Skip to content

Latest commit

 

History

History
51 lines (30 loc) · 2.38 KB

timm09_12.md

File metadata and controls

51 lines (30 loc) · 2.38 KB
Raw

数据

原文链接：huggingface.co/docs/timm/reference/data

timm.data.create_dataset

< source >

( name root split = 'validation' search_split = True class_map = None load_bytes = False is_training = False download = False batch_size = None seed = 42 repeats = 0 **kwargs )

数据集工厂方法

括号中每个参数后面都是支持的数据集类型之一：

  • folder - 默认，基于 timm 文件夹（或 tar）的 ImageDataset

  • torch - 基于 torchvision 的数据集

  • HFDS - Hugging Face 数据集

  • TFDS - 通过 IterableImageDataset 在 IterabeDataset 接口中包装的 Tensorflow 数据集

  • WDS - Webdataset

  • 所有 - 以上任何一个

timm.data.create_loader

< source >

( dataset input_size batch_size is_training = False use_prefetcher = True no_aug = False re_prob = 0.0 re_mode = 'const' re_count = 1 re_split = False scale = None ratio = None hflip = 0.5 vflip = 0.0 color_jitter = 0.4 auto_augment = None num_aug_repeats = 0 num_aug_splits = 0 interpolation = 'bilinear' mean = (0.485, 0.456, 0.406) std = (0.229, 0.224, 0.225) num_workers = 1 distributed = False crop_pct = None crop_mode = None collate_fn = None pin_memory = False fp16 = False img_dtype = torch.float32 device = device(type='cuda') tf_preprocessing = False use_multi_epochs_loader = False persistent_workers = True worker_seeding = 'all' )

timm.data.create_transform

< source >

( input_size is_training = False use_prefetcher = False no_aug = False scale = None ratio = None hflip = 0.5 vflip = 0.0 color_jitter = 0.4 auto_augment = None interpolation = 'bilinear' mean = (0.485, 0.456, 0.406) std = (0.229, 0.224, 0.225) re_prob = 0.0 re_mode = 'const' re_count = 1 re_num_splits = 0 crop_pct = None crop_mode = None tf_preprocessing = False separate = False )

timm.data.resolve_data_config

< source >

( args = None pretrained_cfg = None model = None use_test_size = False verbose = False )