{
    "info": {
        "author": "Yukio Fukuzawa",
        "author_email": "y.fukuzawa@massey.ac.nz",
        "bugtrack_url": null,
        "classifiers": [
            "License :: OSI Approved :: MIT License",
            "Operating System :: OS Independent",
            "Programming Language :: Python"
        ],
        "description": "Real valued sequence to sequence autoencoder\n---\nMost sequence to sequence autoencoders I can find are suitable for categorical sequences,\n such as translation.\n \nThis auto encoder is for real valued sequences.\n\nThe input and output can be multi dimensional, have different dimensions and could even be totally different\n\n\n### Compatibility\n- This package is compatible with both Python 2 and 3\n- Example scripts in `main.py` and `main.ipynb` are also compatible with both\n- I have no plan to use tensorflow 2.0, so all the nonsense about future obsolescence is muted\n\n\n### Installation\n\n```bash\npip install seq2seq\n```\n\n\n### Usage\nFirst create a factory:\n```python\nfrom models import  NDS2SAEFactory\n\nn_iterations = 3000\nfactory = NDS2SAEFactory()\nfactory.set_output('toy.zip')\nfactory.input_dim = 2 #  Input is 2 dimensiona;\nfactory.output_dim = 1 #  Output is one dimensional\nfactory.layer_sizes = [50, 30]\n\n# Use exponential decaying learning rate, in here it starts at 0.02 then decreases exponentially to 0.00001 \nfactory.lrtype = 'expdecay'\nfactory.lrargs = dict(start_lr=0.02, finish_lr=0.00001, decay_steps=n_iterations)\n\n# Alternatively, set a constant rate is also possible, e.g. 0.001\n# factory.lrtype = 'constant'\n# factory.lrargs = dict(lr=0.001)\n\n# For the dropout layer. Default is None. If None, the dropout layer is not used\nfactory.keep_prob = 0.7\n\n# The hidden layer will be symmetric (in this case: 50:30:30:50)\n# otherwise it'll be repeated (50:30:50:30)\nfactory.symmetric = True\n\n# Save or load (and resume) from this zip file\nencoder = factory.build()\n```\n\nCreate a training sample generator and a validation sample generator. Both should have the same signature:\n```python\ndef generate_samples(batch_size):\n    \"\"\"\n    :return in_seq: a list of input sequences. Each sequence must be a np.ndarray\n            out_seq: a list of output sequences. Each sequence must be a np.ndarray\n                These sequences don't need to be the same length and don't need any padding\n                The encoder will take care of that\n            last_batch: True if this batch is the last of the iteration.\n                E.g. if you have 70000 samples and the batch size is 20000, you'll have 4 batches\n                     the last batch contains 10000 samples. You should return False for the first 3\n                     batches and True for the last one\n    \"\"\"\n    ...\n    return in_seq, out_seq, last_batch\n```\n\n#### Train\n\n```python\nencoder.train(train_generator, valid_generator, n_iterations=3000, batch_size=100, display_step=100)\n```\n\n#### Predict\n\n```python\n# test_seq is a list of np.ndarrays\npredicted = encoder.predict(test_seq)\n\n# predicted is a list of np.ndarrays. Each sequence will have the same length (due to padding)\n# Look for the stop token to truncate the padding out\n```\n\n#### Encode\n\n```python\n# test_seq is a list of np.ndarrays\nencoded = encoder.encode(test_seq)\n\n# encoded is a list of hidden-layer states corresponding to each input sequence\n```\n\n### Jupyter notebook\nOpen [main.ipynb](main.ipynb) to run the example\n\n### Licence\nMIT\n\nPRs are welcome",
        "description_content_type": "text/markdown",
        "docs_url": null,
        "download_url": "",
        "downloads": {
            "last_day": -1,
            "last_month": -1,
            "last_week": -1
        },
        "home_page": "https://github.com/fzyukio/multidimensional-variable-length-seq2seq-autoencoder",
        "keywords": "",
        "license": "MIT",
        "maintainer": "",
        "maintainer_email": "",
        "name": "seq2seq",
        "package_url": "https://pypi.org/project/seq2seq/",
        "platform": "",
        "project_url": "https://pypi.org/project/seq2seq/",
        "project_urls": {
            "Homepage": "https://github.com/fzyukio/multidimensional-variable-length-seq2seq-autoencoder"
        },
        "release_url": "https://pypi.org/project/seq2seq/0.0.5/",
        "requires_dist": null,
        "requires_python": "",
        "summary": "Real valued sequence to sequence autoencoder",
