encoder.blocks.23.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2316945472,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.adaLN_modulation.0.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1403678720,
      1428844544
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/337",
    "source_zip_crc32": "0bfbb443"
  },
  "noisy_encoder.blocks.23.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.23.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2342111424,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.23.adaLN_modulation.0.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1428844544,
      1428869120
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/338",
    "source_zip_crc32": "998d8842"
  },
  "noisy_encoder.merger.norm.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.merger.norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2342136128,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "merger.norm.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1428869120,
      1428873216
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/339",
    "source_zip_crc32": "a61dd372"
  },
  "noisy_encoder.blocks.2.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2342140352,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "blocks.2.norm2.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1428873216,
      1428877312
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/34",
    "source_zip_crc32": "c380665d"
  },
  "noisy_encoder.merger.norm.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.merger.norm.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2342144576,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "merger.norm.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1428877312,
      1428881408
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/340",
    "source_zip_crc32": "5eb91c71"
  },
  "noisy_encoder.merger.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.merger.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2342148800,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "merger.linear_fc1.weight",
    "output_header_bytes": 27152,
    "output_offsets": [
      1428881408,
      1495990272
    ],
    "dtype": "float32",
    "shape": [
      4096,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/341",
    "source_zip_crc32": "3348cbb0"
  },
  "noisy_encoder.merger.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.merger.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2409257792,
    "output_file": "weights/noisy_encoder/model-00001-of-00002.safetensors",
    "output_key": "merger.linear_fc1.bias",
    "output_header_bytes": 27152,
    "output_offsets": [
      1495990272,
      1496006656
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/342",
    "source_zip_crc32": "2ab86d2d"
  },
  "noisy_encoder.merger.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.merger.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2409274304,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "merger.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      0,
      41943040
    ],
    "dtype": "float32",
    "shape": [
      2560,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/343",
    "source_zip_crc32": "7ed14ab8"
  },
  "noisy_encoder.merger.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.merger.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2451217472,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "merger.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      41943040,
      41953280
    ],
    "dtype": "float32",
    "shape": [
      2560
    ],
    "zip_record": "vlm_uniddt_512/data/344",
    "source_zip_crc32": "acc1d601"
  },
  "noisy_encoder.deepstack_merger_list.0.norm.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.0.norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2451227840,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.0.norm.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      41953280,
      41969664
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/345",
    "source_zip_crc32": "db5a3ed7"
  },
  "noisy_encoder.deepstack_merger_list.0.norm.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.0.norm.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2451244352,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.0.norm.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      41969664,
      41986048
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/346",
    "source_zip_crc32": "62dd48e9"
  },
  "noisy_encoder.deepstack_merger_list.0.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.0.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2451260864,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.0.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      41986048,
      109094912
    ],
    "dtype": "float32",
    "shape": [
      4096,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/347",
    "source_zip_crc32": "06eb0b38"
  },
  "noisy_encoder.deepstack_merger_list.0.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.0.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2518369856,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.0.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      109094912,
      109111296
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/348",
    "source_zip_crc32": "46b2a990"
  },
  "noisy_encoder.deepstack_merger_list.0.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.0.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2518386368,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.0.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      109111296,
      151054336
    ],
    "dtype": "float32",
    "shape": [
      2560,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/349",
    "source_zip_crc32": "d6195577"
  },
  "noisy_encoder.blocks.2.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2560329536,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.attn.qkv.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      151054336,
      163637248
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/35",
    "source_zip_crc32": "6fb924fe"
  },
  "noisy_encoder.deepstack_merger_list.0.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.0.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2572912576,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.0.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      163637248,
      163647488
    ],
    "dtype": "float32",
    "shape": [
      2560
    ],
    "zip_record": "vlm_uniddt_512/data/350",
    "source_zip_crc32": "d6723d5a"
  },
  "noisy_encoder.deepstack_merger_list.1.norm.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.1.norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2572922944,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.1.norm.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      163647488,
      163663872
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/351",
    "source_zip_crc32": "49ed30e9"
  },
  "noisy_encoder.deepstack_merger_list.1.norm.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.1.norm.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2572939456,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.1.norm.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      163663872,
      163680256
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/352",
    "source_zip_crc32": "f50fcb8c"
  },
  "noisy_encoder.deepstack_merger_list.1.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.1.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2572955968,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.1.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      163680256,
      230789120
    ],
    "dtype": "float32",
    "shape": [
      4096,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/353",
    "source_zip_crc32": "bbd90785"
  },
  "noisy_encoder.deepstack_merger_list.1.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.1.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2640064960,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.1.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      230789120,
      230805504
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/354",
    "source_zip_crc32": "f0fe3828"
  },
  "noisy_encoder.deepstack_merger_list.1.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.1.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2640081472,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.1.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      230805504,
      272748544
    ],
    "dtype": "float32",
    "shape": [
      2560,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/355",
    "source_zip_crc32": "cf563cc5"
  },
  "noisy_encoder.deepstack_merger_list.1.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.1.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2682024640,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.1.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      272748544,
      272758784
    ],
    "dtype": "float32",
    "shape": [
      2560
    ],
    "zip_record": "vlm_uniddt_512/data/356",
    "source_zip_crc32": "fdbd3d90"
  },
  "noisy_encoder.deepstack_merger_list.2.norm.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.2.norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2682035008,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.2.norm.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      272758784,
      272775168
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/357",
    "source_zip_crc32": "3f79af34"
  },
  "noisy_encoder.deepstack_merger_list.2.norm.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.2.norm.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2682051520,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.2.norm.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      272775168,
      272791552
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/358",
    "source_zip_crc32": "2307b421"
  },
  "noisy_encoder.deepstack_merger_list.2.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.2.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2682068032,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.2.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      272791552,
      339900416
    ],
    "dtype": "float32",
    "shape": [
      4096,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/359",
    "source_zip_crc32": "26406d97"
  },
  "noisy_encoder.blocks.2.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2749177024,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.attn.qkv.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      339900416,
      339912704
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/36",
    "source_zip_crc32": "75114455"
  },
  "noisy_encoder.deepstack_merger_list.2.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.2.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2749189440,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.2.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      339912704,
      339929088
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/360",
    "source_zip_crc32": "316023e1"
  },
  "noisy_encoder.deepstack_merger_list.2.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.2.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2749205952,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.2.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      339929088,
      381872128
    ],
    "dtype": "float32",
    "shape": [
      2560,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/361",
    "source_zip_crc32": "10cdf0d9"
  },
  "noisy_encoder.deepstack_merger_list.2.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.deepstack_merger_list.2.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2791149120,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "deepstack_merger_list.2.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      381872128,
      381882368
    ],
    "dtype": "float32",
    "shape": [
      2560
    ],
    "zip_record": "vlm_uniddt_512/data/362",
    "source_zip_crc32": "3f1b7ca6"
  },
  "noisy_encoder.t_embedder.mlp.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.t_embedder.mlp.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2791159488,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "t_embedder.mlp.0.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      381882368,
      382930944
    ],
    "dtype": "float32",
    "shape": [
      1024,
      256
    ],
    "zip_record": "vlm_uniddt_512/data/363",
    "source_zip_crc32": "e8980f0f"
  },
  "noisy_encoder.t_embedder.mlp.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.t_embedder.mlp.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2792208192,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "t_embedder.mlp.0.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      382930944,
      382935040
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/364",
    "source_zip_crc32": "7c068751"
  },
  "noisy_encoder.t_embedder.mlp.2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.t_embedder.mlp.2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2792212416,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "t_embedder.mlp.2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      382935040,
      387129344
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/365",
    "source_zip_crc32": "fb891b1c"
  },
  "noisy_encoder.t_embedder.mlp.2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.t_embedder.mlp.2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 2796406848,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "t_embedder.mlp.2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      387129344,
      387133440
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/366",
    "source_zip_crc32": "58f9a504"
  },
  "noisy_encoder.blocks.2.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 4404664896,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.attn.proj.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      387133440,
      391327744
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/37",
    "source_zip_crc32": "57d8d13c"
  },
  "noisy_encoder.blocks.2.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 4802098112,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.attn.proj.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      391327744,
      391331840
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/38",
    "source_zip_crc32": "3d04f834"
  },
  "noisy_encoder.blocks.2.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 5163883840,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.mlp.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      391331840,
      408109056
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/39",
    "source_zip_crc32": "cad5f7d1"
  },
  "noisy_encoder.blocks.0.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 5584375488,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.0.norm1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      408109056,
      408113152
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/4",
    "source_zip_crc32": "24ef48f1"
  },
  "noisy_encoder.blocks.2.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 5584379712,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.mlp.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      408113152,
      408129536
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/40",
    "source_zip_crc32": "ef15d411"
  },
  "noisy_encoder.blocks.2.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 5988110528,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.mlp.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      408129536,
      424906752
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/41",
    "source_zip_crc32": "89f93353"
  },
  "noisy_encoder.blocks.2.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 6308997696,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.mlp.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      424906752,
      424910848
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/42",
    "source_zip_crc32": "11761235"
  },
  "noisy_encoder.blocks.2.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 6613111744,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.adaLN_modulation.0.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      424910848,
      450076672
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/43",
    "source_zip_crc32": "a5dca667"
  },
