{ "info": { "author": "Kumar Nityan Suman", "author_email": "nityan.suman@gmail.com", "bugtrack_url": null, "classifiers": [ "Intended Audience :: Developers", "Intended Audience :: Education", "Intended Audience :: Science/Research", "License :: OSI Approved :: MIT License", "Operating System :: OS Independent", "Programming Language :: Python :: 3", "Programming Language :: Python :: 3.1", "Programming Language :: Python :: 3.2", "Programming Language :: Python :: 3.3", "Programming Language :: Python :: 3.4", "Programming Language :: Python :: 3.5", "Programming Language :: Python :: 3.6", "Programming Language :: Python :: 3.7", "Topic :: Software Development :: Libraries", "Topic :: Software Development :: Libraries :: Python Modules" ], "description": "

\"LOGO\"/

\n\n\"PyPI\" \"PyPI \"GitHub \"PyPI [![Codacy Badge](https://api.codacy.com/project/badge/Grade/d1e35c252db741b28144f5b7b9ffd7d2)](https://www.codacy.com/app/nityansuman/tensorhub?utm_source=github.com&utm_medium=referral&utm_content=nityansuman/tensorhub&utm_campaign=Badge_Grade)\n\n\n## You have just found TensorHub.\n\nThe open source library to help you develop and train ML models, easy and fast as never before with `TensorFlow 2.0`.\n`TensorHub` is a global collection of `building blocks` and `ready to serve models`.\n\nIt is a wrapper library of deep learning models and neural lego blocks designed to make deep learning more accessible and accelerate ML research.\n\nUse `TensorHub` if you need a deep learning library that:\n\n+ **Reproducibility** - Reproduce the results of existing pre-training models (such as Google BERT, XLNet)\n\n+ **Model modularity** - TensorHub divided into multiple components: ready-to-serve models, layers, neural-blocks etc. Ample modules are implemented in each component. Clear and robust interface allows users to combine modules with as few restrictions as possible.\n\n+ **Prototyping** - Code less build more. Apply `TensorHub` to create fast prototypes with the help of modulear blocks, custom layers, custom activation support.\n\n+ **Platform Independent** - Supports both `Keras` and `TensorFlow 2.0`. Run your model on CPU, single GPU or using a distributed training strategy.\n\n------------------\n\n\n## Getting started: 30 seconds to TensorHub\n\n**Here is the `Sequential` model for `Image Classification`:**\n\n```python\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.layers import Conv2D, Dense\n\n# Use Tensorhub to accelerate your prototyping\nfrom tensorhub.layers import InceptionV1 # Custom Inception Layer\nfrom tensorhub.models.image.classifiers import CNNClassifier, VGG16 # Cooked Models\nfrom tensorhub.utilities.activations import gelu, softmax # Custom Activations Supported\n\n## Initiate a sequential model\nmodel = Sequential()\n\n## Stacking layers is as easy as `.add()`\nmodel.add(Conv2D(32, (3, 3), activation=gelu, input_shape=(100, 100, 3)))\nmodel.add(Conv2D(32, (3, 3), activation=gelu))\nmodel.add(MaxPooling2D(pool_size=(2, 2)))\nmodel.add(Dropout(0.25))\n\n## Add custom layer like any other standard layer\nmodel.add(InceptionV1(32)) \n\nmodel.add(Dense(units=64, activation=gelu, input_dim=100))\nmodel.add(Dense(units=10, activation=softmax))\n\n# Or\n## Use one of our pre-cooked models\nmodel = VGG16(n_classes=10, num_nodes=64, img_height=100, img_width=100)\n\n## Once your model looks good, configure its learning process with `.compile()`\nmodel.compile(\n loss='categorical_crossentropy',\n optimizer='sgd',\n metrics=['accuracy']\n)\n\n## Alternatively, if you need to, you can further configure your compile configuration\nmodel.compile(\n loss=tensorflow.keras.losses.categorical_crossentropy,\n optimizer=tensorflow.keras.optimizers.SGD(lr=0.01, momentum=0.9, nesterov=True),\n metrics=['acc']\n)\n\n## You can now iterate on your training data in batches\n\n## x_train and y_train are Numpy arrays\nmodel.fit(x_train, y_train, epochs=5, batch_size=32)\n\n## Alternatively, you can feed batches to your model manually\nmodel.train_on_batch(x_batch, y_batch)\n\n## Evaluate your performance in one line:\nloss_and_metrics = model.evaluate(x_test, y_test, batch_size=128)\n\n## Or generate predictions on new data\nclasses = model.predict(x_test, batch_size=128)\n```\n\nBuilding a question answering system, an image classification model, a Neural Turing Machine, or any other model is just as fast. The ideas behind deep learning are simple, so why should their implementation be painful?