Top 23 Python Tensorflow Projects
-
keras
Deep Learning for humans
Then it's maybe a version bug problem, try to update to the latest tensorflow and keras version. It seems to appear in this issue and hasn't been resolved, switch to Pytorch maybe ?
-
transformers
🤗Transformers: State-of-the-art Natural Language Processing for Pytorch and TensorFlow 2.0.
4. Repeat
For step 3 you need to send the gradients from each GPU somewhere, and then send back either the averaged gradient or the updated model weights. So when the model is large (say, 3GB for GPT 774M!) that's a lot of GPU-GPU communication!
You're right that for the vast majority of ML cases, the models are small enough that the synchronization cost is negligible, though.
I wrote up some benchmarks here:
-
Real-Time-Voice-Cloning
Clone a voice in 5 seconds to generate arbitrary speech in real-time
you can also https://github.com/CorentinJ/Real-Time-Voice-Cloning use this to create voices with the power of ai
-
data-science-ipython-notebooks
Data science Python notebooks: Deep learning (TensorFlow, Theano, Caffe, Keras), scikit-learn, Kaggle, big data (Spark, Hadoop MapReduce, HDFS), matplotlib, pandas, NumPy, SciPy, Python essentials, AWS, and various command lines.
Latest mention: Learning webscraping, data analysis, and visualization, where should I start? | reddit.com/r/learnpython | 2021-01-23data science ipython notebooks
-
Mask_RCNN
Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
matterport / Mask_RCNN
-
spleeter
Deezer source separation library including pretrained models.
Latest mention: Does anyone know good (free?) software to extract vocals from a song | reddit.com/r/Beatmatch | 2021-01-25You can use spleeter or SpleeterGUI. One of the best there is and free to use.
-
facenet
Face recognition using Tensorflow
ML training is practically impossible on micro-controllers. Inferencing on the other hand is quite doable, especially if aided by a [TPU coprocessor](https://coral.ai/products/accelerator/). Supposedly with the TPU you can do some quantization-aware training, but I haven't tried this. I am working on a security system that does facial recognition to recognize me and some friends and considers anyone else as an intruder. How I am doing this is by retraining [Facenet](https://github.com/davidsandberg/facenet) with my facial embeddings. Use something like Haar Cascade in OpenCV to get the bounding box for a face and put it through the model to extract face embeddings. You can then save these embeddings as a sort of databases for the faces you want it to recognize during the inferencing phase. After that you can impose something like a SVM classifier to say who in your face database it is. One thing I will note is that the problem is even easier if you are only concerned with one face - in which case it is technically face identification - not recognition. If that is the case, you only need to do a difference calculation between the embeddings you saved during training and the result output from inferencing. If you do end up using the TPU, you can connect to it over USB from inside a container (I only know how to do this in Docker though) too. Hope this was helpful. I am actually looking to use a k8s cluster eventually too as a sort of smart hub for my security system and other devices so I can handle much more traffic (not sure if this is overkill or not on the pi 4s).
-
DeepCreamPy
Decensoring Hentai with Deep Neural Networks
-
tflearn
Deep learning library featuring a higher-level API for TensorFlow.
-
nni
An open source AutoML toolkit for automate machine learning lifecycle, including feature engineering, neural architecture search, model compression and hyper-parameter tuning.
microsoft / nni
-
bert-as-service
Mapping a variable-length sentence to a fixed-length vector using BERT model
Latest mention: Needed 100% to pass a safety quiz, need to wait a week to retake | reddit.com/r/mildlyinfuriating | 2021-01-12You joke but
-
datasets
🤗 The largest hub of ready-to-use NLP datasets for ML models with fast, easy-to-use and efficient data manipulation tools
Latest mention: [P] 611 text datasets in 467 languages in the new v1.2 release of HuggingFace datasets library | reddit.com/r/MachineLearning | 2021-01-05There will be 13 more bytthe end of this week, from Microsoft CodeXGlue, I had not the time to fix my PR earlier : https://github.com/huggingface/datasets/pull/997 .
-
MMdnn
MMdnn is a set of tools to help users inter-operate among different deep learning frameworks. E.g. model conversion and visualization. Convert models between Caffe, Keras, MXNet, Tensorflow, CNTK, PyTorch Onnx and CoreML.
Yes. But that's missing the point. Almost everything in the space uses a framework other than Core ML. Therefore most people need good development support for PyTorch/Tensorflow etc., not Core ML. The fact that Apple has a tool to import/convert models is nice, but not relevant. Also, there is onnx as an exchange format between the frameworks, and tools like MMdnn and others.
-
textgenrnn
Easily train your own text-generating neural network of any size and complexity on any text dataset with a few lines of code.