  "noisy_encoder.blocks.2.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.2.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 6942387520,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.2.adaLN_modulation.0.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      450076672,
      450101248
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/44",
    "source_zip_crc32": "5a0c426d"
  },
  "noisy_encoder.blocks.3.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 7346136256,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.norm1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      450101248,
      450105344
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/45",
    "source_zip_crc32": "6241f8ee"
  },
  "noisy_encoder.blocks.3.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 7749864512,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.norm1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      450105344,
      450109440
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/46",
    "source_zip_crc32": "b4e42408"
  },
  "noisy_encoder.blocks.3.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 8111650240,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.norm2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      450109440,
      450113536
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/47",
    "source_zip_crc32": "ee7e62c4"
  },
  "noisy_encoder.blocks.3.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 8504893248,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.norm2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      450113536,
      450117632
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/48",
    "source_zip_crc32": "29bc75b6"
  },
  "noisy_encoder.blocks.3.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 8898136256,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.attn.qkv.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      450117632,
      462700544
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/49",
    "source_zip_crc32": "006a0794"
  },
  "noisy_encoder.blocks.0.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 9272500800,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.0.norm2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      462700544,
      462704640
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/5",
    "source_zip_crc32": "da425d97"
  },
  "noisy_encoder.blocks.3.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 9272505024,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.attn.qkv.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      462704640,
      462716928
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/50",
    "source_zip_crc32": "f5cea859"
  },
  "noisy_encoder.blocks.3.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 9676231744,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.attn.proj.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      462716928,
      466911232
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/51",
    "source_zip_crc32": "ff2069db"
  },
  "noisy_encoder.blocks.3.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 10084140480,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.attn.proj.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      466911232,
      466915328
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/52",
    "source_zip_crc32": "650de7c4"
  },
  "noisy_encoder.blocks.3.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 10388254528,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.mlp.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      466915328,
      483692544
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/53",
    "source_zip_crc32": "f0007f27"
  },
  "noisy_encoder.blocks.3.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 10709141696,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.mlp.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      483692544,
      483708928
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/54",
    "source_zip_crc32": "0fb554fa"
  },
  "noisy_encoder.blocks.3.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 11013268032,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.mlp.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      483708928,
      500486144
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/55",
    "source_zip_crc32": "b62f551c"
  },
  "noisy_encoder.blocks.3.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 11433769408,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.mlp.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      500486144,
      500490240
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/56",
    "source_zip_crc32": "1fb46a30"
  },
  "noisy_encoder.blocks.3.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 11837497664,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.adaLN_modulation.0.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      500490240,
      525656064
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/57",
    "source_zip_crc32": "0672c44b"
  },
  "noisy_encoder.blocks.3.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.3.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 12224445120,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.3.adaLN_modulation.0.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      525656064,
      525680640
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/58",
    "source_zip_crc32": "bb94792b"
  },
  "noisy_encoder.blocks.4.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 12617708608,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.norm1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      525680640,
      525684736
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/59",
    "source_zip_crc32": "9afab620"
  },
  "noisy_encoder.blocks.0.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 13010951616,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.0.norm2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      525684736,
      525688832
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/6",
    "source_zip_crc32": "15540d14"
  },
  "noisy_encoder.blocks.4.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 13010955840,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.norm1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      525688832,
      525692928
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/60",
    "source_zip_crc32": "4cf32e4e"
  },
  "noisy_encoder.blocks.4.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 13372741568,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.norm2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      525692928,
      525697024
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/61",
    "source_zip_crc32": "b8076789"
  },
  "noisy_encoder.blocks.4.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 13776460096,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.norm2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      525697024,
      525701120
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/62",
    "source_zip_crc32": "47e2dffd"
  },
  "noisy_encoder.blocks.4.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 14180178624,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.attn.qkv.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      525701120,
      538284032
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/63",
    "source_zip_crc32": "8cc1d3d0"
  },
  "noisy_encoder.blocks.4.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 14496871488,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.attn.qkv.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      538284032,
      538296320
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/64",
    "source_zip_crc32": "56cc376a"
  },
  "noisy_encoder.blocks.4.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 14800993728,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.attn.proj.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      538296320,
      542490624
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/65",
    "source_zip_crc32": "bca07042"
  },
  "noisy_encoder.blocks.4.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 15109297984,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.attn.proj.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      542490624,
      542494720
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/66",
    "source_zip_crc32": "4a5add8b"
  },
  "noisy_encoder.blocks.4.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 15513026240,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.mlp.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      542494720,
      559271936
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/67",
    "source_zip_crc32": "114d3cdd"
  },
  "noisy_encoder.blocks.4.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 15933527616,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.mlp.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      559271936,
      559288320
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/68",
    "source_zip_crc32": "ea49198b"
  },
  "noisy_encoder.blocks.4.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 16295325632,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.mlp.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      559288320,
      576065536
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/69",
    "source_zip_crc32": "4fb876cd"
  },
  "noisy_encoder.blocks.0.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 16705341760,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.0.attn.qkv.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      576065536,
      588648448
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/7",
    "source_zip_crc32": "ce15399d"
  },
  "noisy_encoder.blocks.4.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 16717924800,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.mlp.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      588648448,
      588652544
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/70",
    "source_zip_crc32": "18655b35"
  },
  "noisy_encoder.blocks.4.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 17111167808,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.adaLN_modulation.0.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      588652544,
      613818368
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/71",
    "source_zip_crc32": "de94011c"
  },
  "noisy_encoder.blocks.4.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.4.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 17498115264,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.4.adaLN_modulation.0.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      613818368,
      613842944
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/72",
    "source_zip_crc32": "c79e428a"
  },
  "noisy_encoder.blocks.5.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 17901854272,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.norm1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      613842944,
      613847040
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/73",
    "source_zip_crc32": "7216c31a"
  },
  "noisy_encoder.blocks.5.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 18305572800,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.norm1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      613847040,
      613851136
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/74",
    "source_zip_crc32": "136754e9"
  },
  "noisy_encoder.blocks.5.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 18609686848,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.norm2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      613851136,
      613855232
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/75",
    "source_zip_crc32": "2af18b29"
  },
  "noisy_encoder.blocks.5.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 18913800896,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.norm2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      613855232,
      613859328
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/76",
    "source_zip_crc32": "ef10a461"
  },
  "noisy_encoder.blocks.5.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19129200768,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.attn.qkv.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      613859328,
      626442240
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/77",
    "source_zip_crc32": "dd4edb43"
  },
  "noisy_encoder.blocks.5.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19255069184,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.attn.qkv.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      626442240,
      626454528
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/78",
    "source_zip_crc32": "4b57627d"
  },
  "noisy_encoder.blocks.5.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19368366976,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.attn.proj.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      626454528,
      630648832
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/79",
    "source_zip_crc32": "54e890f9"
  },
  "noisy_encoder.blocks.0.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19485846784,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.0.attn.qkv.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      630648832,
      630661120
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/8",
    "source_zip_crc32": "f439752e"
  },
  "noisy_encoder.blocks.5.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19485859200,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.attn.proj.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      630661120,
      630665216
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/80",
    "source_zip_crc32": "55d680bb"
  },
  "noisy_encoder.blocks.5.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19636891392,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.mlp.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      630665216,
      647442432
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/81",
    "source_zip_crc32": "02f0470a"
  },
  "noisy_encoder.blocks.5.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19795265664,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.mlp.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      647442432,
      647458816
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/82",
    "source_zip_crc32": "393ffeac"
  },
  "noisy_encoder.blocks.5.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19993525760,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.mlp.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      647458816,
      664236032
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/83",
    "source_zip_crc32": "abca5098"
  },
  "noisy_encoder.blocks.5.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20104720256,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.mlp.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      664236032,
      664240128
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/84",
    "source_zip_crc32": "1ac05ee8"
  },
  "noisy_encoder.blocks.5.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20265226496,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.adaLN_modulation.0.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      664240128,
      689405952
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/85",
    "source_zip_crc32": "c20d70b2"
  },
  "noisy_encoder.blocks.5.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.5.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20488636032,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.5.adaLN_modulation.0.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      689405952,
      689430528
    ],
    "dtype": "float32",
    "shape": [
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/86",
    "source_zip_crc32": "02dcd8f6"
  },
  "noisy_encoder.blocks.6.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20583077888,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.norm1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      689430528,
      689434624
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/87",
    "source_zip_crc32": "c18e15ba"
  },
  "noisy_encoder.blocks.6.norm1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.norm1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20743584128,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.norm1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      689434624,
      689438720
