Bump version of Python, torch and pytorch-lightning

README.md

@@ -1,26 +1,24 @@
-# PyCave
+# TorchGMM
 
-![PyPi](https://img.shields.io/pypi/v/pycave?label=version)
-![License](https://img.shields.io/pypi/l/pycave)
+<!-- ![PyPi](https://img.shields.io/pypi/v/torchgmm?label=version)
+![License](https://img.shields.io/pypi/l/torchgmm) -->
 
-PyCave allows you to run traditional machine learning models on CPU, GPU, and even on multiple
-nodes. All models are implemented in [PyTorch](https://pytorch.org/) and provide an `Estimator` API
+TorchGMM allows to run Gaussian Mixture Models on single or multiple CPUs/GPUs.
+The repository is a fork from [PyCave](https://github.com/borchero/pycave) and [LightKit](https://github.com/borchero/lightkit), two amazing packages developed by [Olivier Borchert](https://github.com/borchero) that are not being maintained anymore.
+While PyCave implements additional models such as Markov Chains, TorchGMM focuses only on Gaussian Mixture Models.
+
+The models are implemented in [PyTorch](https://pytorch.org/) and [PyTorch Lightning](https://lightning.ai/docs/pytorch/stable/), and provide an `Estimator` API
 that is fully compatible with [scikit-learn](https://scikit-learn.org/stable/).
 
-For Gaussian mixture model, PyCave allows for 100x speed ups when using a GPU and enables to train
+For Gaussian mixture model, TorchGMM allows for 100x speed ups when using a GPU and enables to train
 on markedly larger datasets via mini-batch training. The full suite of benchmarks run to compare
-PyCave models against scikit-learn models is available on the
+TorchGMM models against scikit-learn models is available on the
 [documentation website](https://pycave.borchero.com/sites/benchmark.html).
 
-_PyCave version 3 is a complete rewrite of PyCave which is tested much more rigorously, depends on
-well-maintained libraries and is tuned for better performance. While you are, thus, highly
-encouraged to upgrade, refer to [pycave-v2.borchero.com](https://pycave-v2.borchero.com) for
-documentation on PyCave 2._
-
 ## Features
 
 - Support for GPU and multi-node training by implementing models in PyTorch and relying on
-  [PyTorch Lightning](https://www.pytorchlightning.ai/)
+  [PyTorch Lightning](https://lightning.ai/docs/pytorch/stable/)
 - Mini-batch training for all models such that they can be used on huge datasets
 - Well-structured implementation of models
 
@@ -31,21 +29,21 @@ documentation on PyCave 2._
 
 ## Installation
 
-PyCave is available via `pip`:
+TorchGMM is available via `pip`:
 
 ```bash
-pip install pycave
+pip install torchgmm
 ```
 
 If you are using [Poetry](https://python-poetry.org/):
 
 ```bash
-poetry add pycave
+poetry add torchgmm
 ```
 
 ## Usage
 
-If you've ever used scikit-learn, you'll feel right at home when using PyCave. First, let's create
+If you've ever used scikit-learn, you'll feel right at home when using TorchGMM. First, let's create
 some artificial data to work with:
 
 ```python
@@ -62,7 +60,7 @@ This dataset consists of three clusters with 8-dimensional datapoints. If you wa
 model, to find the clusters' centroids, it's as easy as:
 
 ```python
-from pycave.clustering import KMeans
+from torchgmm.clustering import KMeans
 
 estimator = KMeans(3)
 estimator.fit(X)
@@ -79,7 +77,7 @@ and which methods are available.
 
 ### GPU and Multi-Node training
 
-For GPU- and multi-node training, PyCave leverages PyTorch Lightning. The hardware that training
+For GPU- and multi-node training, TorchGMM leverages PyTorch Lightning. The hardware that training
 runs on is determined by the
 [Trainer](https://pytorch-lightning.readthedocs.io/en/latest/api/pytorch_lightning.trainer.trainer.html#pytorch_lightning.trainer.trainer.Trainer)
 class. It's
@@ -104,12 +102,11 @@ In fact, **you do not need to change anything else in your code**.
 
