Skip to content

Commit

Permalink
Custom dataset (clementchadebec#49)
Browse files Browse the repository at this point in the history 
* add custom datasets

* make sanity check on loader

* add DatasetOutput class

* unify data loading with other models

* allow custom datasets as inputs

* update tests

* update tests

* add tutorial

* update gitignore

* update README

* clean up inerpolate and reconstruct

* add interpolate and reconstruct

* update README

* apply black

* small update in test

* remove not needed imports

* add install on main

* clean up notebook

* black

* remove DoubleBatchDataset

* remove DoubleBatchDataset

* black isort
  • Loading branch information
@clementchadebec
clementchadebec authored and Louis Vincent committed Jan 24, 2023
1 parent c562fef commit 2c547e1
Show file tree
Hide file tree
Showing 22 changed files with 712 additions and 243 deletions.
4 changes: 4 additions & 0 deletions .gitignore
View file
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,10 @@ examples/scripts/plots/*
examples/notebooks/my_model_with_custom_archi/
examples/notebooks/my_model/
examples/net.py
examples/notebooks/mlruns/*
mlruns/*
examples/notebooks/data_folders/*

examples/scripts/configs/*
examples/scripts/reproducibility/reproducibility/*

Expand Down
6 changes: 4 additions & 2 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -156,8 +156,8 @@ To launch a model training, you only need to call a `TrainingPipeline` instance.
... )
>>> # Launch the Pipeline
>>> pipeline(
... train_data=your_train_data, # must be torch.Tensor or np.array
... eval_data=your_eval_data # must be torch.Tensor or np.array
... train_data=your_train_data, # must be torch.Tensor, np.array or torch datasets
... eval_data=your_eval_data # must be torch.Tensor, np.array or torch datasets
... )
```

Expand Down Expand Up @@ -408,6 +408,8 @@ provide tutorials:

- [making_your_own_autoencoder.ipynb](https://github.com/clementchadebec/benchmark_VAE/tree/main/examples/notebooks) shows you how to pass your own networks to the models implemented in pythae [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/making_your_own_autoencoder.ipynb)

- [custom_dataset.ipynb](https://github.com/clementchadebec/benchmark_VAE/tree/main/examples/notebooks) shows you how to use custom datasets with any of the models implemented in pythae [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/custom_dataset.ipynb)

- [hf_hub_models_sharing.ipynb](https://github.com/clementchadebec/benchmark_VAE/tree/main/examples/notebooks) shows you how to upload and download models for the HuggingFace Hub [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/hf_hub_models_sharing.ipynb)

- [wandb_experiment_monitoring.ipynb](https://github.com/clementchadebec/benchmark_VAE/tree/main/examples/notebooks) shows you how to monitor you experiments using `wandb` [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/clementchadebec/benchmark_VAE/blob/main/examples/notebooks/wandb_experiment_monitoring.ipynb)
Expand Down
338 changes: 338 additions & 0 deletions examples/notebooks/custom_dataset.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,338 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Tutorial - Using your own Dataset\n",
"\n",
"In this notebook, we will see how to use your own `Dataset` within `pythae`"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# Install the library\n",
"%pip install git+https://github.com/clementchadebec/benchmark_VAE.git"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Load the data"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import torchvision.datasets as datasets\n",
"\n",
"%load_ext autoreload\n",
"%autoreload 2"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"mnist_trainset = datasets.MNIST(root='../data', train=True, download=True, transform=None)\n",
"\n",
"train_dataset = mnist_trainset.data[:10000].reshape(-1, 1, 28, 28) / 255.\n",
"train_targets = mnist_trainset.targets[:10000]\n",
"eval_dataset = mnist_trainset.data[-10000:].reshape(-1, 1, 28, 28) / 255.\n",
"eval_targets = mnist_trainset.targets[-10000:]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Example\n",
"Below is presented an example where we build a dataset inheriting from [`torchvision.datasets.ImageFolder`](https://pytorch.org/vision/stable/generated/torchvision.datasets.ImageFolder.html#torchvision.datasets.ImageFolder) that iterate on images located in a folder. In particular, this loader will avoid loading all the data at once."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# We will save the data in folders to mimic the desired example\n",
"\n",
"import os\n",
"import torch\n",
"import numpy as np\n",
"import imageio\n",
"\n",
"if not os.path.exists(\"data_folders\"):\n",
" os.mkdir(\"data_folders\")\n",
"if not os.path.exists(\"data_folders/train\"):\n",
" os.mkdir(\"data_folders/train\")\n",
"if not os.path.exists(\"data_folders/eval\"):\n",
" os.mkdir(\"data_folders/eval\")\n",
"\n",
"for i in range(len(train_dataset)):\n",
" img = 255.0*train_dataset[i][0].unsqueeze(-1)\n",
