v1.2.5

What's Changed

• feature: update dist launch by @cnstark in #56

Full Changelog: v1.2.4...v1.2.5

v1.2.4

What's Changed

• feature: save cfg.txt and copy cfg file when launching training by @cnstark in #55

Full Changelog: v1.2.3...v1.2.4

v1.2.3

What's Changed

• feature: imagenet example by @cnstark in #43
• fix: examples training code by @cnstark in #53
• feature: re match for loading ckpt by @cnstark in #54

Full Changelog: 1.2.2...v1.2.3

v1.2.2

What's Changed

• feature: save bast model by @cnstark in #52

Full Changelog: v1.2.1...v1.2.2

v1.2.1

fix bug
Full Changelog: v1.2...v1.2.1

v1.2

What's Changed

• fix: add param 'n' for updating meter by @cnstark in #44
• feature: DDP find_unused_parameters by @cnstark in #49
• feature: support collate_fn in Dataloader by @zezhishao in #50

New Contributors

• @zezhishao made their first contribution in #50

Full Changelog: v1.1...v1.2

v1.1

What's Changed

• fix: fix seed on DDP by @cnstark in #34
• feature: detailed configs for random seed by @cnstark in #35
• feature: support custom optim and lr scheduler by @cnstark in #38
• feature: env cfg by @cnstark in #39
• fix: set spu mode when 'GPU_NUM' is 0 by @cnstark in #41
• version 1.1 by @cnstark in #42

Full Changelog: v1.0...v1.1

v1.0

Welcome to easytorch

What's Changed

• feature: add dist utils, unify single gpu and ddp by @cnstark in #2
• feature: merge nn_interface by @cnstark in #3
• refactor: refactor files by @cnstark in #4
• feature: support logging by @cnstark in #5
• feature: support finetune by @cnstark in #6
• feature: support val interval by @cnstark in #7
• feature: full ckpt save strategy by @cnstark in #8
• feature: add easyoptim, change runner by @YuhaoZeng in #11
• fix: multi_cos bug when 'warmup_begin' is zero by @cnstark in #12
• feature: choose ckpt on inference by @cnstark in #10
• feature: add func 'init_training' before training loop by @cnstark in #9
• feature: split common functions by @cnstark in #14
• feature: init validation by @cnstark in #15
• feature: friendly launcher by @cnstark in #16
• fix: fix val tensorboard by @cnstark in #17
• Feature/add notes by @cnstark in #18
• Feature/time predictor by @cnstark in #20
• feature: val callback by @cnstark in #21
• feature: support non cuda env environment by @cnstark in #22
• Create git-mirror.yml by @cnstark in #23
• fix: fix code by lgtm by @cnstark in #24
• feature: setup random by @cnstark in #25
• feature: save config file by @cnstark in #26
• feature: deploy by @cnstark in #19
• fix lr scheduler bug, support None Type default value now by @YuhaoZeng in #28
• feature: ddp multi node by @cnstark in #29
• feature: tqdm by @cnstark in #30
• feature: lamb optimizer by @cnstark in #31
• feature: release by @cnstark in #32

New Contributors

• @YuhaoZeng made their first contribution in #11

Full Changelog: https://github.com/cnstark/easytorch/commits/v1.0