{
"dataset_param": {
"dataset_type": "vace",
"basic_parameters": {
"data_path": "./examples/vace/feature_extract/data.txt",
"data_storage_mode": "vace"
},
"preprocess_parameters": {
"video_processor_type": "VACEVideoProcessor",
"video_reader_type": "DecordVideo",
"image_reader_type": "Image",
"image_processer_type": "image2image",
"num_frames": 17,
"frame_interval": 1,
"auto_interval": false,
"max_height": 480,
"max_width": 832,
"max_hxw": 399360,
"train_fps": 24,
"speed_factor": 1.0,
"cfg": 0.0,
"keep_last": false,
"zero_start": true,
"vae_stride": [4, 8, 8],
"vae_patch_size": [1, 2, 2],
"force_resolution": false,
"train_pipeline": {
"video": [
{
"trans_type": "ResizeCropVideo",
"param": {
"antialias": true,
"size": "auto",
"interpolation_mode": "bicubic"
}
},
{
"trans_type": "ToTensorVideo"
},
{
"trans_type": "norm_fun",
"param": {
"mean": 0.5,
"std": 0.5
}
}
],
"image": [
{
"trans_type": "ToTensorVideo"
},
{
"trans_type": "norm_fun",
"param": {
"mean": 0.5,
"std": 0.5
}
},
{
"trans_type": "ResizeToFill",
"param": {
"size": "auto",
"interpolation_mode": "bilinear",
"align_corners": false
}
}
]
}
},
"use_text_processer": true,
"enable_text_preprocessing": true,
"text_preprocess_methods":[
{
"method": "basic_clean"
},
{
"method": "whitespace_clean"
}
],
"tokenizer_config":
{
"autotokenizer_name": "AutoTokenizer",
"hub_backend": "hf",
"from_pretrained": "./Wan2.1-VACE-1.3B/google/umt5-xxl",
"model_max_length": 512
},
"support_chinese": true
},
"dataloader_param": {
"dataloader_mode": "sampler",
"sampler_type": "stateful_distributed_sampler",
"shuffle": false,
"drop_last": true,
"pin_memory": true,
"group_frame": false,
"group_resolution": false,
"group_data": true,
"initial_global_step_for_sampler": 0
}
}