diff options
| -rw-r--r-- | .gitignore | 1 | ||||
| -rw-r--r-- | python-rctorchprivate.spec | 300 | ||||
| -rw-r--r-- | sources | 1 |
3 files changed, 302 insertions, 0 deletions
@@ -0,0 +1 @@ +/rctorchprivate-0.9819998.tar.gz diff --git a/python-rctorchprivate.spec b/python-rctorchprivate.spec new file mode 100644 index 0000000..b618ab7 --- /dev/null +++ b/python-rctorchprivate.spec @@ -0,0 +1,300 @@ +%global _empty_manifest_terminate_build 0 +Name: python-rctorchprivate +Version: 0.9819998 +Release: 1 +Summary: A Python 3 toolset for creating and optimizing Echo State Networks. This library is an extension and expansion of the previous library written by Reinier Maat: https://github.com/1Reinier/Reservoir +License: Harvard +URL: https://github.com/blindedjoy/RcTorch-private +Source0: https://mirrors.nju.edu.cn/pypi/web/packages/bf/70/4da4e6d44e451dd964b55b1b4b5e7d2dab919609d4168920ab5dc4090970/rctorchprivate-0.9819998.tar.gz +BuildArch: noarch + + +%description +A Pytorch toolset for creating and optimizing Echo State Networks. +>License: 2020-2021 MIT +>Authors: Hayden Joy, Marios Mattheakis +Contains: +- A ESN Reservoir architecture class "rc.py" +- Bayesian Optimization (BO) class "rc_bayes.py" with optimized routines for Echo State Nets through `Botorch` (GPU optimized), can train multiple RCs in parellel durring BO + - an implimentation of the TURBO-1 algorithm as outlined in this paper: https://github.com/uber-research/TuRBO +- Capable of solving differential equations (the population equation, the bernoulli equation, a simple harmonic oscillator and a nonlinear oscillator) +Reference to prior instantiation: +This library is an extension and expansion of a previous library written by Reinier Maat: https://github.com/1Reinier/Reservoir +2018 International Joint Conference on Neural Networks (IJCNN), pp. 1-7. IEEE, 2018 +https://arxiv.org/abs/1903.05071 +## For example usage please see the notebooks folder. +# Installation +## Using pip +Like most standard libraries, `rctorch` is hosted on [PyPI](https://pypi.org/project/RcTorch/). To install the latest stable relesase, +```bash +pip install -U rctorch # '-U' means update to latest version +``` +## Example Usages +### Imports +```python +from rctorch import * +import torch +``` +### Load data +RcTorch has several built in datasets. Among these is the forced pendulum dataset. Here we demonstrate +```python +fp_data = rctorch.data.load("forced_pendulum", train_proportion = 0.2) +force_train, force_test = fp_data["force"] +target_train, input_test = fp_data["target"] +#Alternatively you can use sklearn's train_test_split. +``` +### Hyper-parameters +```python +#declare the hyper-parameters +>>> hps = {'connectivity': 0.4, + 'spectral_radius': 1.13, + 'n_nodes': 202, + 'regularization': 1.69, + 'leaking_rate': 0.0098085, + 'bias': 0.49} +``` +### Setting up your very own EchoStateNetwork +```python +my_rc = RcNetwork(**hps, random_state = 210, feedback = True) +#fitting the data: +my_rc.fit(y = target_train) +#making our prediction +score, prediction = my_rc.test(y = target_test) +my_rc.combined_plot() +``` + +Feedback allows the network to feed in the prediction at the previous timestep as an input. This helps the RC to make longer and more stable predictions in many situations. +### Bayesian Optimization +Unlike most other reservoir neural network packages ours offers the automatically tune hyper-parameters. +```python +#any hyper parameter can have 'log_' in front of it's name. RcTorch will interpret this properly. +bounds_dict = {"log_connectivity" : (-2.5, -0.1), + "spectral_radius" : (0.1, 3), + "n_nodes" : (300,302), + "log_regularization" : (-3, 1), + "leaking_rate" : (0, 0.2), + "bias": (-1,1), + } +rc_specs = {"feedback" : True, + "reservoir_weight_dist" : "uniform", + "output_activation" : "tanh", + "random_seed" : 209} +rc_bo = RcBayesOpt(bounds = bounds_dict, + scoring_method = "nmse", + n_jobs = 1, + cv_samples = 3, + initial_samples= 