\n\nFor a more in-depth tutorial about Keras, you can check out:\n\n+ [Getting started with Text Classification](https://github.com/nityansuman/tensorhub/tree/master/examples/training-a-text-classifier-using-tensorhub-models.ipynb)\n+ [Getting started with Custom Layers](https://github.com/nityansuman/tensorhub/tree/master/examples/creating-custom-models.ipynb)\n\nIn the [examples folder](https://github.com/nityansuman/tensorhub/tree/master/examples) of this repository, you will find much more advanced examples coming your way very soon.\n\n------------------\n\n\n## What's coming in V1.0\n+ Cooked Models\n + Image Classification (Supports Transfer Learning with ImageNet Weights)\n + Xception\n + VGG16\n + VGG19\n + ResNet50\n + InceptionV3\n + InceptionResNetV2\n + MobileNet\n + DenseNet\n + NASNet\n + SqueezeNet (Without Transfer Learning) *\n\n + Text Classification\n + RNN Model\n + LSTM Model\n + GRU Model\n + Text-CNN\n\n + Neural Machine Translation *\n + Encoder-Decoder Sequence Translation Model\n + Translation with Attention\n\n + Text Generation *\n + RNN, LSTM, GRU Based Model\n\n + Named Entity Recogniton *\n + RNN, LSTM, GRU Based Model\n\n+ Custom Modules/Layers\n + Standard Layers\n + Linear\n + Inception V1 Layer\n + Inception V1 with Reduction Layer\n + Inception V2 Layer *\n + Inception V3 Layer *\n + Attention layers\n + Bahdanau Attention\n + Luong Attention\n + Self-Attention *\n\n+ Utilities\n + Text\n + Custom Word and Character Tokenizer\n + Load Pre-trained Embeddings\n + Create Vocabulary Matrix\n + Image *\n + Image Augmentation\n + Activations\n + RELU\n + SELU\n + GELU\n + ELU\n + Tanh\n + Sigmoid\n + Hard Sigmoid\n + Softmax\n + Softplus\n + Softsign\n + Exponential\n + Linear\n + Trainer (Generic TF2.0 train and validation pipelines) *\n\nNote: `*` - Support coming soon\n\n------------------\n\n\n## Installation\n\nBefore installing `TensorHub`, please install its backend engines: TensorFlow (*TensorFlow 2.0 is Recommended*).\n\n+ [Install TensorFlow and Get Started!](https://www.tensorflow.org/install)\n\n+ [Build, deploy, and experiment easily with TensorFlow](https://www.tensorflow.org/)\n\nYou may also consider installing the following **optional dependencies**:\n\n+ [cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/) (*Recommended if you plan on running Keras on GPU*).\n+ HDF5 and [h5py](http://docs.h5py.org/en/latest/build.html) (*Required if you plan on saving Keras models to disk*).\n\nThen, you can install TensorHub itself.\n\n```sh\nsudo pip install tensorhub==1.0.0a3\n```\n\nIf you are using a virtualenv, you may want to avoid using sudo:\n\n```sh\npip install tensorhub==1.0.0a3\n```\n\n+ **Alternatively: Install TensorHub from the GitHub source:**\n\nFirst, clone TensorHub using `git`:\n\n```sh\ngit clone https://github.com/nityansuman/tensorhub.git\n```\n\nThen, `cd` to the TensroHub folder and run the install command:\n```sh\ncd tensorhub\nsudo python setup.py install\n```\n\n------------------\n\n\n## Support\n\nYou can also post **bug reports and feature requests** (only) in [GitHub issues](https://github.com/nityansuman/tensorhub/issues). Make sure to read [our guidelines](https://github.com/nityansuman/tensorhub/blob/master/CONTRIBUTING.md) first.\n\nWe are eager to collaborate with you. Feel free to open an issue on or send along a pull request.\nIf you like the work, show your appreciation by \"FORK\", \"STAR\", or \"SHARE\".