{
"dataset_param": {
"dataset_type": "t2v",
"basic_parameters": {
"data_path": "./examples/opensoraplan1.2/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,
"vae_scale_factor": [4,8,8],
"max_height": 480,
"max_width": 640,
"train_fps": 24,
"speed_factor": 1.0,
"drop_short_ratio": 1.0,
"cfg": 0.1,
"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": "your model path",
"model_max_length": 512
}
},
"dataloader_param": {
"dataloader_mode": "sampler",
"sampler_type": "LengthGroupedSampler",
"shuffle": true,
"drop_last": true,
"pin_memory": false,
"group_frame": false,
"group_resolution": false,
"collate_param": {
"model_name": "open_sora_plan"
}
}
}