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/88",
    "source_zip_crc32": "d2935a36"
  },
  "noisy_encoder.blocks.6.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20941831936,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.norm2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      689438720,
      689442816
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/89",
    "source_zip_crc32": "a68e6602"
  },
  "noisy_encoder.blocks.0.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.0.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21036253312,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.0.attn.proj.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      689442816,
      693637120
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/9",
    "source_zip_crc32": "7fd9cba9"
  },
  "noisy_encoder.blocks.6.norm2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.norm2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21040447744,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.norm2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      693637120,
      693641216
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/90",
    "source_zip_crc32": "23a0c509"
  },
  "noisy_encoder.blocks.6.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21200953984,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.attn.qkv.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      693641216,
      706224128
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/91",
    "source_zip_crc32": "8a6fc519"
  },
  "noisy_encoder.blocks.6.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21411780608,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.attn.qkv.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      706224128,
      706236416
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/92",
    "source_zip_crc32": "96c69958"
  },
  "noisy_encoder.blocks.6.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21506210176,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.attn.proj.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      706236416,
      710430720
    ],
    "dtype": "float32",
    "shape": [
      1024,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/93",
    "source_zip_crc32": "079aca13"
  },
  "noisy_encoder.blocks.6.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21670906624,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.attn.proj.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      710430720,
      710434816
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/94",
    "source_zip_crc32": "0ba23f5f"
  },
  "noisy_encoder.blocks.6.mlp.linear_fc1.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.mlp.linear_fc1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21869154432,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.mlp.linear_fc1.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      710434816,
      727212032
    ],
    "dtype": "float32",
    "shape": [
      4096,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/95",
    "source_zip_crc32": "2fbbc412"
  },
  "noisy_encoder.blocks.6.mlp.linear_fc1.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.mlp.linear_fc1.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21980348928,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.mlp.linear_fc1.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      727212032,
      727228416
    ],
    "dtype": "float32",
    "shape": [
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/96",
    "source_zip_crc32": "4f9e9b89"
  },
  "noisy_encoder.blocks.6.mlp.linear_fc2.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.mlp.linear_fc2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22140867456,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.mlp.linear_fc2.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      727228416,
      744005632
    ],
    "dtype": "float32",
    "shape": [
      1024,
      4096
    ],
    "zip_record": "vlm_uniddt_512/data/97",
    "source_zip_crc32": "e0f33825"
  },
  "noisy_encoder.blocks.6.mlp.linear_fc2.bias": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.mlp.linear_fc2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22355888384,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.mlp.linear_fc2.bias",
    "output_header_bytes": 9712,
    "output_offsets": [
      744005632,
      744009728
    ],
    "dtype": "float32",
    "shape": [
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/98",
    "source_zip_crc32": "62d51714"
  },
  "noisy_encoder.blocks.6.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.noisy_encoder.blocks.6.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22450309760,
    "output_file": "weights/noisy_encoder/model-00002-of-00002.safetensors",
    "output_key": "blocks.6.adaLN_modulation.0.weight",
    "output_header_bytes": 9712,
    "output_offsets": [
      744009728,
      769175552
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1024
    ],
    "zip_record": "vlm_uniddt_512/data/99",
    "source_zip_crc32": "970f0fb7"
  },
  "diffusion_decoder.blocks.12.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 521152,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.attn.conditon_kv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      0,
      18874368
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1000",
    "source_zip_crc32": "b09ff054"
  },
  "diffusion_decoder.blocks.12.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19395648,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.attn.conditon_kv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      18874368,
      18886656
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/1001",
    "source_zip_crc32": "0fcd7c6f"
  },
  "diffusion_decoder.blocks.12.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19408064,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      18886656,
      28323840
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1002",
    "source_zip_crc32": "60c3f7bf"
  },
  "diffusion_decoder.blocks.12.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 28845376,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      28323840,
      28329984
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1003",
    "source_zip_crc32": "f57c3123"
  },
  "diffusion_decoder.blocks.12.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 28851648,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      28329984,
      66078720
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1004",
    "source_zip_crc32": "74a84ee1"
  },
  "diffusion_decoder.blocks.12.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 66600512,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      66078720,
      103827456
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1005",
    "source_zip_crc32": "f40e050f"
  },
  "diffusion_decoder.blocks.12.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 104349376,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      103827456,
      141576192
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/1006",
    "source_zip_crc32": "75c280ee"
  },
  "diffusion_decoder.blocks.12.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 142098240,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.adaLN_modulation.0.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      141576192,
      198199296
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1007",
    "source_zip_crc32": "f937f721"
  },
  "diffusion_decoder.blocks.12.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 198721472,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.12.adaLN_modulation.0.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      198199296,
      198236160
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/1008",
    "source_zip_crc32": "7e80c4d0"
  },
  "diffusion_decoder.blocks.13.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 198758464,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      198236160,
      198242304
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1009",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.13.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 198768960,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      198242304,
      198248448
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1010",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.13.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 198775232,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      198248448,
      226560000
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1011",
    "source_zip_crc32": "c3688bda"
  },
  "diffusion_decoder.blocks.13.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 227086912,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      226560000,
      226578432
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/1012",
    "source_zip_crc32": "3276feb9"
  },
  "diffusion_decoder.blocks.13.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 227105472,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      226578432,
      226578944
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/1013",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.13.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 227106112,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      226578944,
      226579456
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/1014",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.13.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 227106752,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.conditon_kv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      226579456,
      245453824
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1015",
    "source_zip_crc32": "08e9f620"
  },
  "diffusion_decoder.blocks.13.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 245981248,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.conditon_kv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      245453824,
      245466112
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/1016",
    "source_zip_crc32": "ce1901b9"
  },
  "diffusion_decoder.blocks.13.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 245993664,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      245466112,
      254903296
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1017",
    "source_zip_crc32": "6219604f"
  },
  "diffusion_decoder.blocks.13.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 255430976,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      254903296,
      254909440
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1018",
    "source_zip_crc32": "dbfcd297"
  },
  "diffusion_decoder.blocks.13.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 255437248,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      254909440,
      292658176
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1019",
    "source_zip_crc32": "cc166257"
  },
  "diffusion_decoder.blocks.13.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 293190336,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      292658176,
      330406912
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1020",
    "source_zip_crc32": "b1b0bc2e"
  },
  "diffusion_decoder.blocks.13.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 330939200,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      330406912,
      368155648
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/1021",
    "source_zip_crc32": "8802feaa"
  },
  "diffusion_decoder.blocks.13.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 368688064,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.adaLN_modulation.0.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      368155648,
      424778752
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1022",
    "source_zip_crc32": "0cf6a7b6"
  },
  "diffusion_decoder.blocks.13.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.13.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 425311296,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.13.adaLN_modulation.0.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      424778752,
      424815616
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/1023",
    "source_zip_crc32": "119cc71d"
  },
  "diffusion_decoder.blocks.14.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 425348288,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      424815616,
      424821760
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1024",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.14.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 425354560,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      424821760,
      424827904
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1025",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.14.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 425360832,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      424827904,
      453139456
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1026",
    "source_zip_crc32": "4dbcf6c1"
  },
  "diffusion_decoder.blocks.14.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 453672512,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      453139456,
      453157888
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/1027",
    "source_zip_crc32": "58521164"
  },
  "diffusion_decoder.blocks.14.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 453691072,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      453157888,
      453158400
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/1028",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.14.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 453691712,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      453158400,
      453158912
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/1029",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.14.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 453696576,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.conditon_kv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      453158912,
      472033280
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1030",
    "source_zip_crc32": "5d279a95"
  },
  "diffusion_decoder.blocks.14.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 472571072,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.conditon_kv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      472033280,
      472045568
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/1031",
    "source_zip_crc32": "be3f80c7"
  },
  "diffusion_decoder.blocks.14.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 472583488,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      472045568,
      481482752
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1032",
    "source_zip_crc32": "ebc55527"
  },
  "diffusion_decoder.blocks.14.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 482020800,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      481482752,
      481488896
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1033",
    "source_zip_crc32": "68b0bb9e"
  },
  "diffusion_decoder.blocks.14.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 482027072,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      481488896,
      519237632
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1034",
    "source_zip_crc32": "6aa9e71b"
  },
  "diffusion_decoder.blocks.14.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 519775936,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      519237632,
      556986368
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1035",
    "source_zip_crc32": "7e0bd796"
  },
  "diffusion_decoder.blocks.14.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 557524800,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      556986368,
      594735104
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/1036",
    "source_zip_crc32": "35334c22"
  },
  "diffusion_decoder.blocks.14.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 595273664,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.adaLN_modulation.0.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      594735104,
      651358208
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1037",
    "source_zip_crc32": "fee1fe9a"
  },
  "diffusion_decoder.blocks.14.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.14.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 651896896,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.14.adaLN_modulation.0.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      651358208,
      651395072
