Skip to main content

continuation-jax : Continuaion Framework for lambda

Continuation methods of Deep Neural Networks Tags: optimization, deep-learning, homotopy, bifurcation-analysis, continuation

Code style: black PyPI version License: MIT build

Install using pip:

Package: https://pypi.org/project/continuation-jax/

pip install continuation-jax

Import

import cjax

Math operations on Pytrees

>>> import cjax
>>> from cjax.utils import math_trees
>>> math_trees.pytree_element_mul([2,3,5], 2)
[4, 6, 10]
>>> math_trees.pytree_sub([2,3,5], [1,1,1])
[DeviceArray(1, dtype=int32), DeviceArray(2, dtype=int32), DeviceArray(4, dtype=int32)]
>>> math_trees.pytree_zeros_like({'a':12, 'b':45, 'c':[1,1]})
{'a': 0, 'b': 0, 'c': [0, 0]}

Examples:

"""
Main file to run contination on the user defined problem. Examples can be found in the examples/ directory.


Continuation is topological procedure to train a neural network. This module tracks all
the critical points or fixed points and dumps them to  output file provided in hparams.json file.

  Typical usage example:

  continuation = ContinuationCreator(
            problem=problem, hparams=hparams
        ).get_continuation_method()
        continuation.run()


"""
from cjax.continuation.creator.continuation_creator import ContinuationCreator
from examples.model_simple_classifier.model_classifier import ModelContClassifier
from cjax.utils.abstract_problem import ProblemWraper
import json
from jax.config import config
from datetime import datetime
import mlflow
from cjax.utils.visualizer import pick_array, bif_plot

config.update("jax_debug_nans", True)

# TODO: use **kwargs to reduce params

if __name__ == "__main__":
    problem = ModelContClassifier()
    problem = ProblemWraper(problem)

    with open(problem.HPARAMS_PATH, "r") as hfile:
        hparams = json.load(hfile)
    mlflow.set_tracking_uri(hparams['meta']["mlflow_uri"])
    mlflow.set_experiment(hparams['meta']["name"])

    with mlflow.start_run(run_name=hparams['meta']["method"]+"-"+hparams["meta"]["optimizer"]) as run:
        mlflow.log_dict(hparams, artifact_file="hparams/hparams.json")
        mlflow.log_text("", artifact_file="output/_touch.txt")
        artifact_uri = mlflow.get_artifact_uri("output/")
        hparams["meta"]["output_dir"] = artifact_uri
        print(f"URI: {artifact_uri}")
        start_time = datetime.now()

        if hparams["n_perturbs"] > 1:
            for perturb in range(hparams["n_perturbs"]):
                print(f"Running perturb {perturb}")
                continuation = ContinuationCreator(
                    problem=problem, hparams=hparams, key=perturb
                ).get_continuation_method()
                continuation.run()
        else:
            continuation = ContinuationCreator(
                problem=problem, hparams=hparams
            ).get_continuation_method()
            continuation.run()

        end_time = datetime.now()
        print(f"Duration: {end_time-start_time}")

        figure = bif_plot(hparams["meta"]["output_dir"], pick_array)
        mlflow.log_figure(figure, artifact_file="plots/fig.png")

Note on Hyperparameters

Papers:

Contact:

harshnpathak@gmail.com

Metadata

Release files for continuation-jax 0.0.7

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for continuation-jax 0.0.7
File Size Uploaded Provenance Actions
continuation_jax-0.0.7.tar.gz 57.4 kB Without provenance Details for continuation_jax-0.0.7.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for continuation-jax 0.0.7
File Interpreter ABI Platform Actions
continuation_jax-0.0.7-py3-none-any.whl Python 3 none any Details for continuation_jax-0.0.7-py3-none-any.whl

Total release size: 163.7 kB

Release files / continuation_jax-0.0.7.tar.gz

Download URL continuation_jax-0.0.7.tar.gz
Size 57.4 kB
Tags Source
SHA-256 checksum
How to use checksums
f29543d8090e4b91e23c7f19d03a21c8408042b582c88c98f25e159c318f963c
BLAKE2b-256 checksum
How to use checksums
29f6cd677235ea59bd7428d2a8c93fb5dc3101460d3b0a25f90474a3f671354c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/3.4.1 importlib_metadata/3.10.0 pkginfo/1.7.0 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.61.0 CPython/3.9.2

Release files / continuation_jax-0.0.7-py3-none-any.whl

Download URL continuation_jax-0.0.7-py3-none-any.whl
Size 106.4 kB
Tags Python 3
SHA-256 checksum
How to use checksums
01e44ab1dd2e4fa294b87633e8d8883df30ad34a493887df1569524ec992e228
BLAKE2b-256 checksum
How to use checksums
aa33c6501ec1b7d94f36c969fcf5251067cbf76cc5946a5de0e041c1cf688999
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/3.4.1 importlib_metadata/3.10.0 pkginfo/1.7.0 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.61.0 CPython/3.9.2

Security for continuation-jax 0.0.7

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page