accmt 1.2.5

Accelerator Module and Trainer based on Accelerate library for simple distributed train processes, inspired by PyTorch Lightning.

AcceleratorModule

Module based on Accelerate 🤗 for distributed training accross multiple GPUs, with focus on readability and ease to customize experiments. We also integrate modified versions of DataCollators from Transformers library for huggingface standard tokenizers to integrate with different environments.

NOTE: Some features might not be tested and could cause problems. Feel free to open an issue or send a PR to fix any problem found.

AcceleratorModule will take care of the heavy lifting of distributed training on many GPUs. Accelerate is quite simple, and it has many adventages over PyTorch Lightning, mainly because it doesn't abstract the low level part of the training loop, so you can customize it however you want. The main idea of this little project is to have a standard way to make distributed training. This module let's you:

• Define the logic involved for training.
• Define the logic involved for evaluation.
• Save checkpoints to recover training progress.
• Save best model by evaluating best average validation loss at the end of every epoch.
• Define the hyperparameters in a simple YAML file.
• Visualize training progress using TensorBoard (train and validation losses in one graph).
• And more...

Installation

AcceleratorModule is available via pip:

pip install accmt

Module Structure

Import AcceleratorModule:

from accmt import AcceleratorModule

The AcceleratorModule class has 3 main methods:

• forward: Defines the flow of data.
• training_step: Defines the training logic up to the loss function.
• validation_step: Defines the validation logic up to the loss function.

The structure looks like this:

class ExampleModule(AcceleratorModule):
    def __init__(self):
        self.model = ...

    def forward(self, x):
        return self.model(x)

    def training_step(self, batch, status_dict):
        x, y = batch
        # ...
        return train_loss

    def validation_step(self, batch, status_dict):
        x, y = batch
        # ...
        return val_loss

More information about module structure here.

To train this Module, you need a Trainer class:

from accmt import Trainer

trainer = Trainer(
    hps_config="hps_config.yaml",
    checkpoint="checkpoint_folder"
)

More information about trainer here.

HPS config file

This is a YAML file containing hyperparameters for your training. The structure looks like the following:

hps:
  epochs: 40
  batch_size: 35
  optim:
    type: AdamW
    lr: 1e-3
    weight_decay: 1e-3
  scheduler:
    type: OneCycleLR
    max_lr: 1e-3

An optimizer (optim) is necessary, while a scheduler is optional (do not specify if you don't want to).

Available optimizer types are the following:

Optimizer	Source
Adam	PyTorch
Adadelta	PyTorch
Adagrad	PyTorch
Adamax	PyTorch
AdamW	PyTorch
Adafactor	HuggingFace
ASGD	PyTorch
LBFGS	PyTorch
NAdam	PyTorch
RAdam	PyTorch
RMSprop	PyTorch
Rprop	PyTorch
SGD	PyTorch
SparseAdam	PyTorch

Available schedulers types are the following:

Scheduler	Source
StepLR	PyTorch
LinearLR	PyTorch
ExponentialLR	PyTorch
CosineAnnealingLR	PyTorch
CyclicLR	PyTorch
OneCycleLR	PyTorch
CosineAnnealingWarmRestarts	PyTorch
CosineWithWarmup	HuggingFace
Constant	HuggingFace
ConstantWithWarmup	HuggingFace
CosineWithHardRestartsWithWarmup	HuggingFace
InverseSQRT	HuggingFace
LinearWithWarmup	HuggingFace
PolynomialDecayWithWarmup	HuggingFace

Finally, we can train our model by using the .fit() function, providing our AcceleratorModule and the train and validation datasets (from PyTorch):

trainer.fit(module, train_dataset, val_dataset)

More information about HPS config file here.

Run

To run training, we use Accelerator 🤗 in the background, so you must use the corresponding CLI:

accelerate launch train.py

You can use any Accelerate configuration that you want 🤗 (DDP, FSDP or DeepSpeed).

Checkpointing

Checkpointing is a default process in ACCMT, and it's customizable with some parameters in the Trainer constructor:

trainer = Trainer(
    # ... Other parameters.
    checkpoint_every="2ep", # Checkpoint every N epochs, in this case, every 2 epochs.
    checkpoint="checkpoint_model", # Checkpoint directory.
    resume=True # Whether you want to resume from checkpoint (True), or start from scratch (False).
)

Save model

Model saving is an integrated feature of ACCMT. You can enable it by specifying a directory where to save the model.

