lucidrains/ema-pytorch issues and pull requests

#37 - should the value `update_after_step` be larger then the value of warmup steps?

Issue - State: open - Opened by JohnHerry about 2 months ago

#36 - expected dtype float for `end` but got dtype c10::BFloat16

Issue - State: closed - Opened by ZhiyuanChen about 2 months ago - 1 comment

#35 - Memory leak in PostHocEMA during checkpointing

Issue - State: closed - Opened by xandergos 2 months ago - 4 comments

#34 - Avoid deep copies when saving post-hoc EMA checkpoints

Pull Request - State: closed - Opened by kalekundert 3 months ago - 1 comment

#33 - Post-hoc EMA with non-deepcopyable model

Issue - State: closed - Opened by kalekundert 3 months ago - 7 comments

#32 - Add `demote_types` function to minimize memory usage

Pull Request - State: closed - Opened by amorehead 4 months ago - 1 comment

#31 - Ensure `current_decay` is on the correct device in `update_moving_average`

Pull Request - State: closed - Opened by HiDolen 5 months ago - 1 comment

#30 - Rename `allow_different_devices` to `auto_move_device`

Pull Request - State: closed - Opened by HiDolen 5 months ago - 9 comments

#29 - How BatchNorm is handled?

Issue - State: open - Opened by liqimai 6 months ago

#28 - `PostHocEMA.copy_params_from_ema_to_model()` is not correctly named

Issue - State: closed - Opened by liqimai 6 months ago - 1 comment

#27 - fix FutureWarning with `torch.load`

Pull Request - State: closed - Opened by HiDolen 6 months ago - 2 comments

#26 - FutureWarning with `torch.load` in `PostHocEMA` Example

Issue - State: closed - Opened by HiDolen 6 months ago - 3 comments

#25 - torch.no_grad on net?

Issue - State: closed - Opened by TKassis 8 months ago - 2 comments

#24 - Alternative way to set requires_grad to False

Pull Request - State: closed - Opened by mjwen 9 months ago - 1 comment

#23 - Tensor -> tensor

Pull Request - State: closed - Opened by una-dinosauria 9 months ago - 1 comment

#22 - fix bug with gamma in synthesize_ema_model

Pull Request - State: closed - Opened by zhvng 11 months ago - 2 comments

#21 - option to specify post hoc ema dtype

Pull Request - State: closed - Opened by zhvng 11 months ago - 1 comment

#20 - Leaf/computed variables requires_grad error

Issue - State: open - Opened by crlandsc 11 months ago

#19 - fix: switch to is_floating_point and is_complex

Pull Request - State: closed - Opened by samedii 11 months ago - 3 comments

#18 - fix: add bfloat16 to allowed dtypes

Pull Request - State: closed - Opened by samedii 11 months ago - 4 comments

#17 - post hoc ema from https://arxiv.org/pdf/2312.02696.pdf

Pull Request - State: closed - Opened by lucidrains 12 months ago

#16 - Function parameters confusing in inplace_copy and inplace_lerp

Issue - State: closed - Opened by sheshanshag almost 1 year ago - 1 comment

#15 - add step-dependent karras beta

Pull Request - State: closed - Opened by elyxlz about 1 year ago - 1 comment

#14 - Post-hoc ema?

Issue - State: closed - Opened by elyxlz about 1 year ago - 11 comments

#13 - Update EMA docstring

Pull Request - State: closed - Opened by undefx about 1 year ago

#12 - print detailed error message

Pull Request - State: closed - Opened by developer0hye about 1 year ago

#11 - question about usage

Issue - State: closed - Opened by fire717 over 1 year ago - 4 comments

#10 - track FP16

Pull Request - State: closed - Opened by zhaohui-yang almost 2 years ago - 1 comment

#9 - Using EMA in a DDP setup

Issue - State: closed - Opened by JamesMcCullochDickens almost 2 years ago - 2 comments

#8 - Implementation error

Issue - State: closed - Opened by sbl1996 almost 2 years ago - 1 comment

#7 - Saving EMA also saves online model?

Issue - State: closed - Opened by Netruk44 about 2 years ago - 4 comments

#6 - Support for non-deepcopyable models

Issue - State: closed - Opened by stefan-baumann over 2 years ago - 5 comments

Ecosyste.ms: Issues

GitHub / lucidrains/ema-pytorch issues and pull requests