From 1f80cb4b68a596056f2c4755c77f9b12cef22515 Mon Sep 17 00:00:00 2001 From: Victor Hall Date: Sat, 25 Feb 2023 16:19:23 -0500 Subject: [PATCH] optimizer doc --- README.md | 6 +++++- doc/OPTIMIZER.md | 39 +++++++++++++++++++++++++++++++++++++++ 2 files changed, 44 insertions(+), 1 deletion(-) create mode 100644 doc/OPTIMIZER.md diff --git a/README.md b/README.md index b35b5b8..2d8586f 100644 --- a/README.md +++ b/README.md @@ -35,6 +35,8 @@ Make sure to check out the [tools repo](https://github.com/victorchall/EveryDrea [Advanced Tweaking](doc/ATWEAKING.md) - More stuff to tweak once you are comfortable +[Advanced Optimizer Tweaking](/doc/OPTIMIZER.md) - Even more stuff to tweak if you are *very adventurous* + [Chaining training sessions](doc/CHAINING.md) - Modify training parameters by chaining training sessions together end to end [Shuffling Tags](doc/SHUFFLING_TAGS.md) @@ -49,4 +51,6 @@ Make sure to check out the [tools repo](https://github.com/victorchall/EveryDrea [Free tier Google Colab notebook](https://colab.research.google.com/github/victorchall/EveryDream2trainer/blob/main/Train_Colab.ipynb) -[RunPod / Vast](/doc/CLOUD_SETUP.md) \ No newline at end of file +[RunPod / Vast](/doc/CLOUD_SETUP.md) + +[Docker image link](https://github.com/victorchall/EveryDream2trainer/pkgs/container/everydream2trainer) \ No newline at end of file diff --git a/doc/OPTIMIZER.md b/doc/OPTIMIZER.md new file mode 100644 index 0000000..c07f695 --- /dev/null +++ b/doc/OPTIMIZER.md @@ -0,0 +1,39 @@ +# Advanced optimizer tweaking + +You can set advanced optimizer settings using this arg: + + --optimizer_config optimizer.json + +or in train.json + + "optimizer_config": "optimizer.json" + +A default `optimizer.json` is supplied which you can modify + +This has expanded tweaking. This doc is incomplete, but there is information on the web on betas and weight decay setting you can search for. + +If you do not set optimizer_config, the defaults are `adamw8bit` with standard betas of `(0.9,0.999)`, weight decay `0.01`, and epsilon `1e-8`. The hyperparameters are originally from XavierXiao's Dreambooth code and based off Compvis Stable Diffusion code. + +## Optimizers + +In `optimizer.json` the `optimizer` value is the type of optimizer to use. Below are the supported optimizers. + +* adamw + +Standard full precision AdamW optimizer exposed by PyTorch. Not recommended. Slower and uses more memory than adamw8bit. Widely documented on the web. + +* adamw8bit + +Tim Dettmers / bitsandbytes AdamW 8bit optimizer. This is the default and recommended setting. Widely documented on the web. + +* lion + +Lucidrains' [implementation](https://github.com/lucidrains/lion-pytorch) of the [lion optimizer](https://arxiv.org/abs/2302.06675). Click links to read more. Unknown what hyperparameters will work well, but paper shows potentially quicker learning. *Highly experimental, but tested and works.* + +## Optimizer parameters + +LR can be set in `optimizer.json` and excluded from the main CLI arg or train.json but if you use the main CLI arg or set it in the main train.json it will override the setting. This was done to make sure existing behavior will not break. To set LR in the `optimizer.json` make sure to delete `"lr": 1.3e-6` in your main train.json and exclude the CLI arg. + +Betas, weight decay, and epsilon are documented in the [AdamW paper](https://arxiv.org/abs/1711.05101) and there is a wealth of information on the web, but consider those experimental to tweak. I cannot provide advice on what might be useful to tweak here. + +Note `lion` does not use epsilon. \ No newline at end of file