25, + **rc_specs + ) +``` + +%package -n python3-rctorchprivate +Summary: A Python 3 toolset for creating and optimizing Echo State Networks. This library is an extension and expansion of the previous library written by Reinier Maat: https://github.com/1Reinier/Reservoir +Provides: python-rctorchprivate +BuildRequires: python3-devel +BuildRequires: python3-setuptools +BuildRequires: python3-pip +%description -n python3-rctorchprivate +A Pytorch toolset for creating and optimizing Echo State Networks. +>License: 2020-2021 MIT +>Authors: Hayden Joy, Marios Mattheakis +Contains: +- A ESN Reservoir architecture class "rc.py" +- Bayesian Optimization (BO) class "rc_bayes.py" with optimized routines for Echo State Nets through `Botorch` (GPU optimized), can train multiple RCs in parellel durring BO + - an implimentation of the TURBO-1 algorithm as outlined in this paper: https://github.com/uber-research/TuRBO +- Capable of solving differential equations (the population equation, the bernoulli equation, a simple harmonic oscillator and a nonlinear oscillator) +Reference to prior instantiation: +This library is an extension and expansion of a previous library written by Reinier Maat: https://github.com/1Reinier/Reservoir +2018 International Joint Conference on Neural Networks (IJCNN), pp. 1-7. IEEE, 2018 +https://arxiv.org/abs/1903.05071 +## For example usage please see the notebooks folder. +# Installation +## Using pip +Like most standard libraries, `rctorch` is hosted on [PyPI](https://pypi.org/project/RcTorch/). To install the latest stable relesase, +```bash +pip install -U rctorch # '-U' means update to latest version +``` +## Example Usages +### Imports +```python +from rctorch import * +import torch +``` +### Load data +RcTorch has several built in datasets. Among these is the forced pendulum dataset. Here we demonstrate +```python +fp_data = rctorch.data.load("forced_pendulum", train_proportion = 0.2) +force_train, force_test = fp_data["force"] +target_train, input_test = fp_data["target"] +#Alternatively you can use sklearn's train_test_split. +``` +### Hyper-parameters +```python +#declare the hyper-parameters +>>> hps = {'connectivity': 0.4, + 'spectral_radius': 1.13, + 'n_nodes': 202, + 'regularization': 1.69, + 'leaking_rate': 0.0098085, + 'bias': 0.49} +``` +### Setting up your very own EchoStateNetwork +```python +my_rc = RcNetwork(**hps, random_state = 210, feedback = True) +#fitting the data: +my_rc.fit(y = target_train) +#making our prediction +score, prediction = my_rc.test(y = target_test) +my_rc.combined_plot() +``` + +Feedback allows the network to feed in the prediction at the previous timestep as an input. This helps the RC to make longer and more stable predictions in many situations. +### Bayesian Optimization +Unlike most other reservoir neural network packages ours offers the automatically tune hyper-parameters. +```python +#any hyper parameter can have 'log_' in front of it's name. RcTorch will interpret this properly. +bounds_dict = {"log_connectivity" : (-2.5, -0.1), + "spectral_radius" : (0.1, 3), + "n_nodes" : (300,302), + "log_regularization" : (-3, 1), + "leaking_rate" : (0, 0.2), + "bias": (-1,1), + } +rc_specs = {"feedback" : True, + "reservoir_weight_dist" : "uniform", + "output_activation" : "tanh", + "random_seed" : 209} +rc_bo = RcBayesOpt(bounds = bounds_dict, + scoring_method = "nmse", + n_jobs = 1, + cv_samples = 3, + initial_samples= 25, + **rc_specs + ) +``` + +%package help +Summary: Development documents and examples for rctorchprivate +Provides: python3-rctorchprivate-doc +%description help +A Pytorch toolset for creating and optimizing Echo State Networks. +>License: 2020-2021 MIT +>Authors: Hayden Joy, Marios Mattheakis +Contains: +- A ESN Reservoir architecture class "rc.py" +- Bayesian Optimization (BO) class "rc_bayes.py" with optimized routines for Echo State Nets through `Botorch` (GPU optimized), can train multiple RCs in parellel