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/1038",
    "source_zip_crc32": "a7b49e09"
  },
  "diffusion_decoder.blocks.15.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 651933888,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      651395072,
      651401216
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1039",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.15.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 651944384,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      651401216,
      651407360
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1040",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.15.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 651950656,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      651407360,
      679718912
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1041",
    "source_zip_crc32": "0630205e"
  },
  "diffusion_decoder.blocks.15.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 680262336,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      679718912,
      679737344
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/1042",
    "source_zip_crc32": "6df9c9dc"
  },
  "diffusion_decoder.blocks.15.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 680280896,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      679737344,
      679737856
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/1043",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.15.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 680281536,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      679737856,
      679738368
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/1044",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.15.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 680282176,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.conditon_kv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      679738368,
      698612736
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1045",
    "source_zip_crc32": "8f5ce67d"
  },
  "diffusion_decoder.blocks.15.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 699156672,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.conditon_kv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      698612736,
      698625024
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/1046",
    "source_zip_crc32": "6f9fa76b"
  },
  "diffusion_decoder.blocks.15.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 699169088,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      698625024,
      708062208
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1047",
    "source_zip_crc32": "ee848101"
  },
  "diffusion_decoder.blocks.15.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 708606400,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      708062208,
      708068352
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1048",
    "source_zip_crc32": "170730f0"
  },
  "diffusion_decoder.blocks.15.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 708612672,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      708068352,
      745817088
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1049",
    "source_zip_crc32": "76f37a31"
  },
  "diffusion_decoder.blocks.15.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 758944576,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      745817088,
      783565824
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1050",
    "source_zip_crc32": "c99ee048"
  },
  "diffusion_decoder.blocks.15.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 796693440,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      783565824,
      821314560
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/1051",
    "source_zip_crc32": "bf98cdc5"
  },
  "diffusion_decoder.blocks.15.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 834442304,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.adaLN_modulation.0.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      821314560,
      877937664
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1052",
    "source_zip_crc32": "53cde601"
  },
  "diffusion_decoder.blocks.15.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.15.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 891065536,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "blocks.15.adaLN_modulation.0.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      877937664,
      877974528
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/1053",
    "source_zip_crc32": "3efa3b22"
  },
  "diffusion_decoder.out_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.out_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 891102528,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "out_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      877974528,
      878367744
    ],
    "dtype": "float32",
    "shape": [
      64,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1054",
    "source_zip_crc32": "45dfa76a"
  },
  "diffusion_decoder.out_proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.out_proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 891495872,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "out_proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      878367744,
      878368000
    ],
    "dtype": "float32",
    "shape": [
      64
    ],
    "zip_record": "vlm_uniddt_512/data/1055",
    "source_zip_crc32": "8a77eb44"
  },
  "diffusion_decoder.t_embedder.mlp.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.t_embedder.mlp.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 891496256,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "t_embedder.mlp.0.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      878368000,
      879940864
    ],
    "dtype": "float32",
    "shape": [
      1536,
      256
    ],
    "zip_record": "vlm_uniddt_512/data/1056",
    "source_zip_crc32": "0fa00c6a"
  },
  "diffusion_decoder.t_embedder.mlp.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.t_embedder.mlp.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 893069248,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "t_embedder.mlp.0.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      879940864,
      879947008
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1057",
    "source_zip_crc32": "7d6508c0"
  },
  "diffusion_decoder.t_embedder.mlp.2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.t_embedder.mlp.2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 893075520,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "t_embedder.mlp.2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      879947008,
      889384192
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1058",
    "source_zip_crc32": "812501af"
  },
  "diffusion_decoder.t_embedder.mlp.2.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.t_embedder.mlp.2.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 902512832,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "t_embedder.mlp.2.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      889384192,
      889390336
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/1059",
    "source_zip_crc32": "b6b8f7f3"
  },
  "diffusion_decoder.deltas": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.deltas",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19113065920,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "deltas",
    "output_header_bytes": 11432,
    "output_offsets": [
      889390336,
      889390384
    ],
    "dtype": "float32",
    "shape": [
      3,
      1,
      4
    ],
    "zip_record": "vlm_uniddt_512/data/765",
    "source_zip_crc32": "da2fb1e3"
  },
  "diffusion_decoder.patch_embed.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.patch_embed.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19113066112,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "patch_embed.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      889390384,
      889783600
    ],
    "dtype": "float32",
    "shape": [
      1536,
      16,
      2,
      2
    ],
    "zip_record": "vlm_uniddt_512/data/766",
    "source_zip_crc32": "6315a56d"
  },
  "diffusion_decoder.patch_embed.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.patch_embed.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19113459456,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "patch_embed.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      889783600,
      889789744
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/767",
    "source_zip_crc32": "042ac913"
  },
  "diffusion_decoder.condition_proj.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_proj.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19113465728,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_proj.0.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      889789744,
      905518384
    ],
    "dtype": "float32",
    "shape": [
      1536,
      2560
    ],
    "zip_record": "vlm_uniddt_512/data/768",
    "source_zip_crc32": "8d445e5e"
  },
  "diffusion_decoder.condition_proj.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_proj.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19129194496,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_proj.0.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      905518384,
      905524528
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/769",
    "source_zip_crc32": "08b9aadf"
  },
  "diffusion_decoder.condition_refiner.0.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19141783808,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      905524528,
      905530672
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/770",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.0.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19141790080,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      905530672,
      905536816
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/771",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.0.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19141796352,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      905536816,
      933848368
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/772",
    "source_zip_crc32": "93e8c921"
  },
  "diffusion_decoder.condition_refiner.0.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19170108032,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      933848368,
      933866800
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/773",
    "source_zip_crc32": "88428b3b"
  },
  "diffusion_decoder.condition_refiner.0.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19170126592,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      933866800,
      933867312
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/774",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.0.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19170127232,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      933867312,
      933867824
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/775",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.0.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19170127872,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      933867824,
      943305008
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/776",
    "source_zip_crc32": "030c12b8"
  },
  "diffusion_decoder.condition_refiner.0.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19179565184,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      943305008,
      943311152
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/777",
    "source_zip_crc32": "50fb48be"
  },
  "diffusion_decoder.condition_refiner.0.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19179571456,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      943311152,
      981059888
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/778",
    "source_zip_crc32": "f87f1da8"
  },
  "diffusion_decoder.condition_refiner.0.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19217320320,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      981059888,
      1018808624
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/779",
    "source_zip_crc32": "2b7f4563"
  },
  "diffusion_decoder.condition_refiner.0.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.0.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19255081600,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.0.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1018808624,
      1056557360
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/780",
    "source_zip_crc32": "02ff3f78"
  },
  "diffusion_decoder.condition_refiner.1.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19292830464,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1056557360,
      1056563504
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/781",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.1.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19292836736,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1056563504,
      1056569648
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/782",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.1.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19292843008,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1056569648,
      1084881200
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/783",
    "source_zip_crc32": "32d89dda"
  },
  "diffusion_decoder.condition_refiner.1.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19321154688,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      1084881200,
      1084899632
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/784",
    "source_zip_crc32": "501ee62c"
  },
  "diffusion_decoder.condition_refiner.1.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19321173248,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1084899632,
      1084900144
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/785",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.1.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19321173888,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1084900144,
      1084900656
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/786",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.1.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19321174528,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1084900656,
      1094337840
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/787",
    "source_zip_crc32": "d70c55e8"
  },
  "diffusion_decoder.condition_refiner.1.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19330611840,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      1094337840,
      1094343984
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/788",
    "source_zip_crc32": "ccf9ab1b"
  },
  "diffusion_decoder.condition_refiner.1.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19330618112,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1094343984,
      1132092720
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/789",
    "source_zip_crc32": "70343675"
  },
  "diffusion_decoder.condition_refiner.1.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19372561408,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1132092720,
      1169841456
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/790",
    "source_zip_crc32": "72ba4006"
  },
  "diffusion_decoder.condition_refiner.1.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.1.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19410310272,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.1.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1169841456,
      1207590192
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/791",
    "source_zip_crc32": "5260470e"
  },
  "diffusion_decoder.condition_refiner.2.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19448059136,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1207590192,
      1207596336
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/792",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.2.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19448065408,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1207596336,
      1207602480
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/793",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.2.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19448071680,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1207602480,
      1235914032
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/794",
    "source_zip_crc32": "17abfb4f"
  },
