Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/warner-benjamin/optimi

Fast, Modern, Memory Efficient, and Low Precision PyTorch Optimizers
https://github.com/warner-benjamin/optimi

deep-learning optimizers pytorch

Last synced: about 7 hours ago
JSON representation

Fast, Modern, Memory Efficient, and Low Precision PyTorch Optimizers

Awesome Lists containing this project

README 

        
# optimī



### Fast, Modern, Memory Efficient, and Low Precision PyTorch Optimizers



optimi enables accurate low precision training via Kahan summation, integrates gradient release and optimizer accumulation for additional memory efficiency, supports fully decoupled weight decay, and features fast implementations of modern optimizers.



## Low Precision Training with Kahan Summation



optimi optimizers can nearly reach or match the performance of mixed precision when [training in BFloat16 by using Kahan summation](https://optimi.benjaminwarner.dev/kahan_summation).



Training in BFloat16 with Kahan summation can reduce non-activation training memory usage by [37.5 to 45.5 percent](https://optimi.benjaminwarner.dev/kahan_summation/#memory-savings) when using an Adam optimizer. BFloat16 training increases single GPU [training speed by ~10 percent](https://optimi.benjaminwarner.dev/kahan_summation/#training-speedup) at the same batch size.



## Gradient Release: Fused Backward and Optimizer Step



optimi optimizers can perform the [optimization step layer-by-layer during the backward pass](https://optimi.benjaminwarner.dev/gradient_release), immediately freeing gradient memory.



Unlike the current PyTorch implementation, optimi’s gradient release optimizers are a drop-in replacement for standard optimizers and seamlessly work with exisiting hyperparmeter schedulers.



## Optimizer Accumulation: Gradient Release and Accumulation



optimi optimizers can approximate gradient accumulation with gradient release by [accumulating gradients into the optimizer states](https://optimi.benjaminwarner.dev/optimizer_accumulation).



## Fully Decoupled Weight Decay



In addition to supporting PyTorch-style decoupled weight decay, optimi optimizers also support [fully decoupled weight decay](https://optimi.benjaminwarner.dev/fully_decoupled_weight_decay).



Fully decoupled weight decay decouples weight decay from the learning rate, more accurately following [*Decoupled Weight Decay Regularization*](https://arxiv.org/abs/1711.05101). This can help simplify hyperparameter tuning as the optimal weight decay is no longer tied to the learning rate.



## Foreach Implementations



All optimi optimizers have fast [foreach implementations](https://optimi.benjaminwarner.dev/foreach), which can significantly outperform the for-loop versions. optimi reuses the gradient buffer for temporary variables to reduce foreach memory usage.



## Documentation



## Install



optimi is available to install from pypi.



```bash

pip install torch-optimi

```



## Usage



To use an optimi optimizer with Kahan summation and fully decoupled weight decay:



```python

import torch

from torch import nn

from optimi import AdamW



# create or cast model in low precision (bfloat16)

model = nn.Linear(20, 1, dtype=torch.bfloat16)



# initialize any optimi optimizer with parameters & fully decoupled weight decay

# Kahan summation is automatically enabled since model & inputs are bfloat16

opt = AdamW(model.parameters(), lr=1e-3, weight_decay=1e-5, decouple_lr=True)



# forward and backward, casting input to bfloat16 if needed

loss = model(torch.randn(20, dtype=torch.bfloat16))

loss.backward()



# optimizer step

opt.step()

opt.zero_grad()

```



To use with PyTorch-style weight decay with float32 or mixed precision:



```python

# create model

model = nn.Linear(20, 1)



# initialize any optimi optimizer with parameters

opt = AdamW(model.parameters(), lr=1e-3, weight_decay=1e-2)

```



To use with gradient release:



```python

# initialize any optimi optimizer with `gradient_release=True`

# and call `prepare_for_gradient_release` on model and optimizer

opt = AdamW(model.parameters(), lr=1e-3, gradient_release=True)

prepare_for_gradient_release(model, opt)



# setup a learning rate scheduler like normal

scheduler = CosineAnnealingLR(opt, ...)



# calling backward on the model will peform the optimzier step

loss = model(torch.randn(20, dtype=torch.bfloat16))

loss.backward()



# optimizer step and zero_grad are no longer needed, and will

# harmlessly no-op if called by an existing training framework

# opt.step()

# opt.zero_grad()



# step the learning rate scheduler like normal

scheduler.step()



# optionally remove gradient release hooks when done training

remove_gradient_release(model)

```



To use with optimizer accumulation:



```python

# initialize any optimi optimizer with `gradient_release=True`

# and call `prepare_for_gradient_release` on model and optimizer

opt = AdamW(model.parameters(), lr=1e-3, gradient_release=True)

prepare_for_gradient_release(model, opt)



# update model parameters every four steps after accumulating

# gradients directly into the optimizer states

accumulation_steps = 4



# setup a learning rate scheduler for gradient accumulation

scheduler = CosineAnnealingLR(opt, ...)



# use existing PyTorch dataloader

for idx, batch in enumerate(dataloader):

    # `optimizer_accumulation=True` accumulates gradients into

    # optimizer states. set `optimizer_accumulation=False` to

    # update parameters by performing a full gradient release step

    opt.optimizer_accumulation = (idx+1) % accumulation_steps != 0



    # calling backward on the model will peform the optimizer step

    # either accumulating gradients or updating model parameters

    loss = model(batch)

    loss.backward()



    # optimizer step and zero_grad are no longer needed, and will

    # harmlessly no-op if called by an existing training framework

    # opt.step()

    # opt.zero_grad()



    # step the learning rate scheduler after accumulating gradients

    if not opt.optimizer_accumulation:

        scheduler.step()



# optionally remove gradient release hooks when done training

remove_gradient_release(model)

```



## Differences from PyTorch



optimi optimizers do not support compilation, differentiation, complex numbers, or have capturable versions.



optimi Adam optimizers do not support AMSGrad and SGD does not support Nesterov momentum. Optimizers which debias updates (Adam optimizers and Adan) calculate the debias term per parameter group, not per parameter.



## Optimizers



optimi implements the following optimizers: [Adam](https://optimi.benjaminwarner.dev/optimizers/adam), [AdamW](https://optimi.benjaminwarner.dev/optimizers/adamw), [Adan](https://optimi.benjaminwarner.dev/optimizers/adan), [Lion](https://optimi.benjaminwarner.dev/optimizers/lion), [RAdam](https://optimi.benjaminwarner.dev/optimizers/radam), [Ranger](https://optimi.benjaminwarner.dev/optimizers/ranger), [SGD](https://optimi.benjaminwarner.dev/optimizers/sgd), & [StableAdamW](https://optimi.benjaminwarner.dev/optimizers/stableadamw)