Flowvision V0.1.0 Stable Release

Flowvision V0.1.0 Stable Release

New Features

• Support trunc_normal_ in flowvision.layers.weight_init #92
• Support DeiT model #115
• Support PolyLRScheduler and TanhLRScheduler in flowvision.scheduler #85
• Add resmlp_12_224_dino model and pretrained weight #128
• Support ConvNeXt model #93
• Add ReXNet weights #132

Bug Fixes

• Fix F.normalize usage in SSD #116
• Fix bug in EfficientNet and Res2Net #122
• Fix error pretrained weight usage in vit_small_patch32_384 and res2net50_48w_2s #128

Improvements

• Refator trunc_normal_ and linspace usage in Swin-T, Cross-Former, PVT and CSWin models #100
• Refator Vision Transformer model #115
• Refine flowvision.models.ModelCreator to support ModelCreator.model_list func #123
• Refator README #124
• Refine load_state_dict_from_url in flowvision.models.utils to support downloading pretrained weights to cache dir ~/.oneflow/flowvision_cache #127
• Rebuild a cleaner model zoo and test all the model with pretrained weights released in flowvision #128

Docs Update

• Update Vision Transformer docs #115
• Add Getting Started docs #124
• Add resmlp_12_224_dino docs #128
• Fix VGG docs bug #128
• Add ConvNeXt docs #93

Contributors

A total of 5 developers contributed to this release. Thanks @rentainhe, @simonJJJ, @kaijieshi7, @lixiang007666, @Ldpe2G