Currently trying to use federated analytics (and eventually federated learning) at work. We are exploring PyTorch Federated and TensorFlow Federated. When I watched the TensorFlow Federated Tutorials on Google TechTalk, all tutorials were being connected to simulated data sets available locally and they confirmed that TensorFlow Federated is only ready for simulations and not for production. Do we know when TensorFlow Federated will be ready for production?
TFF is still only for research and simulations for now.
Related
I am very interested in federated systems and i was trying one of the pre trained multilingual models such as this notebook Multi_Lingual_Training_and_models.
I was looking for any tutorials using TFF or Flower frameworks that handle csv datasets.
So could you suggest any tutorials or github repositories to help me do that with TFF or Flower!
I have built an XGBoost Classifier and RandomForest Classifier model for the audio classification project. I want to deploy these models which are saved in pickle (.pkl) format on AWS Sagemaker. From what I have observed, there isn't a lot of resources available online. Can anyone guide me with the steps and if possible also provide the code? I already have the models built and I'm just left with deploying it on Sagemaker.
By saying that you want to deploy to sagemaker, I assume you mean a sagemaker endpoint.
The answer is sagemaker inference toolkit. It's basically about educating sagemaker how to load and do inference. More details here: https://github.com/aws/sagemaker-inference-toolkit and here is an example implementation: https://github.com/aws/amazon-sagemaker-examples/tree/master/advanced_functionality/multi_model_bring_your_own
I'm using a bunch of images to train my Tensorflow Image recognition project using this tutorial https://codelabs.developers.google.com/codelabs/tensorflow-for-poets/index.html#4
Actually I need a lot of Cpu to train my model and it takes a lot of time on my laptop.
I have registered a Google ML account and started this tutorial:
https://cloud.google.com/ml/docs/quickstarts/training
Everything is set up and running but this is for mnist sample code. There is no image_retraining sample code like the retrain.py from tensorflow.
Looking for some examples on how to to run the Tensorflow Image Recognition script retrain in Google ML.
IBM Watson has a capability where you can train the classifiers on Watson using your images but I am unable to find a similar capability on Google Cloud Vision API? What I want is that I upload 10-15 classes of images and on the bases of upload images classify any images loaded after that. IBM Bluemix (Watson) has this capability but their pricing is significantly higher than Google. I am open to other services as well, if prices ares below Google's
As far as I know Google Cloud Vision API cannot be trained with your custom data. However, there is a service called vize.ai, where you can define your custom classes and upload the images, the training is for free and the prices for API usage are below Google's and IBM's.
Disclaimer: I'm vize.it co-founder
Edit: Link changed
You can train your own models using Cloud AutoML Vision. There are 2 different ways to do this:
Cloud-hosted models.
Edge exportable models.
With some work you can train a model for free using TensorFlow - see the model training section.
However, they have released an already trained model, so if you're lucky and what you want to classify already overlaps with their model, then no training is needed.
Azure has started this now, google for "azure custom vision" this is still a preview service but with good accuracy at least for our workload which is preschool children images.
I'm working on a project where I train a text classifier and I need to create a web app to let the user enter text for classification. Currently all the code is written in Python and I'm using scikit-learn library. I've encountered a problem installing the scikit-learn on heroku, in order for my Python code to run on the server. I don't mind changing everything (Python language, Flask web framework, scikit learning library, heroku web-app hosting services), I just need to get this thing to work :)
Do any one of you in CV community had any experience in making a web-app that uses a learning library online? The web app hosting should be a free one though, as this project is not commercial, and also it would be very nice to have Python behind the scenes.
N.B. The classifiers that should be supported by the library are multiclass svm and naive bayes.
How about trying google app engine? It has python (2.5 and 2.7) and can be free.