{
"dataset_param": {
"dataset_type": "i2v",
"basic_parameters": {
"data_path": "./examples/opensoraplan1.3/data.txt",
"data_storage_mode": "combine"
},
"preprocess_parameters": {
"video_processor_type": "OpensoraplanVideoProcessor",
"video_reader_type": "DecordVideo",
"image_reader_type": "Image",
"num_frames": 93,
"min_num_frames": 29,
"frame_interval": 1,
"max_hxw": 236544,
"min_hxw": 102400,
"train_fps": 18,
"speed_factor": 1.0,
"drop_short_ratio": 0.0,
"cfg": 0.1,
"hw_stride": 32,
"vae_scale_factor": [4,8,8],
"force_resolution": false,
"seed": 1234,
"hw_aspect_thr": 2.0,
"use_aesthetic": true,
"train_pipeline": {
"video": [
{
"trans_type": "MaxHWResizeVideo",
"param": {
"align_corners": true,
"antialias": true,
"transform_size": "auto"
}
},
{
"trans_type": "SpatialStrideCropVideo",
"param": {
"stride": 32
}
}
]
},
"train_pipeline_after_resize": {
"video": [
{
"trans_type": "ToTensorAfterResize"
},
{
"trans_type": "ae_norm"
}
]
}
},
"use_text_processer": true,
"tokenizer_config": {
"hub_backend": "hf",
"autotokenizer_name": "AutoTokenizer",
"from_pretrained": "your model path",
"model_max_length": 512
},
"mask_type_ratio_dict_video": {
"i2v": 1
}
},
"dataloader_param": {
"dataloader_mode": "sampler",
"sampler_type": "LengthGroupedSampler",
"shuffle": true,
"drop_last": true,
"pin_memory": false,
"group_frame": false,
"group_resolution": false,
"group_data": true,
"initial_global_step_for_sampler": 0,
"collate_param": {
"model_name": "open_sora_plan"
}
}
}