In federated learning, stragglers are devices which sends back an update to the parameter server with a considerable delay. I would like to simulate stragglers in TensorFlow federated. However, I don't see any documentation or examples on this topic for this framework.
Can you suggest any link or indication on how to simulate different training times for devices in the same FL session?
Related
Currently I am studying the usage of Apache Spark 3.0 with Rapids GPU Acceleration. In the official spark-rapids docs I came across this page which states:
There are cases where you may want to get access to the raw data on the GPU, preferably without copying it. One use case for this is exporting the data to an ML framework after doing feature extraction.
To me this sounds as if one could make data that is already available on the GPU from some upstream Spark ETL process directly available to a framework such as Tensorflow or PyTorch. If this is the case how can I access the data from within any of these frameworks? If I am misunderstanding something here, what is the quote exactly referring to?
The link you references really only allows you to get access to the data still sitting on the GPU, but using that data in another framework, like Tensorflow or PyTorch is not that simple.
TL;DR; Unless you have a library explicitly setup to work with the RAPIDS accelerator you probably want to run your ETL with RAPIDS, then save it, and launch a new job to train your models using that data.
There are still a number of issues that you would need to solve. We have worked on these in the case of XGBoost, but it has not been something that we have tried to tackle for Tensorflow or PyTorch yet.
The big issues are
Getting the data to the correct process. Even if the data is on the GPU, because of security, it is tied to a given user process. PyTorch and Tensorflow generally run as python processes and not in the same JVM that Spark is running in. This means that the data has to be sent to the other process. There are several ways to do this, but it is non-trivial to try and do it as a zero-copy operation.
The format of the data is not what Tensorflow or PyTorch want. The data for RAPIDs is in an arrow compatible format. Tensorflow and PyTorch have APIs for importing data in standard formats from the CPU, but it might take a bit of work to get the data into a format that the frameworks want and to find an API to let you pull it in directly from the GPU.
Sharing GPU resources. Spark only recently added in support for scheduling GPUs. Prior to that people would just launch a single spark task per executor and a single python process so that the python process would own the entire GPU when doing training or inference. With the RAPIDS accelerator the GPU is not free any more and you need a way to share the resources. RMM provides some of this if both libraries are updated to use it and they are in the same process, but in the case of Pytorch and and Tensoflow they are typically in python processes so figuring out how to share the GPU is hard.
I have implemented a neural network model using Python and Tensorflow, which normally runs on my own computer.
Now I would like to train it on new datasets on the Google Cloud Platform. Do you think it is possible? Do I need to change my code?
Thank you very much for your help!
Google Cloud offers the Cloud ML Engine service, which allows to train your models and perform predictions without the need of running and maintaining an instance with the required software.
In order to run the TensorFlow NN models you already have, you will not need to change your code, you will only have to package the trainer appropriately, as described in the documentation, and run a ML Engine job that performs the training itself. Once you have your model, you can also deploy it in the same service and later get predictions with different features depending on your requirements (urgency in getting the predictions, data set sources, etc.).
Alternatively, as suggested in the comments, you can always launch a Compute Engine instance and run there your TensorFlow model as if you were doing it locally in your computer. However, I would strongly recommend the approach I proposed earlier, as you will be saving some money, because you will only be charged for your usage (training jobs and/or predictions) and do not need to configure an instance from scratch.
Is my understanding correct that model_deploy lets the user train a model using multiple devices on a single machine? The basic premise seems that the clone devices do variable sharing and variables get distributed to param servers in a round-robin fashion.
On the other hand distributed tensorflow framework enables the user to train a model through a cluster. A Cluster lets the user train a model using multiple devices across multiple servers.
I think the Slim documentation is very slim and the point has been raised couple of times already: Configuration/Flags for TF-Slim across multiple GPU/Machines
Thank you.
I'm trying to launch my deep network on multiple devices. I'm using the slim framework provided in tensorflow/models/slim. But I have difficulty in understanding the terms in the source code.
What are the meanings of clone, replica, parameter-server(ps) and worker? Can anyone explain them to a real system muggle like me?
I have just started experimenting with Deep Learning and Computer Vision technologies. I came across this awesome tutorial. I have setup the TensorFlow environment using docker and trained my own sets of objects and it provided greater accuracy when I tested it out.
Now I want to make the same more real-time. For example, instead of giving an image of an object as the input, I want to utilize a webcam and make it recognize the object with the help of TensorFlow. Can you guys guide me with the right place to start with this work?
You may want to look at TensorFlow Serving so that you can decouple compute from sensors (and distribute the computation), or our C++ api. Beyond that, tensorflow was written emphasizing throughput rather than latency, so batch samples as much as you can. You don't need to run tensorflow at every frame, so input from a webcam should definitely be in the realm of possibilities. Making the network smaller, and buying better hardware are popular options.