  "diffusion_decoder.condition_refiner.2.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19476383360,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      1235914032,
      1235932464
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/795",
    "source_zip_crc32": "0cb6f561"
  },
  "diffusion_decoder.condition_refiner.2.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19476401920,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1235932464,
      1235932976
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/796",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.2.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19476402560,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1235932976,
      1235933488
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/797",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.2.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19476403200,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1235933488,
      1245370672
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/798",
    "source_zip_crc32": "934d7897"
  },
  "diffusion_decoder.condition_refiner.2.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19485840512,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      1245370672,
      1245376816
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/799",
    "source_zip_crc32": "c8fd76d5"
  },
  "diffusion_decoder.condition_refiner.2.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19485863424,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1245376816,
      1283125552
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/800",
    "source_zip_crc32": "ee7bd8aa"
  },
  "diffusion_decoder.condition_refiner.2.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19523612288,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1283125552,
      1320874288
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/801",
    "source_zip_crc32": "e22bac84"
  },
  "diffusion_decoder.condition_refiner.2.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.2.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19561361152,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.2.mlp.down_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1320874288,
      1358623024
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/802",
    "source_zip_crc32": "a6655061"
  },
  "diffusion_decoder.condition_refiner.3.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19599110016,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.norm1.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1358623024,
      1358629168
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/803",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.3.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19599116288,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.norm2.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1358629168,
      1358635312
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/804",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.condition_refiner.3.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19599122560,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.attn.qkv.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1358635312,
      1386946864
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/805",
    "source_zip_crc32": "22b5ba5d"
  },
  "diffusion_decoder.condition_refiner.3.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19627434240,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.attn.qkv.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      1386946864,
      1386965296
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/806",
    "source_zip_crc32": "04f9974b"
  },
  "diffusion_decoder.condition_refiner.3.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19627452800,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.attn.q_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1386965296,
      1386965808
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/807",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.3.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19627453440,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.attn.k_norm.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1386965808,
      1386966320
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/808",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.condition_refiner.3.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19627454080,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.attn.proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1386966320,
      1396403504
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/809",
    "source_zip_crc32": "a5571fb5"
  },
  "diffusion_decoder.condition_refiner.3.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19653668736,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.attn.proj.bias",
    "output_header_bytes": 11432,
    "output_offsets": [
      1396403504,
      1396409648
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/810",
    "source_zip_crc32": "fe6f5c0e"
  },
  "diffusion_decoder.condition_refiner.3.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19653675008,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.mlp.gate_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1396409648,
      1434158384
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/811",
    "source_zip_crc32": "5fee6783"
  },
  "diffusion_decoder.condition_refiner.3.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19691423872,
    "output_file": "weights/diffusion_decoder/model-00001-of-00003.safetensors",
    "output_key": "condition_refiner.3.mlp.up_proj.weight",
    "output_header_bytes": 11432,
    "output_offsets": [
      1434158384,
      1471907120
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/812",
    "source_zip_crc32": "c51f0392"
  },
  "diffusion_decoder.condition_refiner.3.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.condition_refiner.3.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19729172736,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "condition_refiner.3.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      0,
      37748736
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/813",
    "source_zip_crc32": "6a2f8585"
  },
  "diffusion_decoder.blocks.0.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19766921600,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      37748736,
      37754880
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/814",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.0.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19766927872,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      37754880,
      37761024
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/815",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.0.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19766934144,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      37761024,
      66072576
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/816",
    "source_zip_crc32": "91e57208"
  },
  "diffusion_decoder.blocks.0.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19795245824,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      66072576,
      66091008
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/817",
    "source_zip_crc32": "ffbe4e18"
  },
  "diffusion_decoder.blocks.0.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19795264384,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      66091008,
      66091520
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/818",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.0.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19795265024,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      66091520,
      66092032
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/819",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.0.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19795282176,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      66092032,
      84966400
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/820",
    "source_zip_crc32": "b2b4a586"
  },
  "diffusion_decoder.blocks.0.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19814156672,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      84966400,
      84978688
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/821",
    "source_zip_crc32": "204b8a69"
  },
  "diffusion_decoder.blocks.0.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19814169088,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      84978688,
      94415872
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/822",
    "source_zip_crc32": "f7d6e4b0"
  },
  "diffusion_decoder.blocks.0.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19823606400,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      94415872,
      94422016
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/823",
    "source_zip_crc32": "bd8debd0"
  },
  "diffusion_decoder.blocks.0.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19823612672,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      94422016,
      132170752
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/824",
    "source_zip_crc32": "b5700c2d"
  },
  "diffusion_decoder.blocks.0.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19861361536,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.mlp.up_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      132170752,
      169919488
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/825",
    "source_zip_crc32": "b40a5055"
  },
  "diffusion_decoder.blocks.0.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19899110400,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      169919488,
      207668224
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/826",
    "source_zip_crc32": "62c7056d"
  },
  "diffusion_decoder.blocks.0.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19936859264,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.adaLN_modulation.0.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      207668224,
      264291328
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/827",
    "source_zip_crc32": "e6603973"
  },
  "diffusion_decoder.blocks.0.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.0.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19993482496,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.0.adaLN_modulation.0.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      264291328,
      264328192
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/828",
    "source_zip_crc32": "6573b9c3"
  },
  "diffusion_decoder.blocks.1.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 19993519488,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      264328192,
      264334336
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/829",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.1.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20010303104,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      264334336,
      264340480
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/830",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.1.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20010309376,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      264340480,
      292652032
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/831",
    "source_zip_crc32": "ef09a755"
  },
  "diffusion_decoder.blocks.1.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20038621056,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      292652032,
      292670464
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/832",
    "source_zip_crc32": "f43d89f5"
  },
  "diffusion_decoder.blocks.1.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20038639616,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      292670464,
      292670976
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/833",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.1.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20038640256,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      292670976,
      292671488
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/834",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.1.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20038640896,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      292671488,
      311545856
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/835",
    "source_zip_crc32": "9a9d7449"
  },
  "diffusion_decoder.blocks.1.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20057515392,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      311545856,
      311558144
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/836",
    "source_zip_crc32": "a51e7b8c"
  },
  "diffusion_decoder.blocks.1.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20057527808,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      311558144,
      320995328
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/837",
    "source_zip_crc32": "addc4aae"
  },
  "diffusion_decoder.blocks.1.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20066965120,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      320995328,
      321001472
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/838",
    "source_zip_crc32": "9e7657fb"
  },
  "diffusion_decoder.blocks.1.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20066971392,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      321001472,
      358750208
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/839",
    "source_zip_crc32": "988f1fd5"
  },
  "diffusion_decoder.blocks.1.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20104724480,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.mlp.up_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      358750208,
      396498944
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/840",
    "source_zip_crc32": "3d861ee6"
  },
  "diffusion_decoder.blocks.1.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20142473344,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      396498944,
      434247680
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/841",
    "source_zip_crc32": "763329ab"
  },
  "diffusion_decoder.blocks.1.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20180222208,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.adaLN_modulation.0.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      434247680,
      490870784
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/842",
    "source_zip_crc32": "1acfd04c"
  },
  "diffusion_decoder.blocks.1.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.1.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20236845440,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.1.adaLN_modulation.0.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      490870784,
      490907648
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/843",
    "source_zip_crc32": "50aa35d4"
  },
  "diffusion_decoder.blocks.2.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20236882432,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      490907648,
      490913792
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/844",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.2.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20236888704,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      490913792,
      490919936
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/845",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.2.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20236894976,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      490919936,
      519231488
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/846",
    "source_zip_crc32": "96f1b05e"
  },
  "diffusion_decoder.blocks.2.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20265206656,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      519231488,
      519249920
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/847",
    "source_zip_crc32": "805deb7d"
  },
  "diffusion_decoder.blocks.2.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20265225216,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      519249920,
      519250432
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/848",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.2.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20265225856,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      519250432,
      519250944