 ### Implemented Models
 
-Currently, PyCave implements three different models:
+Currently, TorchGMM implements two different models:
 
 - [GaussianMixture](https://pycave.borchero.com/sites/generated/bayes/gmm/pycave.bayes.GaussianMixture.html)
-- [MarkovChain](https://pycave.borchero.com/sites/generated/bayes/markov_chain/pycave.bayes.MarkovChain.html)
 - [K-Means](https://pycave.borchero.com/sites/generated/clustering/kmeans/pycave.clustering.KMeans.html)
 
 ## License
 
-PyCave is licensed under the [MIT License](https://github.com/borchero/pycave/blob/main/LICENSE).
+TorchGMM is licensed under the [MIT License](https://github.com/marcovarrone/torchgmm/blob/main/LICENSE).

pyproject.toml

@@ -4,21 +4,20 @@ classifiers = [
   "Development Status :: 5 - Production/Stable",
   "Topic :: Scientific/Engineering :: Artificial Intelligence",
 ]
-description = "Traditional Machine Learning Models in PyTorch."
-documentation = "https://pycave.borchero.com"
+description = "Gaussian Mixture Models in PyTorch."
 license = "MIT"
-name = "pycave"
+name = "torchgmm"
 readme = "README.md"
-repository = "https://github.com/borchero/pycave"
+repository = "https://github.com/marcovarrone/torchgmm"
 version = "0.0.0"
 
 [tool.poetry.dependencies]
 lightkit = "^0.5.0"
 numpy = "^1.20.3"
 python = ">=3.8,<3.11"
-pytorch-lightning = "^1.6.0"
-torch = "^1.8.0"
-torchmetrics = ">=0.6,<0.12"
+pytorch-lightning = "<2.2"
+torch = "<2.2.0"
+torchmetrics = ">=0.6,<1.4.0"
 
 [tool.poetry.group.pre-commit.dependencies]
 black = "^22.12.0"
@@ -76,7 +75,7 @@ target-version = ["py38", "py39", "py310"]
 [tool.isort]
 force_alphabetical_sort_within_sections = true
 include_trailing_comma = true
-known_first_party = "pycave,tests"
+known_first_party = "torchgmm,tests"
 line_length = 99
 lines_between_sections = 0
 profile = "black"

tests/_data/gmm.py

@@ -1,8 +1,8 @@
 # pylint: disable=missing-function-docstring
 from typing import Tuple
 import torch
-from pycave.bayes.core import CovarianceType
-from pycave.bayes.gmm import GaussianMixtureModel, GaussianMixtureModelConfig
+from torchgmm.bayes.core import CovarianceType
+from torchgmm.bayes.gmm import GaussianMixtureModel, GaussianMixtureModelConfig
 
 
 def sample_gmm(

tests/bayes/core/benchmark_log_normal.py

@@ -5,7 +5,7 @@
 from sklearn.mixture._gaussian_mixture import _compute_precision_cholesky  # type: ignore
 from sklearn.mixture._gaussian_mixture import _estimate_log_gaussian_prob  # type: ignore
 from torch.distributions import MultivariateNormal
-from pycave.bayes.core import cholesky_precision, log_normal
+from torchgmm.bayes.core import cholesky_precision, log_normal
 
 
 def test_log_normal_spherical(benchmark: BenchmarkFixture):

tests/bayes/core/benchmark_precision_cholesky.py

@@ -3,7 +3,7 @@
 import torch
 from pytest_benchmark.fixture import BenchmarkFixture  # type: ignore
 from sklearn.mixture._gaussian_mixture import _compute_precision_cholesky  # type: ignore
-from pycave.bayes.core import cholesky_precision
+from torchgmm.bayes.core import cholesky_precision
 
 
 def test_cholesky_precision_spherical(benchmark: BenchmarkFixture):

tests/bayes/core/test_normal.py

@@ -4,8 +4,8 @@
 from sklearn.mixture._gaussian_mixture import _compute_log_det_cholesky  # type: ignore
 from sklearn.mixture._gaussian_mixture import _compute_precision_cholesky  # type: ignore
 from torch.distributions import MultivariateNormal
-from pycave.bayes.core import cholesky_precision, covariance, log_normal, sample_normal
-from pycave.bayes.core._jit import _cholesky_logdet  # type: ignore
+from torchgmm.bayes.core import cholesky_precision, covariance, log_normal, sample_normal
+from torchgmm.bayes.core._jit import _cholesky_logdet  # type: ignore
 from tests._data.normal import (
     sample_data,
     sample_diag_covars,

tests/bayes/gmm/benchmark_gmm_estimator.py

@@ -5,8 +5,8 @@
 import torch
 from pytest_benchmark.fixture import BenchmarkFixture  # type: ignore
 from sklearn.mixture import GaussianMixture as SklearnGaussianMixture  # type: ignore
-from pycave.bayes import GaussianMixture
-from pycave.bayes.core.types import CovarianceType
+from torchgmm.bayes import GaussianMixture
+from torchgmm.bayes.core.types import CovarianceType
 from tests._data.gmm import sample_gmm
 