" img_folder = os.path.join(\"data_folders\", \"train\", f\"{train_targets[i]}\")\n",
" if not os.path.exists(img_folder):\n",
" os.mkdir(img_folder)\n",
" imageio.imwrite(os.path.join(img_folder, \"%08d.jpg\" % i), np.repeat(img, repeats=3, axis=-1).type(torch.uint8))\n",
"\n",
"for i in range(len(eval_dataset)):\n",
" img = 255.0*eval_dataset[i][0].unsqueeze(-1)\n",
" img_folder = os.path.join(\"data_folders\", \"eval\", f\"{eval_targets[i]}\")\n",
" if not os.path.exists(img_folder):\n",
" os.mkdir(img_folder)\n",
" imageio.imwrite(os.path.join(img_folder, \"%08d.jpg\" % i), np.repeat(img, repeats=3, axis=-1).type(torch.uint8))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# Define transforms to be applied on the data when reloaded\n",
"from torchvision import datasets, transforms\n",
"data_transform = transforms.Compose([\n",
" transforms.Grayscale(num_output_channels=1),\n",
" transforms.ToTensor() # the data must be tensors\n",
"])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Define your `CustomDataset`\n",
"\n",
"In this example, we build a custom dataset inheriting from ImageFolder. The only thing you have to keep in mind when building a custom dataset that you want to use in `pythae` is that the `__getitem__` method must output a `DatasetOutput` instance containing at least `data` as key. If this is not the case, you will not be able to combine your Dataset with the `pipelines`. "
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from pythae.data.datasets import DatasetOutput\n",
"\n",
"class MyCustomDataset(datasets.ImageFolder):\n",
"\n",
" def __init__(self, root, transform=None, target_transform=None):\n",
" super().__init__(root=root, transform=transform, target_transform=target_transform)\n",
"\n",
" def __getitem__(self, index):\n",
" X, _ = super().__getitem__(index)\n",
"\n",
" return DatasetOutput(\n",
" data=X\n",
" )"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"train_dataset = MyCustomDataset(\n",
" root=\"data_folders/train\",\n",
" transform=data_transform,\n",
")\n",
"\n",
"eval_dataset = MyCustomDataset(\n",
" root=\"data_folders/eval\", \n",
" transform=data_transform\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Use your CustomDataset to train a `pythae.models`\n",
"\n",
"Now, the datasets can be passed to the `training_pipeline` to train any model implemented in `pythae`."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from pythae.models import VAE, VAEConfig\n",
"from pythae.trainers import BaseTrainerConfig\n",
"from pythae.pipelines.training import TrainingPipeline"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"config = BaseTrainerConfig(\n",
" output_dir='my_model',\n",
" learning_rate=1e-3,\n",
" batch_size=100,\n",
" num_epochs=10, # Change this to train the model a bit more\n",
")\n",
"\n",
"\n",
"model_config = VAEConfig(\n",
" input_dim=(1, 28, 28),\n",
" latent_dim=16\n",
")\n",
"\n",
"model = VAE(\n",
" model_config=model_config\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"pipeline = TrainingPipeline(\n",
" training_config=config,\n",
" model=model\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"pipeline(\n",
" train_data=train_dataset, # here we use the custom train dataset\n",
" eval_data=eval_dataset # here we use the custom eval dataset\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Let's have a look to the trained model"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"from pythae.models import AutoModel"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"last_training = sorted(os.listdir('my_model'))[-1]\n",
"trained_model = AutoModel.load_from_folder(os.path.join('my_model', last_training, 'final_model'))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from pythae.samplers import NormalSampler"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# create normal sampler\n",
"normal_samper = NormalSampler(\n",
" model=trained_model\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# sample\n",
"gen_data = normal_samper.sample(\n",
" num_samples=25\n",
")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import matplotlib.pyplot as plt"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# show results with normal sampler\n",
"fig, axes = plt.subplots(nrows=5, ncols=5, figsize=(10, 10))\n",
"\n",
"for i in range(5):\n",
" for j in range(5):\n",
" axes[i][j].imshow(gen_data[i*5 +j].cpu().squeeze(0), cmap='gray')\n",
" axes[i][j].axis('off')\n",
"plt.tight_layout(pad=0.)"
]
}
],
"metadata": {
"interpreter": {
"hash": "5e51c5ac46389dd7ba2bd8215d251ab84152720d3cad2ff91113d77594821aef"
},
"kernelspec": {
"display_name": "Python 3.8.12 ('pythae')",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.13"
},
"orig_nbformat": 4
},
"nbformat": 4,
"nbformat_minor": 2
}
4 changes: 4 additions & 0 deletions src/pythae/customexception.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -36,3 +36,7 @@ class BadInheritanceError(Exception):

class ModelError(Exception):
pass


class DatasetError(Exception):
pass
Loading

0 comments on commit 2c547e1

Please sign in to comment.