Latest mention: I forced a bot to read 4000 Dan Andrews Facebook posts and asked it to write its own. Here's what it came up with. | reddit.com/r/melbourne | 2020-12-25For Danbot V1 (The one using 4,000 posts and GPT-2), I used Kevinzg's Facebook Scraper to get my source material, and used minimaxir's text generator to write the posts.
-
best-of-ml-python
🏆 A ranked list of awesome machine learning Python libraries. Updated weekly.
Latest mention: best-of-python: A ranked list of awesome Python libraries and tools | reddit.com/r/Python | 2021-01-14Here ya go: https://github.com/ml-tooling/best-of-ml-python/pull/47
-
mindsdb
Predictive AI layer for existing databases.
-
segmentation_models
Segmentation models with pretrained backbones. Keras and TensorFlow Keras.
Latest mention: Any way to speed up inference prepare operations on host (CPU)? | reddit.com/r/tensorflow | 2021-01-18That is just U-net from this repo, anything aside is slicing images to fit into window and predict call. I measure time of predict() and it is the same as profiler numbers, so definitely my other operations are beyond profiler. C API code is just creating tensors and calling TF_SessionRun plus slice operations with opencv. Can't post code, sorry.
-
gpt-2-simple
Python package to easily retrain OpenAI's GPT-2 text-generating model on new texts
The GPT-2 weights were released by OpenAI when GPT-2 was released. (https://github.com/openai/gpt-2)
Around that time (since no one else was doing it) I released a wrapper to streamline that code and make it much easier to finetune on your own data. (https://github.com/minimaxir/gpt-2-simple)
Nowadays, the easiest way to interact with GPT-2 is to use the transformers library (https://github.com/huggingface/transformers), of which I've created a much better library for GPT-2 that leverages it. (https://github.com/minimaxir/aitextgen)
-
tf-quant-finance
High-performance TensorFlow library for quantitative finance.
Latest mention: Show HN: A high-performance TensorFlow library for quantitative finance | reddit.com/r/patient_hackernews | 2021-01-19 -
lingvo
Lingvo
Yes, there are really good open source speech to text tools (automatic speech recognition (ASR) is the common name for that).
Kaldi (https://kaldi-asr.org/) is probably the most well known, and supports hybrid NN-HMM and lattice-free MMI models. Kaldi is used by many people both in research and in production.
Lingvo (https://github.com/tensorflow/lingvo) is the open source version of Google speech recognition toolkit, with support mostly for end-to-end models.
ESPNet (https://github.com/espnet/espnet) is good and well known for end-to-end models as well.
RASR (https://github.com/rwth-i6/rasr) + RETURNN (https://github.com/rwth-i6/returnn) are very good as well, both for end-to-end models and hybrid NN-HMM, but they are for non-commercial applications only (or you need a commercial licence) (disclaimer: I work at the university chair which develops these frameworks).
-
coremltools
Core ML tools contain supporting tools for Core ML model conversion, editing, and validation.
And did you even know that Apple provides a CoreML conversion toolkit for converting models from PyTorch, TF, SKLearn and Keras to CoreML? You can comfortably train in your preferred tool and then convert the models to CoreML. (https://github.com/apple/coremltools). It currently supports a lot of conversions
-
jina
An easier way to build neural search in the cloud
-
agents
TF-Agents is a library for Reinforcement Learning in TensorFlow
Latest mention: Can somebody give me reinforcement learning code example. | reddit.com/r/MLQuestions | 2021-01-22
Index
What are some of the best open-source Tensorflow projects in Python? This list will help you:
| Project | Stars | |
|---|---|---|
| 1 | keras | 50,548 |
| 2 | transformers | 39,664 |
| 3 | Real-Time-Voice-Cloning | 22,637 |
| 4 | data-science-ipython-notebooks | 20,109 |
| 5 | Mask_RCNN | 19,276 |
| 6 | spleeter | 15,361 |
| 7 | facenet | 11,457 |
| 8 | DeepCreamPy | 10,400 |
| 9 | tflearn | 9,505 |
| 10 | nni | 8,798 |
| 11 | bert-as-service | 8,791 |
| 12 | datasets | 6,551 |
| 13 | MMdnn | 5,180 |
| 14 | textgenrnn | 4,230 |
| 15 | best-of-ml-python | 3,390 |
| 16 | mindsdb | 3,170 |
| 17 | segmentation_models | 2,897 |
| 18 | gpt-2-simple | 2,455 |
| 19 | tf-quant-finance | 2,376 |
| 20 | lingvo | 2,190 |
| 21 | coremltools | 2,085 |
| 22 | jina | 1,917 |
| 23 | agents | 1,767 |