You can also save model in 3 different modes:

• best_valid_loss: Saves the model whenever the validation loss is the best.
• best_train_loss: Saves the model whenever the train loss is the best.
• always: Save the model everytime it's possible.

And you can activate movel saving below a specific loss (e.g. if specified best_valid_loss, then model will be saved when validation loss is below the specified threshold).

trainer = Trainer(
    # ... Other parameters.
    model_path="model", # Path where to save model.
    model_saving="best_valid_loss", # Model saving mode.
    model_saving_below_loss=0.67 # Save model below this threshold (e.g. below 0.67 validation loss).
)

Gradient Accumulation

When training big models, size in memory becomes a huge problem. One way to avoid that is to not always step the optimizer, instead accumulate gradients for a certain amount of steps. This is very easy to do, just configure the parameter grad_accumulation_steps for the amount of steps you want to accumulate gradients before stepping.

Logging training progress

Logging training progress is set by default in ACCMT, as it is essential to track how good our experiments are, and determine if we're good to pause training.

We use Tensorboard by default. Other logger types will be supported in the future.

There are only 2 paremeters to change for this (in the Trainer constructor):

• logging_dir: Specifies a logging dir (default is "logs").
• log_every: Log every N number of steps (default is 1).

To visualize our training progress, we need to initialize TensorBoard in the terminal:

tensorboard --logdir=<logging_dir> [--port=6006] # or any other port

After that, you can open up a browser and type:

http://localhost:6006/

If you're running your training on a remote environment, you would need to set up a SSH tunel in your local machine:

ssh -L <local-port>:localhost:<remote-port> <username>@<remote-ip-address>

Collate Functions

You can implement your own collate function by overriding collate_fn from AcceleratorModule:

class ExampleModule(AcceleratorModule):
    # Rest of the code...

    def collate_fn(self, batch: list):
        # Your collate function logic here.

        return batch # Output taken in training and validation steps.

There is another and simplier way to add collators that I'm going to be building in the future, and that is using a specific DataCollator built into this library.

At the moment, there is only one available DataCollator: DataCollatorForSeq2Seq (inspired by the trasformers library). This is how you would include it in your training:

from accmt.collate_fns import DataCollatorForSeq2Seq

tokenizer = ... # a tokenizer from 'transformers' library.

trainer = Trainer(
    hps_file_config="hps_config.yaml",
    checkpoint="checkpoint_folder",
    collate_fn=DataCollatorForSeq2Seq(tokenizer)
)

This collator will handle padding to the longest example in a batch, as well as adding special tokens for padding in labels to compute loss without learning pad tokens. This is optimal to optimize training, since you're going to have tensors with the adecuated shape, and not a fixed one.

The current implementation is quite similar to transformers library.

Teacher-Student support

A Teacher-Student approach let's you mimic the behaviour of a bigger model (teacher) in a smaller model (student). This is a method for model distillation, useful to save computational resources and accelerate inference.

To load teacher and student models, we can do the following in the module constructor:

class TeacherStudentExampleModule(AcceleratorModule):
    def __init__(self):
        self.teacher = ... # teacher model
        self.model = ...   # student model

        self.teacher.eval() # set teacher to evaluation mode

During training, the teacher model will only provide outputs, and will not have its parameters updated.

NOTE: In order to successfully load models into hardware, we must use self.teacher for teacher model, and self.model for student model.

If using KL Divergence approach for the loss function, our step method will look something like this:

import torch
import torch.nn.functional as F
# other imports...

# other logic for module...

def step(self, batch):
    x = batch
    with torch.no_grad(): # no gradients required for teacher model
        teacher_logits = self.teacher(**x).logits

    student_output = self.model(**x)
    student_logits = student_output.logits

    soft_prob = F.log_softmax(student_logits / self.T, dim=-1)
    soft_targets = F.softmax(teacher_logits / self.T, dim=-1)

    kd_loss = F.kl_div(soft_prob, soft_targets, reduction="batchmean") * (self.T**2)
    loss = self.alpha * student_output.loss + (1. - self.alpha) * kd_loss

    return loss

Notes

I will continue to update this repository to add more features overtime. If you want to contribute to this little project, feel free to make a PR 🤗.