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/849",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.2.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20290392448,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      519250944,
      538125312
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/850",
    "source_zip_crc32": "0416fd7e"
  },
  "diffusion_decoder.blocks.2.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20309266944,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      538125312,
      538137600
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/851",
    "source_zip_crc32": "9d9dd181"
  },
  "diffusion_decoder.blocks.2.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20309279360,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      538137600,
      547574784
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/852",
    "source_zip_crc32": "50634453"
  },
  "diffusion_decoder.blocks.2.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20318716672,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      547574784,
      547580928
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/853",
    "source_zip_crc32": "36d01d28"
  },
  "diffusion_decoder.blocks.2.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20318722944,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      547580928,
      585329664
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/854",
    "source_zip_crc32": "07681ac1"
  },
  "diffusion_decoder.blocks.2.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20356471808,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.mlp.up_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      585329664,
      623078400
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/855",
    "source_zip_crc32": "c359a343"
  },
  "diffusion_decoder.blocks.2.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20394220672,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      623078400,
      660827136
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/856",
    "source_zip_crc32": "45c92663"
  },
  "diffusion_decoder.blocks.2.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20431969536,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.adaLN_modulation.0.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      660827136,
      717450240
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/857",
    "source_zip_crc32": "17069dc6"
  },
  "diffusion_decoder.blocks.2.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.2.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20488592768,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.2.adaLN_modulation.0.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      717450240,
      717487104
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/858",
    "source_zip_crc32": "e1d54855"
  },
  "diffusion_decoder.blocks.3.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20488629760,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      717487104,
      717493248
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/859",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.3.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20488660736,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      717493248,
      717499392
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/860",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.3.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20488667008,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      717499392,
      745810944
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/861",
    "source_zip_crc32": "dc9366a6"
  },
  "diffusion_decoder.blocks.3.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20516978688,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      745810944,
      745829376
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/862",
    "source_zip_crc32": "8d7fe6df"
  },
  "diffusion_decoder.blocks.3.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20516997248,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      745829376,
      745829888
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/863",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.3.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20516997888,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      745829888,
      745830400
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/864",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.3.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20516998528,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      745830400,
      764704768
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/865",
    "source_zip_crc32": "c9280c36"
  },
  "diffusion_decoder.blocks.3.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20535873024,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      764704768,
      764717056
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/866",
    "source_zip_crc32": "711ea872"
  },
  "diffusion_decoder.blocks.3.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20535885440,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      764717056,
      774154240
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/867",
    "source_zip_crc32": "82429553"
  },
  "diffusion_decoder.blocks.3.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20545322752,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      774154240,
      774160384
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/868",
    "source_zip_crc32": "ab6c6d9a"
  },
  "diffusion_decoder.blocks.3.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20545329024,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      774160384,
      811909120
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/869",
    "source_zip_crc32": "5f71bd43"
  },
  "diffusion_decoder.blocks.3.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20583082112,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.mlp.up_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      811909120,
      849657856
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/870",
    "source_zip_crc32": "f90bd84a"
  },
  "diffusion_decoder.blocks.3.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20620830976,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      849657856,
      887406592
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/871",
    "source_zip_crc32": "93fbe8f3"
  },
  "diffusion_decoder.blocks.3.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20658579840,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.adaLN_modulation.0.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      887406592,
      944029696
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/872",
    "source_zip_crc32": "257326af"
  },
  "diffusion_decoder.blocks.3.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.3.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20715203072,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.3.adaLN_modulation.0.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      944029696,
      944066560
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/873",
    "source_zip_crc32": "03b08921"
  },
  "diffusion_decoder.blocks.4.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20715240064,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      944066560,
      944072704
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/874",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.4.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20715246336,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      944072704,
      944078848
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/875",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.4.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20715252608,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      944078848,
      972390400
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/876",
    "source_zip_crc32": "91ed505b"
  },
  "diffusion_decoder.blocks.4.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20743564288,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      972390400,
      972408832
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/877",
    "source_zip_crc32": "348f5942"
  },
  "diffusion_decoder.blocks.4.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20743582848,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      972408832,
      972409344
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/878",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.4.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20743583488,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      972409344,
      972409856
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/879",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.4.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20743588352,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      972409856,
      991284224
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/880",
    "source_zip_crc32": "c413d777"
  },
  "diffusion_decoder.blocks.4.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20762462848,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      991284224,
      991296512
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/881",
    "source_zip_crc32": "5f312dc0"
  },
  "diffusion_decoder.blocks.4.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20762475264,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      991296512,
      1000733696
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/882",
    "source_zip_crc32": "c4034960"
  },
  "diffusion_decoder.blocks.4.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20771912576,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1000733696,
      1000739840
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/883",
    "source_zip_crc32": "af482c5f"
  },
  "diffusion_decoder.blocks.4.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20771918848,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1000739840,
      1038488576
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/884",
    "source_zip_crc32": "a7a69967"
  },
  "diffusion_decoder.blocks.4.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20809667712,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.mlp.up_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1038488576,
      1076237312
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/885",
    "source_zip_crc32": "14fa3248"
  },
  "diffusion_decoder.blocks.4.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20847416576,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1076237312,
      1113986048
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/886",
    "source_zip_crc32": "1a26ea03"
  },
  "diffusion_decoder.blocks.4.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20885165440,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.adaLN_modulation.0.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1113986048,
      1170609152
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/887",
    "source_zip_crc32": "6c490036"
  },
  "diffusion_decoder.blocks.4.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.4.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20941788672,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.4.adaLN_modulation.0.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1170609152,
      1170646016
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/888",
    "source_zip_crc32": "2b0b2a5c"
  },
  "diffusion_decoder.blocks.5.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20941825664,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1170646016,
      1170652160
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/889",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.5.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20941836160,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1170652160,
      1170658304
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/890",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.5.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20941842432,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1170658304,
      1198969856
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/891",
    "source_zip_crc32": "8a6c06c6"
  },
  "diffusion_decoder.blocks.5.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20970154112,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1198969856,
      1198988288
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/892",
    "source_zip_crc32": "36232635"
  },
  "diffusion_decoder.blocks.5.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20970172672,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1198988288,
      1198988800
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/893",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.5.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20970173312,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1198988800,
      1198989312
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/894",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.5.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20970173952,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1198989312,
      1217863680
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/895",
    "source_zip_crc32": "8274cd2d"
  },
  "diffusion_decoder.blocks.5.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20989048448,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1217863680,
      1217875968
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/896",
    "source_zip_crc32": "abb45e3a"
  },
  "diffusion_decoder.blocks.5.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20989060864,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1217875968,
      1227313152
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/897",
    "source_zip_crc32": "6537df86"
  },
  "diffusion_decoder.blocks.5.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20998498176,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1227313152,
      1227319296
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/898",
    "source_zip_crc32": "ff391f15"
  },
  "diffusion_decoder.blocks.5.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 20998504448,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1227319296,
      1265068032
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/899",
    "source_zip_crc32": "4de51489"
  },
  "diffusion_decoder.blocks.5.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21040451968,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.mlp.up_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1265068032,
      1302816768
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/900",
    "source_zip_crc32": "8c3101dd"
  },
  "diffusion_decoder.blocks.5.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21078200832,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.mlp.down_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1302816768,
      1340565504
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/901",
    "source_zip_crc32": "55cdec0a"
  },
  "diffusion_decoder.blocks.5.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21115949696,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.adaLN_modulation.0.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1340565504,
      1397188608
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/902",
    "source_zip_crc32": "bc3e5b26"
  },
  "diffusion_decoder.blocks.5.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.5.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21172572928,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.5.adaLN_modulation.0.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1397188608,
      1397225472
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/903",
    "source_zip_crc32": "dfa31a49"
  },
  "diffusion_decoder.blocks.6.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21172609920,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.norm1.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1397225472,
      1397231616
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/904",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.6.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21172616192,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.norm2.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1397231616,
      1397237760
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/905",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.6.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21172622464,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.qkv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1397237760,