\n\n[![Love](https://forthebadge.com/images/badges/built-with-love.svg)](https://GitHub.com/nityansuman/tensorhub/)\n\n\n", "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/nityansuman/tensorhub", "keywords": "", "license": "MIT", "maintainer": "", "maintainer_email": "", "name": "tensorhub", "package_url": "https://pypi.org/project/tensorhub/", "platform": "", "project_url": "https://pypi.org/project/tensorhub/", "project_urls": { "Homepage": "https://github.com/nityansuman/tensorhub" }, "release_url": "https://pypi.org/project/tensorhub/1.0.0a4/", "requires_dist": [ "numpy (>=1.16)", "tensorflow (>=2.0.0a0)" ], "requires_python": "", "summary": "Deep Learning for Everybody.", "version": "1.0.0a4" }, "last_serial": 5539848, "releases": { "1.0.0a3": [ { "comment_text": "", "digests": { "md5": "87036ef7dfe5f3538ef402025e3e0509", "sha256": "7acb8e31a966b89c0707e7b9699d202955d8ed25714fdbce5cd424b6e6d0578b" }, "downloads": -1, "filename": "tensorhub-1.0.0a3-py3-none-any.whl", "has_sig": false, "md5_digest": "87036ef7dfe5f3538ef402025e3e0509", "packagetype": "bdist_wheel", "python_version": "py3", "requires_python": null, "size": 28913, "upload_time": "2019-07-16T09:33:57", "url": "https://files.pythonhosted.org/packages/25/e5/3bf94866fd38836bce65797ec26b842eaa9180d98d7da5779537c4b1134b/tensorhub-1.0.0a3-py3-none-any.whl" }, { "comment_text": "", "digests": { "md5": "a97b25bd59cd405acc513cc851340fa8", "sha256": "948b4fb7a3c18251bc03c56ed175f60b4d3e7f2d921c0e8f22348e63b3f54bc5" }, "downloads": -1, "filename": "tensorhub-1.0.0a3.tar.gz", "has_sig": false, "md5_digest": "a97b25bd59cd405acc513cc851340fa8", "packagetype": "sdist", "python_version": "source", "requires_python": null, "size": 17984, "upload_time": "2019-07-16T09:33:59", "url": "https://files.pythonhosted.org/packages/3b/92/4e3fa5a4493a03372f5a222c9a822e9d93c7da17ab8f625d4943c3c4a46f/tensorhub-1.0.0a3.tar.gz" } ], "1.0.0a4": [ { "comment_text": "", "digests": { "md5": "952bd9633a984a0bbb592ecf7334b10b", "sha256": "bf134b7639623a4ef0c099cb8ff724842588a41c355766f43b6945f55b72cc15" }, "downloads": -1, "filename": "tensorhub-1.0.0a4-py3-none-any.whl", "has_sig": false, "md5_digest": "952bd9633a984a0bbb592ecf7334b10b", "packagetype": "bdist_wheel", "python_version": "py3", "requires_python": null, "size": 28894, "upload_time": "2019-07-16T10:54:26", "url": "https://files.pythonhosted.org/packages/2d/80/8bcaba192572e9993d991cf3198c4850d9dab5fae820dd3ce33a8f7a083c/tensorhub-1.0.0a4-py3-none-any.whl" }, { "comment_text": "", "digests": { "md5": "3874e7cc5168a0b7b3ca722aa51140ca", "sha256": "4f9504ac6a2ba7d7f8b0d6034b3705950ed1817f38f4f942a3c115072ec5b3e3" }, "downloads": -1, "filename": "tensorhub-1.0.0a4.tar.gz", "has_sig": false, "md5_digest": "3874e7cc5168a0b7b3ca722aa51140ca", "packagetype": "sdist", "python_version": "source", "requires_python": null, "size": 17944, "upload_time": "2019-07-16T10:54:29", "url": "https://files.pythonhosted.org/packages/5d/36/58ccb672937c2916fb0dcbb424cc281f2ffa443c71df047534cda6ffa0b7/tensorhub-1.0.0a4.tar.gz" } ] }, "urls": [ { "comment_text": "", "digests": { "md5": "952bd9633a984a0bbb592ecf7334b10b", "sha256": "bf134b7639623a4ef0c099cb8ff724842588a41c355766f43b6945f55b72cc15" }, "downloads": -1, "filename": "tensorhub-1.0.0a4-py3-none-any.whl", "has_sig": false, "md5_digest": "952bd9633a984a0bbb592ecf7334b10b", "packagetype": "bdist_wheel", "python_version": "py3", "requires_python": null, "size": 28894, "upload_time": "2019-07-16T10:54:26", "url": "https://files.pythonhosted.org/packages/2d/80/8bcaba192572e9993d991cf3198c4850d9dab5fae820dd3ce33a8f7a083c/tensorhub-1.0.0a4-py3-none-any.whl" }, { "comment_text": "", "digests": { "md5": "3874e7cc5168a0b7b3ca722aa51140ca", "sha256": "4f9504ac6a2ba7d7f8b0d6034b3705950ed1817f38f4f942a3c115072ec5b3e3" }, "downloads": -1, "filename": "tensorhub-1.0.0a4.tar.gz", "has_sig": false, "md5_digest": "3874e7cc5168a0b7b3ca722aa51140ca", "packagetype": "sdist", "python_version": "source", "requires_python": null, "size": 17944, "upload_time": "2019-07-16T10:54:29", "url": "https://files.pythonhosted.org/packages/5d/36/58ccb672937c2916fb0dcbb424cc281f2ffa443c71df047534cda6ffa0b7/tensorhub-1.0.0a4.tar.gz" } ] }