v0.5.0

What's Changed

• Add simple documentation by @STomoya in #51
• bundle imports for training. by @STomoya in #53
• Optimizer step by @STomoya in #54
• Examples by @STomoya in #52
• GradScaler getter by @STomoya in #55
• Engine by @STomoya in #56
• Add optimizers by @STomoya in #57
• 🐛 Bug Fix by @STomoya in #58
• PyTorch 2.0 by @STomoya in #59
• Update metrics.test_classification by @STomoya in #60
• Parameter sweeper by @STomoya in #61
• 🐛 clip_grad_norm_ by @STomoya in #62
• Support infinite dataset by @STomoya in #64
• Local seed context manager by @STomoya in #65
• 🐛 torch.compile by @STomoya in #66
• 🐛 compile not disabled when compile=False by @STomoya in #67
• Set parallel mode only once. by @STomoya in #68
• Checkpoint random states by @STomoya in #69
• Batch size logger by @STomoya in #70
• Project initializer by @STomoya in #72
• Project by @STomoya in #71
• 🐛 correctly disable compile when used with FSDP by @STomoya in #73
• gather/reduce without distributed initialized. by @STomoya in #74
• Status class update by @STomoya in #75
• Optimizer step by @STomoya in #76
• Bugfix by @STomoya in #78
• Pass drop_last to DataLoader by @STomoya in #79
• ✨ polynomial decay lr scheduler by @STomoya in #82
• 🔨 update FSDP api by @STomoya in #84
• ✨ LayerNorm2d by @STomoya in #87
• ✨ functional API for distributed. by @STomoya in #88
• Omegaconf by @STomoya in #90
• ✨ only on primary decorator by @STomoya in #95
• ✨ hook for GPU memory summary. by @STomoya in #96
• Cleanup by @STomoya in #99
• ✨ NeST by @STomoya in #91
• 🔨 use transforms.v2 namespace by @STomoya in #101

Full Changelog: v0.4.0...v0.5.0