{
"dataset_param": {
"dataset_type": "t2v",
"basic_parameters": {
"data_path": "./examples/wan2.1/feature_extract/data.txt",
"data_storage_mode": "combine"
},
"preprocess_parameters": {
"video_processor_type": "OpensoraplanVideoProcessor",
"video_reader_type": "DecordVideo",
"image_reader_type": "Image",
"num_frames": 81,
"min_num_frames": 1,
"frame_interval": 1,
"auto_interval": false,
"max_height": 480,
"max_width": 832,
"max_hxw": 399360,
"train_fps": 24,
"speed_factor": 1.0,
"drop_short_ratio": 0.0,
"cfg": 0.0,
"hw_stride": 8,
"force_resolution": false,
"hw_aspect_thr": 5.0,
"train_pipeline": {
"video_and_first_frame": [
{
"trans_type": "ResizeVideo",
"param": {
"transform_size": "auto",
"interpolation_mode": "bilinear",
"antialias": true,
"mode": "shortside"
}
},
{
"trans_type": "ToTensorVideo"
}
],
"video_only": [
{
"trans_type": "CenterCropResizeVideo",
"param": {
"use_short_edge": false,
"top_crop": false,
"align_corners": false,
"antialias": true,
"transform_size": "auto"
}
},
{
"trans_type": "norm_fun",
"param": {
"mean": 0.5,
"std": 0.5
}
}
],
"first_frame_clip": [
{
"trans_type": "ResizeVideo",
"param": {
"transform_size": {
"max_height": 224,
"max_width": 224
},
"mode": "resize",
"interpolation_mode": "bicubic",
"align_corners": false,
"antialias": false
}
},
{
"trans_type": "norm_fun",
"param": {
"mean": [0.48145466, 0.4578275, 0.40821073],
"std": [0.26862954, 0.26130258, 0.27577711]
}
}
],
"first_frame_vae": [
{
"trans_type": "ResizeVideo",
"param": {
"transform_size": "auto",
"mode": "resize",
"interpolation_mode": "bilinear",
"align_corners": false,
"antialias": false
}
},
{
"trans_type": "norm_fun",
"param": {
"mean": 0.5,
"std": 0.5
}
}
]
}
},
"use_text_processer": true,
"enable_text_preprocessing": true,
"text_preprocess_methods":{
"method": "whitespace_clean"
},
"tokenizer_config":
{
"autotokenizer_name": "AutoTokenizer",
"hub_backend": "hf",
"from_pretrained": "Wan2.1-I2V-14B-Diffusers/tokenizer",
"model_max_length": 512
}
},
"dataloader_param": {
"dataloader_mode": "sampler",
"sampler_type": "LengthGroupedSampler",
"shuffle": true,
"drop_last": true,
"pin_memory": true,
"group_frame": false,
"group_resolution": false,
"group_data": true,
"initial_global_step_for_sampler": 0
}
}