 
@@ -64,7 +64,7 @@ def test_sklearn(
         (1000000, 64, 64, "diag", 100000),
     ],
 )
-def test_pycave(
+def test_torchgmm(
     benchmark: BenchmarkFixture,
     num_datapoints: int,
     num_features: int,
@@ -107,7 +107,7 @@ def test_pycave(
         (1000000, 64, 64, "tied", 100000),
     ],
 )
-def test_pycave_gpu(
+def test_torchgmm_gpu(
     benchmark: BenchmarkFixture,
     num_datapoints: int,
     num_features: int,

tests/bayes/gmm/test_gmm_estimator.py

@@ -4,8 +4,8 @@
 import pytest
 import torch
 from sklearn.mixture import GaussianMixture as SklearnGaussianMixture  # type: ignore
-from pycave.bayes import GaussianMixture
-from pycave.bayes.core import CovarianceType
+from torchgmm.bayes import GaussianMixture
+from torchgmm.bayes.core import CovarianceType
 from tests._data.gmm import sample_gmm
 
 

tests/bayes/gmm/test_gmm_metrics.py

@@ -3,8 +3,8 @@
 import numpy as np
 import sklearn.mixture._gaussian_mixture as skgmm  # type: ignore
 import torch
-from pycave.bayes.core import CovarianceType
-from pycave.bayes.gmm.metrics import CovarianceAggregator, MeanAggregator, PriorAggregator
+from torchgmm.bayes.core import CovarianceType
+from torchgmm.bayes.gmm.metrics import CovarianceAggregator, MeanAggregator, PriorAggregator
 
 
 def test_prior_aggregator():

tests/bayes/gmm/test_gmm_model.py

@@ -1,6 +1,6 @@
 # pylint: disable=missing-function-docstring
 from torch import jit
-from pycave.bayes.gmm import GaussianMixtureModel, GaussianMixtureModelConfig
+from torchgmm.bayes.gmm import GaussianMixtureModel, GaussianMixtureModelConfig
 
 
 def test_compile():

tests/bayes/markov_chain/test_markov_chain_estimator.py

@@ -3,7 +3,7 @@
 from typing import Tuple
 import pytest
 import torch
-from pycave.bayes import MarkovChain
+from torchgmm.bayes import MarkovChain
 
 
 def test_fit_automatic_config():

tests/bayes/markov_chain/test_markov_chain_model.py

@@ -3,7 +3,7 @@
 import torch
 from torch import jit
 from torch.nn.utils.rnn import pack_padded_sequence
-from pycave.bayes.markov_chain import MarkovChainModel, MarkovChainModelConfig
+from torchgmm.bayes.markov_chain import MarkovChainModel, MarkovChainModelConfig
 
 
 def test_compile():

tests/clustering/kmeans/benchmark_kmeans_estimator.py

@@ -5,8 +5,8 @@
 import torch
 from pytest_benchmark.fixture import BenchmarkFixture  # type: ignore
 from sklearn.cluster import KMeans as SklearnKMeans  # type: ignore
-from pycave.clustering import KMeans
-from pycave.clustering.kmeans.types import KMeansInitStrategy
+from torchgmm.clustering import KMeans
+from torchgmm.clustering.kmeans.types import KMeansInitStrategy
 from tests._data.gmm import sample_gmm
 
 
@@ -61,7 +61,7 @@ def test_sklearn(
         (1000000, 100000, 64, 64, "random"),
     ],
 )
-def test_pycave(
+def test_torchgmm(
     benchmark: BenchmarkFixture,
     num_datapoints: int,
     batch_size: Optional[int],
@@ -101,7 +101,7 @@ def test_pycave(
         (10000000, 1000000, 128, 128, "random"),
     ],
 )
-def test_pycave_gpu(
+def test_torchgmm_gpu(
     benchmark: BenchmarkFixture,
     num_datapoints: int,
     batch_size: Optional[int],

tests/clustering/kmeans/test_kmeans_estimator.py

@@ -4,7 +4,7 @@
 import pytest
 import torch
 from sklearn.cluster import KMeans as SklearnKMeans  # type: ignore
-from pycave.clustering import KMeans
+from torchgmm.clustering import KMeans
 from tests._data.gmm import sample_gmm
 
 

tests/clustering/kmeans/test_kmeans_model.py

@@ -1,7 +1,7 @@
 # pylint: disable=missing-function-docstring
 import torch
 from torch import jit
-from pycave.clustering.kmeans import KMeansModel, KMeansModelConfig
+from torchgmm.clustering.kmeans import KMeansModel, KMeansModelConfig
 
 
 def test_compile():

pycave/bayes/gmm/estimator.py → torchgmm/bayes/gmm/estimator.py

@@ -5,8 +5,8 @@
 from lightkit import ConfigurableBaseEstimator
 from lightkit.data import collate_tensor, DataLoader, dataset_from_tensors, TensorLike
 from lightkit.estimator import PredictorMixin
-from pycave.bayes.core import CovarianceType
-from pycave.clustering import KMeans
+from torchgmm.bayes.core import CovarianceType
+from torchgmm.clustering import KMeans
 from .lightning_module import (
     GaussianMixtureKmeansInitLightningModule,
     GaussianMixtureLightningModule,
@@ -30,7 +30,7 @@ class GaussianMixture(
     `Wikipedia <https://en.wikipedia.org/wiki/Mixture_model>`_.
 