      1425549312
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/906",
    "source_zip_crc32": "f179d6c6"
  },
  "diffusion_decoder.blocks.6.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21200934144,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.qkv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1425549312,
      1425567744
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/907",
    "source_zip_crc32": "4015bf6d"
  },
  "diffusion_decoder.blocks.6.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21200952704,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.q_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1425567744,
      1425568256
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/908",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.6.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21200953344,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.k_norm.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1425568256,
      1425568768
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/909",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.6.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21213537024,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.conditon_kv.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1425568768,
      1444443136
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/910",
    "source_zip_crc32": "3d420cc3"
  },
  "diffusion_decoder.blocks.6.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21232411520,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.conditon_kv.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1444443136,
      1444455424
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/911",
    "source_zip_crc32": "faf36e8e"
  },
  "diffusion_decoder.blocks.6.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21232423936,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1444455424,
      1453892608
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/912",
    "source_zip_crc32": "b887b1eb"
  },
  "diffusion_decoder.blocks.6.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21241861248,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.attn.proj.bias",
    "output_header_bytes": 10432,
    "output_offsets": [
      1453892608,
      1453898752
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/913",
    "source_zip_crc32": "c8d8f155"
  },
  "diffusion_decoder.blocks.6.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21241867520,
    "output_file": "weights/diffusion_decoder/model-00002-of-00003.safetensors",
    "output_key": "blocks.6.mlp.gate_proj.weight",
    "output_header_bytes": 10432,
    "output_offsets": [
      1453898752,
      1491647488
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/914",
    "source_zip_crc32": "cefede14"
  },
  "diffusion_decoder.blocks.6.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21279616384,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.6.mlp.up_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      0,
      37748736
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/915",
    "source_zip_crc32": "009c2820"
  },
  "diffusion_decoder.blocks.6.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21317365248,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.6.mlp.down_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      37748736,
      75497472
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/916",
    "source_zip_crc32": "3c61a25b"
  },
  "diffusion_decoder.blocks.6.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21355114112,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.6.adaLN_modulation.0.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      75497472,
      132120576
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/917",
    "source_zip_crc32": "8fa53a6e"
  },
  "diffusion_decoder.blocks.6.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.6.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21411737344,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.6.adaLN_modulation.0.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      132120576,
      132157440
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/918",
    "source_zip_crc32": "8772a471"
  },
  "diffusion_decoder.blocks.7.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21411774336,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.norm1.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      132157440,
      132163584
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/919",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.7.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21411793024,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.norm2.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      132163584,
      132169728
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/920",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.7.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21411799296,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.qkv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      132169728,
      160481280
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/921",
    "source_zip_crc32": "06030de8"
  },
  "diffusion_decoder.blocks.7.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21440110976,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.qkv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      160481280,
      160499712
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/922",
    "source_zip_crc32": "0eea7619"
  },
  "diffusion_decoder.blocks.7.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21440129536,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.q_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      160499712,
      160500224
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/923",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.7.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21440130176,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.k_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      160500224,
      160500736
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/924",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.7.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21440130816,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.conditon_kv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      160500736,
      179375104
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/925",
    "source_zip_crc32": "4509b684"
  },
  "diffusion_decoder.blocks.7.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21459005312,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.conditon_kv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      179375104,
      179387392
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/926",
    "source_zip_crc32": "7eb7c1b0"
  },
  "diffusion_decoder.blocks.7.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21459017728,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      179387392,
      188824576
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/927",
    "source_zip_crc32": "b6578943"
  },
  "diffusion_decoder.blocks.7.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21468455040,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.attn.proj.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      188824576,
      188830720
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/928",
    "source_zip_crc32": "4ce83fc4"
  },
  "diffusion_decoder.blocks.7.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21468461312,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.mlp.gate_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      188830720,
      226579456
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/929",
    "source_zip_crc32": "568799a2"
  },
  "diffusion_decoder.blocks.7.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21510404608,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.mlp.up_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      226579456,
      264328192
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/930",
    "source_zip_crc32": "5b6e72aa"
  },
  "diffusion_decoder.blocks.7.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21548153472,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.mlp.down_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      264328192,
      302076928
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/931",
    "source_zip_crc32": "a38673dd"
  },
  "diffusion_decoder.blocks.7.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21585902336,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.adaLN_modulation.0.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      302076928,
      358700032
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/932",
    "source_zip_crc32": "7783833c"
  },
  "diffusion_decoder.blocks.7.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.7.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21642525568,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.7.adaLN_modulation.0.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      358700032,
      358736896
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/933",
    "source_zip_crc32": "1e66a677"
  },
  "diffusion_decoder.blocks.8.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21642562560,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.norm1.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      358736896,
      358743040
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/934",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.8.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21642568832,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.norm2.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      358743040,
      358749184
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/935",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.8.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21642575104,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.qkv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      358749184,
      387060736
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/936",
    "source_zip_crc32": "dc527bcc"
  },
  "diffusion_decoder.blocks.8.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21670886784,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.qkv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      387060736,
      387079168
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/937",
    "source_zip_crc32": "4844e1f8"
  },
  "diffusion_decoder.blocks.8.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21670905344,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.q_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      387079168,
      387079680
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/938",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.8.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21670905984,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.k_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      387079680,
      387080192
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/939",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.8.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21670910848,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.conditon_kv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      387080192,
      405954560
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/940",
    "source_zip_crc32": "ef1f6c61"
  },
  "diffusion_decoder.blocks.8.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21689785344,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.conditon_kv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      405954560,
      405966848
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/941",
    "source_zip_crc32": "16e21b09"
  },
  "diffusion_decoder.blocks.8.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21689797760,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      405966848,
      415404032
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/942",
    "source_zip_crc32": "8fa2ab13"
  },
  "diffusion_decoder.blocks.8.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21699235072,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.attn.proj.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      415404032,
      415410176
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/943",
    "source_zip_crc32": "067d3cdd"
  },
  "diffusion_decoder.blocks.8.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21699241344,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.mlp.gate_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      415410176,
      453158912
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/944",
    "source_zip_crc32": "1cdc83e6"
  },
  "diffusion_decoder.blocks.8.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21736990208,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.mlp.up_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      453158912,
      490907648
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/945",
    "source_zip_crc32": "93cec26f"
  },
  "diffusion_decoder.blocks.8.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21774739072,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.mlp.down_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      490907648,
      528656384
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/946",
    "source_zip_crc32": "2cda5b07"
  },
  "diffusion_decoder.blocks.8.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21812487936,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.adaLN_modulation.0.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      528656384,
      585279488
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/947",
    "source_zip_crc32": "7e9af0b1"
  },
  "diffusion_decoder.blocks.8.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.8.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21869111168,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.8.adaLN_modulation.0.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      585279488,
      585316352
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/948",
    "source_zip_crc32": "3883096e"
  },
  "diffusion_decoder.blocks.9.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21869148160,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.norm1.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      585316352,
      585322496
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/949",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.9.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21885931776,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.norm2.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      585322496,
      585328640
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/950",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.9.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21885938048,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.qkv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      585328640,
      613640192
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/951",
    "source_zip_crc32": "dfa179ae"
  },
  "diffusion_decoder.blocks.9.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21914249728,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.qkv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      613640192,
      613658624
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/952",
    "source_zip_crc32": "67371871"
  },
  "diffusion_decoder.blocks.9.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21914268288,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.q_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      613658624,
      613659136
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/953",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.9.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21914268928,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.k_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      613659136,
      613659648
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/954",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.9.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21914269568,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.conditon_kv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      613659648,
      632534016
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/955",
