model_index.json

{
  "_class_name": "IMProvPipeline",
  "_diffusers_version": "0.24.0",
  "_name_or_path": "/shared/jiarui/results/video-diffusers/output/mae_p16_text_ed_enc_cat_n1024_arxiv_l400m_224_600k_bs1024x8/amp/hf_pipeline",
  "image_decoder": [
    "improv.modeling.meta_arch.modules",
    "VQDecoder"
  ],
  "image_encoder": [
    "improv.modeling.meta_arch.modules",
    "PatchEmbedEncoder"
  ],
  "mask_image_model": [
    "improv.modeling.meta_arch.modules",
    "TransformerMAE"
  ],
  "scheduler": [
    "improv.pipelines.pipeline_improv",
    "MAEScheduler"
  ],
  "text_encoder": [
    "transformers",
    "CLIPTextModel"
  ],
  "tokenizer": [
    "transformers",
    "CLIPTokenizer"
  ]
}