     See also:
-        .. currentmodule:: pycave.bayes.gmm
+        .. currentmodule:: torchgmm.bayes.gmm
         .. autosummary::
             :nosignatures:
             :template: classes/pytorch_module.rst
@@ -80,7 +80,7 @@ def __init__(
             num_workers: The number of workers to use for loading the data. Only used if a PyTorch
                 dataset is passed to :meth:`fit` or related methods.
             trainer_params: Initialization parameters to use when initializing a PyTorch Lightning
-                trainer. By default, it disables various stdout logs unless PyCave is configured to
+                trainer. By default, it disables various stdout logs unless TorchGMM is configured to
                 do verbose logging. Checkpointing and logging are disabled regardless of the log
                 level. This estimator further sets the following overridable defaults:
 

pycave/bayes/gmm/lightning_module.py → torchgmm/bayes/gmm/lightning_module.py

@@ -3,8 +3,8 @@
 import torch
 from pytorch_lightning.callbacks import EarlyStopping
 from torchmetrics import MeanMetric
-from pycave.bayes.core import cholesky_precision
-from pycave.utils import NonparametricLightningModule
+from torchgmm.bayes.core import cholesky_precision
+from torchgmm.utils import NonparametricLightningModule
 from .metrics import CovarianceAggregator, MeanAggregator, PriorAggregator
 from .model import GaussianMixtureModel
 

pycave/bayes/gmm/metrics.py → torchgmm/bayes/gmm/metrics.py

@@ -1,7 +1,7 @@
 from typing import Any, Callable, Optional
 import torch
 from torchmetrics import Metric
-from pycave.bayes.core import covariance_shape, CovarianceType
+from torchgmm.bayes.core import covariance_shape, CovarianceType
 
 
 class PriorAggregator(Metric):

pycave/bayes/gmm/model.py → torchgmm/bayes/gmm/model.py

@@ -4,8 +4,8 @@
 import torch
 from lightkit.nn import Configurable
 from torch import jit, nn
-from pycave.bayes.core import covariance, covariance_shape, CovarianceType
-from pycave.bayes.core._jit import jit_log_normal, jit_sample_normal
+from torchgmm.bayes.core import covariance, covariance_shape, CovarianceType
+from torchgmm.bayes.core._jit import jit_log_normal, jit_sample_normal
 
 
 @dataclass

pycave/bayes/gmm/types.py → torchgmm/bayes/gmm/types.py

@@ -7,7 +7,7 @@
 
 - **random**: Samples responsibilities of datapoints at random and subsequently initializes means
   and covariances from these.
-- **kmeans**: Runs K-Means via :class:`pycave.clustering.KMeans` and uses the centroids as the
+- **kmeans**: Runs K-Means via :class:`torchgmm.clustering.KMeans` and uses the centroids as the
   initial component means. For computing the covariances, responsibilities are given as the
   one-hot cluster assignments.
 - **kmeans++**: Runs only the K-Means++ initialization procedure to sample means in a smart

pycave/bayes/markov_chain/estimator.py → torchgmm/bayes/markov_chain/estimator.py

@@ -21,7 +21,7 @@ class MarkovChain(ConfigurableBaseEstimator[MarkovChainModel]):  # type: ignore
     available on `Wikipedia <https://en.wikipedia.org/wiki/Markov_chain>`_.
 
     See also:
-        .. currentmodule:: pycave.bayes.markov_chain
+        .. currentmodule:: torchgmm.bayes.markov_chain
         .. autosummary::
             :nosignatures:
             :template: classes/pytorch_module.rst
@@ -54,7 +54,7 @@ def __init__(
             num_workers: The number of workers to use for loading the data. Only used if a PyTorch
                 dataset is passed to :meth:`fit` or related methods.
             trainer_params: Initialization parameters to use when initializing a PyTorch Lightning
-                trainer. By default, it disables various stdout logs unless PyCave is configured to
+                trainer. By default, it disables various stdout logs unless TorchGMM is configured to
                 do verbose logging. Checkpointing and logging are disabled regardless of the log
                 level. This estimator further enforces the following parameters:
 

pycave/bayes/markov_chain/lightning_module.py → torchgmm/bayes/markov_chain/lightning_module.py

@@ -1,8 +1,8 @@
 import torch
 from torch.nn.utils.rnn import PackedSequence
 from torchmetrics import MeanMetric
-from pycave.bayes.markov_chain.metrics import StateCountAggregator
-from pycave.utils import NonparametricLightningModule
+from torchgmm.bayes.markov_chain.metrics import StateCountAggregator
+from torchgmm.utils import NonparametricLightningModule
 from .model import MarkovChainModel