{
"dataset_param": {
"dataset_type": "feature",
"basic_parameters": {
"data_path": "./sora_features/data.jsonl",
"data_folder": "./sora_features/"
},
"preprocess_parameters": {
"video_processor_type": "OpensoraplanVideoProcessor",
"video_reader_type": "DecordVideo",
"image_reader_type": "Image",
"num_frames": 45,
"min_num_frames": 45,
"frame_interval": 1,
"max_height": 720,
"max_width": 1280,
"train_fps": 16,
"speed_factor": 1.0,
"drop_short_ratio": 0.0,
"cfg": 0.1,
"hw_stride": 32,
"vae_scale_factor": [4,8,8],
"force_resolution": true,
"seed": 42,
"hw_aspect_thr": 2.0,
"use_aesthetic": true,
"train_pipeline": {
"video": [
{
"trans_type": "ToTensorVideo"
},
{
"trans_type": "CenterCropResizeVideo",
"param": {
"use_short_edge": false,
"top_crop": false,
"align_corners": true,
"antialias": true,
"transform_size": "auto"
}
},
{
"trans_type": "ae_norm"
}
],
"image": [
{
"trans_type": "ToTensorVideo"
},
{
"trans_type": "CenterCropResizeVideo",
"param": {
"use_short_edge": false,
"top_crop": true,
"align_corners": true,
"antialias": true,
"transform_size": "auto"
}
},
{
"trans_type": "ae_norm"
}
]
}
},
"use_text_processer": true,
"tokenizer_config": {
"hub_backend": "hf",
"autotokenizer_name": "AutoTokenizer",
"from_pretrained": "./weights/google/mt5-xxl",
"model_max_length": 512
}
},
"dataloader_param": {
"dataloader_mode": "sampler",
"sampler_type": "LengthGroupedSampler",
"num_workers": 4,
"shuffle": true,
"drop_last": true,
"pin_memory": true,
"group_frame": false,
"group_resolution": false,
"group_data": true,
"initial_global_step_for_sampler": 0
}
}