Metadata-Version: 2.1
Name: tfcaidm
Version: 0.0.0a1
Summary: Deep learning pipeline for medical imaging
Home-page: https://github.com/Brandhsu/tfcaidm
Author: Brandhsu
Author-email: brandondhsu@gmail.com
License: GPLv3
Keywords: deep learning,neural networks,medical imaging,computer vision
Platform: UNKNOWN
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: GNU General Public License v3 (GPLv3)
Classifier: Programming Language :: Python :: 3.7
Requires-Python: >=3.6, <3.8
Description-Content-Type: text/markdown
License-File: LICENSE

# TFCAIDM

<div align="center">
    <img src="https://github.com/Brandhsu/tfcaidm-pkg/blob/main/docs/images/tensor.png" height="240" width="284" alt="TFCAIDM Tensor">
    <h4>TensorFlow CAIDM</h4>
    Deep learning pipeline for medical imaging
    <br><br>
    <img src="https://badgen.net/pypi/v/tfcaidm">
    <img src="https://badgen.net/pypi/python/tfcaidm">
    <img src="https://badgen.net/github/license/Brandhsu/tfcaidm">
    <img src="https://badgen.net/badge/code%20style/black?color=black">
</div>

---

## Introduction

<strong>TFCAIDM</strong> is a unified framework for building and training medical imaging deep learning models built on top of [TensorFlow](https://www.tensorflow.org/) and [JarvisMD](https://pypi.org/project/jarvis-md/). The library supports interfacing custom datasets with `jarvis`, model development with `tensorflow`, and built-in reproducibility, traceability, and performance logging for all experiments. User's can train or extend pre-existing models that have been implemented in `MODEL_ZOO.md` or define their own.

<details>

<summary>Available Features</summary>

- [`Reusable state-of-the-art deep learning model blocks`](https://github.com/Brandhsu/tfcaidm-pkg/blob/main/docs/tfcaidm/models/MODEL.md)
- `Support for training multiple models in parallel`
- `High-level interface for customizing datasets, models, loss functions, training routines, etc.`
- `Reproducibility, performance logging, model checkpointing, and hyperparameter tracking`
</details>

<details>

<summary>Upcoming Features</summary>

- `AutoML / efficient hyperparameter search`
- `Distributed data and model training`
- `Vision transformer models`
- `Better documentation`

</details>

<details>

<summary>More Information</summary>

- `YAML configuration files`
- `Hyperparameter tuning`
- `Supported models`
- `Customizability`
- `Viewing results`
- `Benchmarks (coming soon)`

</details>

<br>

Disclaimer: The library is primarily built for users with access to the caidm clusters, though general users are also supported.

---

## Installation

The current library is supported on python 3.7 and tensorflow 2.5+, and the installation instructions provided below assume that your system is already equipped with cuda and nvcc.

<details>
<summary>Local Installation</summary>

Install using the [conda](https://www.anaconda.com/products/individual) virtual environment.

Where `user` is your account username.

```sh
user $ conda create --name tfcaidm python=3.7
user $ conda activate tfcaidm
user (tfcaidm) $ pip install tensorflow
user (tfcaidm) $ pip install jarvis-md
user (tfcaidm) $ pip install tfcaidm
```

</details>

---

## Example

Training a set of models require two separate python scripts: a training submission script and a training routine script.

<details>
<summary>Training Submission</summary>

```python
from jarvis.utils.general import gpus
from tfcaidm import Jobs

# --- Define paths
YML_CONFIG = "pipeline.yml"
TRAIN_ROUTINE_PATH = "main.py"

# --- Submit a training job
Jobs(path=YML_CONFIG).setup(
    producer=__file__,
    consumer=TRAIN_ROUTINE_PATH,
).train_cluster()
```

</details>

<details>
<summary>Automated Training Routine</summary>

```python
from jarvis.train import params
from jarvis.utils.general import gpus
from tfcaidm import Trainer

# --- Autoselect GPU (use only on caidm cluster)
gpus.autoselect()

# --- Get hyperparameters (args passed by environment variables)
hyperparams = params.load()

# --- Train model (dataset and model created within trainer)
trainer = Trainer(hyperparams)
results = trainer.cross_validation(save=True)
trainer.save_results(results)
```

</details>

<details>
<summary>Custom Training Routine</summary>

```python
from jarvis.train import params
from jarvis.utils.general import gpus, overload
from tfcaidm import JClient
from tfcaidm import Model
from tfcaidm import Trainer

# --- Autoselect GPU (use only on caidm cluster)
gpus.autoselect()

# --- Get hyperparameters (args passed by environment variables)
hyperparams = params.load()

# --- Setup custom dataset generator (more details in notebooks)
@overload(JClient)
def create_generator(self, gen_data):
    for xs, ys in gen_data:

        # --- User defined code
        xs = DataAugment(xs)

        yield xs, ys

# --- Setup custom model (more details in notebooks)
@overload(Model)
def create(self):

    # --- User defined code
    model = ViT(...)
    model.compile(...)

    return model

# --- Train model (dataset and model created within trainer)
trainer = Trainer(hyperparams)
results = trainer.cross_validation(save=True)
trainer.save_results(results)

# See notebooks for a breakdown on customizability
```

</details>

<br>

For an example project, see `examples/projects`. For a more detailed walkthrough of the library, see `notebooks`.

---

## Sister Repositories

- [peterchang77/caidm](https://github.com/peterchang77/caidm)
- [peterchang77/dl_tutor](https://github.com/peterchang77/dl_tutor)


