{
  "_class_name": "ErnieImageTransformer2DModel",
  "_diffusers_version": "0.38.0.dev0",
  "_name_or_path": "./tmp_model/transformer",
  "eps": 1e-06,
  "ffn_hidden_size": 12288,
  "hidden_size": 4096,
  "in_channels": 128,
  "num_attention_heads": 32,
  "num_layers": 36,
  "out_channels": 128,
  "patch_size": 1,
  "qk_layernorm": true,
  "quantization_config": {
    "_load_in_4bit": true,
    "_load_in_8bit": false,
    "bnb_4bit_compute_dtype": "bfloat16",
    "bnb_4bit_quant_storage": "uint8",
    "bnb_4bit_quant_type": "nf4",
    "bnb_4bit_use_double_quant": true,
    "llm_int8_enable_fp32_cpu_offload": false,
    "llm_int8_has_fp16_weight": false,
    "llm_int8_skip_modules": [
      "layers.0.self_attention.to_q",
      "layers.0.self_attention.to_k",
      "layers.0.self_attention.to_v",
      "layers.0.self_attention.to_out.0",
      "layers.0.mlp.gate_proj",
      "layers.0.mlp.up_proj",
      "layers.0.mlp.linear_fc2",
      "layers.35.self_attention.to_q",
      "layers.35.self_attention.to_k",
      "layers.35.self_attention.to_v",
      "layers.35.self_attention.to_out.0",
      "layers.35.mlp.gate_proj",
      "layers.35.mlp.up_proj",
      "layers.35.mlp.linear_fc2",
      "text_proj",
      "time_embedding.linear_1",
      "time_embedding.linear_2",
      "adaLN_modulation.1",
      "final_norm.linear",
      "final_linear"
    ],
    "llm_int8_threshold": 6.0,
    "load_in_4bit": true,
    "load_in_8bit": false,
    "quant_method": "bitsandbytes"
  },
  "rope_axes_dim": [
    32,
    48,
    48
  ],
  "rope_theta": 256,
  "text_in_dim": 3072
}