    "source_zip_crc32": "d784c20c"
  },
  "diffusion_decoder.blocks.9.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21933144064,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.conditon_kv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      632534016,
      632546304
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/956",
    "source_zip_crc32": "5d2a29c2"
  },
  "diffusion_decoder.blocks.9.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21933156480,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      632546304,
      641983488
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/957",
    "source_zip_crc32": "87c36bac"
  },
  "diffusion_decoder.blocks.9.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21942593792,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.attn.proj.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      641983488,
      641989632
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/958",
    "source_zip_crc32": "043fe9f4"
  },
  "diffusion_decoder.blocks.9.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21942600064,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.mlp.gate_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      641989632,
      679738368
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/959",
    "source_zip_crc32": "c06b18e1"
  },
  "diffusion_decoder.blocks.9.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 21980365440,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.mlp.up_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      679738368,
      717487104
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/960",
    "source_zip_crc32": "f9f4142f"
  },
  "diffusion_decoder.blocks.9.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22018114304,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.mlp.down_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      717487104,
      755235840
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/961",
    "source_zip_crc32": "bddcc085"
  },
  "diffusion_decoder.blocks.9.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22055863168,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.adaLN_modulation.0.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      755235840,
      811858944
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/962",
    "source_zip_crc32": "7aa965bb"
  },
  "diffusion_decoder.blocks.9.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.9.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22112486400,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.9.adaLN_modulation.0.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      811858944,
      811895808
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/963",
    "source_zip_crc32": "603a5582"
  },
  "diffusion_decoder.blocks.10.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22112523392,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.norm1.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      811895808,
      811901952
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/964",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.10.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22112529664,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.norm2.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      811901952,
      811908096
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/965",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.10.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22112535936,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.qkv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      811908096,
      840219648
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/966",
    "source_zip_crc32": "b53bff16"
  },
  "diffusion_decoder.blocks.10.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22140847616,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.qkv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      840219648,
      840238080
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/967",
    "source_zip_crc32": "cdeccc30"
  },
  "diffusion_decoder.blocks.10.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22140866176,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.q_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      840238080,
      840238592
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/968",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.10.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22140866816,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.k_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      840238592,
      840239104
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/969",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.10.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22157644800,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.conditon_kv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      840239104,
      859113472
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/970",
    "source_zip_crc32": "53cee10d"
  },
  "diffusion_decoder.blocks.10.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22176519296,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.conditon_kv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      859113472,
      859125760
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/971",
    "source_zip_crc32": "37778323"
  },
  "diffusion_decoder.blocks.10.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22176531712,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      859125760,
      868562944
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/972",
    "source_zip_crc32": "fe877e2e"
  },
  "diffusion_decoder.blocks.10.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22185969024,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.attn.proj.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      868562944,
      868569088
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/973",
    "source_zip_crc32": "2747edc1"
  },
  "diffusion_decoder.blocks.10.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22185975296,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.mlp.gate_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      868569088,
      906317824
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/974",
    "source_zip_crc32": "4cff3b34"
  },
  "diffusion_decoder.blocks.10.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22223724160,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.mlp.up_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      906317824,
      944066560
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/975",
    "source_zip_crc32": "b199e61c"
  },
  "diffusion_decoder.blocks.10.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22261473024,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.mlp.down_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      944066560,
      981815296
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/976",
    "source_zip_crc32": "a2c8ad77"
  },
  "diffusion_decoder.blocks.10.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22299221888,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.adaLN_modulation.0.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      981815296,
      1038438400
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/977",
    "source_zip_crc32": "6e06adc3"
  },
  "diffusion_decoder.blocks.10.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.10.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22355845120,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.10.adaLN_modulation.0.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      1038438400,
      1038475264
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/978",
    "source_zip_crc32": "1f75bb20"
  },
  "diffusion_decoder.blocks.11.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22355882112,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.norm1.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1038475264,
      1038481408
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/979",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.11.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22355892608,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.norm2.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1038481408,
      1038487552
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/980",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.11.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22355898880,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.qkv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1038487552,
      1066799104
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/981",
    "source_zip_crc32": "5f8a62ad"
  },
  "diffusion_decoder.blocks.11.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22384210560,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.qkv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      1066799104,
      1066817536
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/982",
    "source_zip_crc32": "19be4f16"
  },
  "diffusion_decoder.blocks.11.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22384229120,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.q_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1066817536,
      1066818048
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/983",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.11.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22384229760,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.k_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1066818048,
      1066818560
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/984",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.11.attn.conditon_kv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.conditon_kv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22384230400,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.conditon_kv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1066818560,
      1085692928
    ],
    "dtype": "float32",
    "shape": [
      3072,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/985",
    "source_zip_crc32": "724490e6"
  },
  "diffusion_decoder.blocks.11.attn.conditon_kv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.conditon_kv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22403104896,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.conditon_kv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      1085692928,
      1085705216
    ],
    "dtype": "float32",
    "shape": [
      3072
    ],
    "zip_record": "vlm_uniddt_512/data/986",
    "source_zip_crc32": "19f09b8b"
  },
  "diffusion_decoder.blocks.11.attn.proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22403117312,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1085705216,
      1095142400
    ],
    "dtype": "float32",
    "shape": [
      1536,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/987",
    "source_zip_crc32": "d9688611"
  },
  "diffusion_decoder.blocks.11.attn.proj.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.attn.proj.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22412554624,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.attn.proj.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      1095142400,
      1095148544
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/988",
    "source_zip_crc32": "ab40d33e"
  },
  "diffusion_decoder.blocks.11.mlp.gate_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.mlp.gate_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22412560896,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.mlp.gate_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1095148544,
      1132897280
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/989",
    "source_zip_crc32": "96469a4a"
  },
  "diffusion_decoder.blocks.11.mlp.up_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.mlp.up_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22475475712,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.mlp.up_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1132897280,
      1170646016
    ],
    "dtype": "float32",
    "shape": [
      6144,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/990",
    "source_zip_crc32": "6230a253"
  },
  "diffusion_decoder.blocks.11.mlp.down_proj.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.mlp.down_proj.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22513224576,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.mlp.down_proj.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1170646016,
      1208394752
    ],
    "dtype": "float32",
    "shape": [
      1536,
      6144
    ],
    "zip_record": "vlm_uniddt_512/data/991",
    "source_zip_crc32": "6a3e7250"
  },
  "diffusion_decoder.blocks.11.adaLN_modulation.0.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.adaLN_modulation.0.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22550973440,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.adaLN_modulation.0.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1208394752,
      1265017856
    ],
    "dtype": "float32",
    "shape": [
      9216,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/992",
    "source_zip_crc32": "2950f139"
  },
  "diffusion_decoder.blocks.11.adaLN_modulation.0.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.11.adaLN_modulation.0.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22607596672,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.11.adaLN_modulation.0.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      1265017856,
      1265054720
    ],
    "dtype": "float32",
    "shape": [
      9216
    ],
    "zip_record": "vlm_uniddt_512/data/993",
    "source_zip_crc32": "62460210"
  },
  "diffusion_decoder.blocks.12.norm1.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.norm1.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22607633664,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.12.norm1.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1265054720,
      1265060864
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/994",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.12.norm2.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.norm2.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22607639936,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.12.norm2.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1265060864,
      1265067008
    ],
    "dtype": "float32",
    "shape": [
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/995",
    "source_zip_crc32": "af9a737c"
  },
  "diffusion_decoder.blocks.12.attn.qkv.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.qkv.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22607646208,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.12.attn.qkv.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1265067008,
      1293378560
    ],
    "dtype": "float32",
    "shape": [
      4608,
      1536
    ],
    "zip_record": "vlm_uniddt_512/data/996",
    "source_zip_crc32": "027cb36b"
  },
  "diffusion_decoder.blocks.12.attn.qkv.bias": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.qkv.bias",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22635957888,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.12.attn.qkv.bias",
    "output_header_bytes": 8760,
    "output_offsets": [
      1293378560,
      1293396992
    ],
    "dtype": "float32",
    "shape": [
      4608
    ],
    "zip_record": "vlm_uniddt_512/data/997",
    "source_zip_crc32": "5cbf6be8"
  },
  "diffusion_decoder.blocks.12.attn.q_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.q_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22635976448,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.12.attn.q_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1293396992,
      1293397504
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/998",
    "source_zip_crc32": "32b9c3f4"
  },
  "diffusion_decoder.blocks.12.attn.k_norm.weight": {
    "source_key": "state_dict.ema_denoiser.diffusion_decoder.blocks.12.attn.k_norm.weight",
    "source_file": "vlm_uniddt_512.ckpt",
    "source_byte_offset": 22635977088,
    "output_file": "weights/diffusion_decoder/model-00003-of-00003.safetensors",
    "output_key": "blocks.12.attn.k_norm.weight",
    "output_header_bytes": 8760,
    "output_offsets": [
      1293397504,
      1293398016
    ],
    "dtype": "float32",
    "shape": [
      128
    ],
    "zip_record": "vlm_uniddt_512/data/999",
    "source_zip_crc32": "32b9c3f4"
  }
}