{
  "_class_name": "MochiTransformer3DModel",
  "_diffusers_version": "0.32.0.dev0",
  "_name_or_path": "mochi-1-diffusers",
  "activation_fn": "swiglu",
  "attention_head_dim": 128,
  "in_channels": 12,
  "max_sequence_length": 256,
  "num_attention_heads": 24,
  "num_layers": 48,
  "out_channels": null,
  "patch_size": 2,
  "pooled_projection_dim": 1536,
  "qk_norm": "rms_norm",
  "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": false,
    "llm_int8_enable_fp32_cpu_offload": false,
    "llm_int8_has_fp16_weight": false,
    "llm_int8_skip_modules": [
      "final_layer",
      "x_embedder.proj",
      "t_embedder",
      "pos_frequencies",
      "t5"
    ],
    "llm_int8_threshold": 6.0,
    "load_in_4bit": true,
    "load_in_8bit": false,
    "quant_method": "bitsandbytes"
  },
  "text_embed_dim": 4096,
  "time_embed_dim": 256
}