durring BO + - an implimentation of the TURBO-1 algorithm as outlined in this paper: https://github.com/uber-research/TuRBO +- Capable of solving differential equations (the population equation, the bernoulli equation, a simple harmonic oscillator and a nonlinear oscillator) +Reference to prior instantiation: +This library is an extension and expansion of a previous library written by Reinier Maat: https://github.com/1Reinier/Reservoir +2018 International Joint Conference on Neural Networks (IJCNN), pp. 1-7. IEEE, 2018 +https://arxiv.org/abs/1903.05071 +## For example usage please see the notebooks folder. +# Installation +## Using pip +Like most standard libraries, `rctorch` is hosted on [PyPI](https://pypi.org/project/RcTorch/). To install the latest stable relesase, +```bash +pip install -U rctorch # '-U' means update to latest version +``` +## Example Usages +### Imports +```python +from rctorch import * +import torch +``` +### Load data +RcTorch has several built in datasets. Among these is the forced pendulum dataset. Here we demonstrate +```python +fp_data = rctorch.data.load("forced_pendulum", train_proportion = 0.2) +force_train, force_test = fp_data["force"] +target_train, input_test = fp_data["target"] +#Alternatively you can use sklearn's train_test_split. +``` +### Hyper-parameters +```python +#declare the hyper-parameters +>>> hps = {'connectivity': 0.4, + 'spectral_radius': 1.13, + 'n_nodes': 202, + 'regularization': 1.69, + 'leaking_rate': 0.0098085, + 'bias': 0.49} +``` +### Setting up your very own EchoStateNetwork +```python +my_rc = RcNetwork(**hps, random_state = 210, feedback = True) +#fitting the data: +my_rc.fit(y = target_train) +#making our prediction +score, prediction = my_rc.test(y = target_test) +my_rc.combined_plot() +``` + +Feedback allows the network to feed in the prediction at the previous timestep as an input. This helps the RC to make longer and more stable predictions in many situations. +### Bayesian Optimization +Unlike most other reservoir neural network packages ours offers the automatically tune hyper-parameters. +```python +#any hyper parameter can have 'log_' in front of it's name. RcTorch will interpret this properly. +bounds_dict = {"log_connectivity" : (-2.5, -0.1), + "spectral_radius" : (0.1, 3), + "n_nodes" : (300,302), + "log_regularization" : (-3, 1), + "leaking_rate" : (0, 0.2), + "bias": (-1,1), + } +rc_specs = {"feedback" : True, + "reservoir_weight_dist" : "uniform", + "output_activation" : "tanh", + "random_seed" : 209} +rc_bo = RcBayesOpt(bounds = bounds_dict, + scoring_method = "nmse", + n_jobs = 1, + cv_samples = 3, + initial_samples= 25, + **rc_specs + ) +``` + +%prep +%autosetup -n rctorchprivate-0.9819998 + +%build +%py3_build + +%install +%py3_install +install -d -m755 %{buildroot}/%{_pkgdocdir} +if [ -d doc ]; then cp -arf doc %{buildroot}/%{_pkgdocdir}; fi +if [ -d docs ]; then cp -arf docs %{buildroot}/%{_pkgdocdir}; fi +if [ -d example ]; then cp -arf example %{buildroot}/%{_pkgdocdir}; fi +if [ -d examples ]; then cp -arf examples %{buildroot}/%{_pkgdocdir}; fi +pushd %{buildroot} +if [ -d usr/lib ]; then + find usr/lib -type f -printf "/%h/%f\n" >> filelist.lst +fi +if [ -d usr/lib64 ]; then + find usr/lib64 -type f -printf "/%h/%f\n" >> filelist.lst +fi +if [ -d usr/bin ]; then + find usr/bin -type f -printf "/%h/%f\n" >> filelist.lst +fi +if [ -d usr/sbin ]; then + find usr/sbin -type f -printf "/%h/%f\n" >> filelist.lst +fi +touch doclist.lst +if [ -d usr/share/man ]; then + find usr/share/man -type f -printf "/%h/%f.gz\n" >> doclist.lst +fi +popd +mv %{buildroot}/filelist.lst . +mv %{buildroot}/doclist.lst . + +%files -n python3-rctorchprivate -f filelist.lst +%dir %{python3_sitelib}/* + +%files help -f doclist.lst +%{_docdir}/* + +%changelog +* Mon May 15 2023 Python_Bot <Python_Bot@openeuler.org> - 0.9819998-1 +- Package Spec generated @@ -0,0 +1 @@ +afb0949432aed7a30eaad46eafa09568 rctorchprivate-0.9819998.tar.gz |