        "version": "0.0.5"
    },
    "last_serial": 5436000,
    "releases": {
        "0.0.1": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "7d0792f49849fc6e2fea2f79aa387824",
                    "sha256": "3283b04c9d659a19517a3b15e49e7b7040edb32cdc466486634dcbcc20d2904b"
                },
                "downloads": -1,
                "filename": "seq2seq-0.0.1.tar.gz",
                "has_sig": false,
                "md5_digest": "7d0792f49849fc6e2fea2f79aa387824",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 2109,
                "upload_time": "2019-06-11T08:36:52",
                "url": "https://files.pythonhosted.org/packages/34/be/2fb3f89cc0504f0cddb77f7bd7aa39355a733e8717365290421bbcd4d22b/seq2seq-0.0.1.tar.gz"
            }
        ],
        "0.0.2": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "cbc2b9dfdb09ba1c02a6ae71069c39da",
                    "sha256": "f70674e093183ce3718c988cfcb1603e865c7fd9925dda9724db0fcc4dd86416"
                },
                "downloads": -1,
                "filename": "seq2seq-0.0.2.tar.gz",
                "has_sig": false,
                "md5_digest": "cbc2b9dfdb09ba1c02a6ae71069c39da",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 2115,
                "upload_time": "2019-06-12T08:47:18",
                "url": "https://files.pythonhosted.org/packages/bf/40/c3a38b1ac832f66113fb0046df1e6ac40a4e4058fe4d4ab20d17cbed4f45/seq2seq-0.0.2.tar.gz"
            }
        ],
        "0.0.3": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "ffc1b1e86e24496bcaebe5fb76f90c04",
                    "sha256": "816d9d3a0af05d70608fba51d73a522ee4ddbc0f3d0b7b33e70548fff98c1bd9"
                },
                "downloads": -1,
                "filename": "seq2seq-0.0.3.tar.gz",
                "has_sig": false,
                "md5_digest": "ffc1b1e86e24496bcaebe5fb76f90c04",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 2111,
                "upload_time": "2019-06-19T11:09:16",
                "url": "https://files.pythonhosted.org/packages/34/f6/178779839e83de09b2cb9d201e588617483e020a517bb932d6a14038a81a/seq2seq-0.0.3.tar.gz"
            }
        ],
        "0.0.4": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "365569760c3b5d27d73b664e01eaf236",
                    "sha256": "4546ded6d4f37204f7ecfac1983515571a0c845d004c2090a99e221158e668a1"
                },
                "downloads": -1,
                "filename": "seq2seq-0.0.4.tar.gz",
                "has_sig": false,
                "md5_digest": "365569760c3b5d27d73b664e01eaf236",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 2110,
                "upload_time": "2019-06-19T22:05:36",
                "url": "https://files.pythonhosted.org/packages/4a/93/e10bf4e0a44cbab2b91fb96cef498f15493d9354e6e3b22ae97f8f8a076a/seq2seq-0.0.4.tar.gz"
            }
        ],
        "0.0.5": [
            {
                "comment_text": "",
                "digests": {
                    "md5": "c0bba6ac44925776b16b68b75f8e2a97",
                    "sha256": "17709120dd208de7c275d737d294707e12e84cef99f62e57595be12fb5b33ddc"
                },
                "downloads": -1,
                "filename": "seq2seq-0.0.5.tar.gz",
                "has_sig": false,
                "md5_digest": "c0bba6ac44925776b16b68b75f8e2a97",
                "packagetype": "sdist",
                "python_version": "source",
                "requires_python": null,
                "size": 2790,
                "upload_time": "2019-06-23T00:53:26",
                "url": "https://files.pythonhosted.org/packages/7a/e3/42b85fdaf1885de998c212afd9dda06aee4f5bded209551f5fa1b1945551/seq2seq-0.0.5.tar.gz"
            }
        ]
    },
    "urls": [
        {
            "comment_text": "",
            "digests": {
                "md5": "c0bba6ac44925776b16b68b75f8e2a97",
                "sha256": "17709120dd208de7c275d737d294707e12e84cef99f62e57595be12fb5b33ddc"
            },
            "downloads": -1,
            "filename": "seq2seq-0.0.5.tar.gz",
            "has_sig": false,
            "md5_digest": "c0bba6ac44925776b16b68b75f8e2a97",
            "packagetype": "sdist",
            "python_version": "source",
            "requires_python": null,
            "size": 2790,
            "upload_time": "2019-06-23T00:53:26",
            "url": "https://files.pythonhosted.org/packages/7a/e3/42b85fdaf1885de998c212afd9dda06aee4f5bded209551f5fa1b1945551/seq2seq-0.0.5.tar.gz"
        }
    ]
}