We performed a comparison between PyTorch and TensorFlow based on real PeerSpot user reviews.
Find out in this report how the two AI Development Platforms solutions compare in terms of features, pricing, service and support, easy of deployment, and ROI."The tool is very user-friendly."
"It's been pretty scalable in terms of using multiple GPUs."
"The framework of the solution is valuable."
"I like that PyTorch actually follows the pythonic way, and I feel that it's quite easy. It's easy to find compared to others who require us to type a long paragraph of code."
"Its interface is the most valuable. The ability to have an interface to train machine learning models and construct them with the high-level interface, without excess busting and reconstructing the same technical elements, is very useful."
"yTorch is gaining credibility in the research space, it's becoming easier to find examples of papers that use PyTorch. This is an advantage for someone who uses PyTorch primarily."
"TensorFlow is a framework that makes it really easy to use for deep learning."
"Optimization is very good in TensorFlow. There are many opportunities to do hyper-parameter training."
"Our clients were not aware they were using TensorFlow, so that aspect was transparent. I think we personally chose TensorFlow because it provided us with more of the end-to-end package that you can use for all the steps regarding billing and our models. So basically data processing, training the model, evaluating the model, updating the model, deploying the model and all of these steps without having to change to a new environment."
"The most valuable feature of TensorFlow is deep learning. It is the best tool for deep learning in the market."
"What made TensorFlow so appealing to us is that you could run it on a cluster computer and on a mobile device."
"It is open-source, and it is being worked on all the time. You don't have to pay all the big bucks like Azure and Databricks. You can just use your local machine with the open-source TensorFlow and create pretty good models."
"It provides us with 35 features like patch normalization layers, and it is easy to implement using the Kras library when the Kaspersky flow is running behind it."
"It empowers us to seamlessly create and deploy machine learning models, offering a versatile solution for implementing sophisticated environments and various types of AI solutions."
"I would like a model to be available. I think Google recently released a new version of EfficientNet. It's a really good classifier, and a PyTorch implementation would be nice."
"There is not enough documentation about some methods and parameters. It is sometimes difficult to find information."
"On the production side of things, having more frameworks would be helpful."
"The training of the models could be faster."
"PyTorch could make certain things more obvious. Even though it does make things like defining loss functions and calculating gradients in backward propagation clear, these concepts may confuse beginners. We find that it's kind of problematic. Despite having methods called on loss functions during backward passes, the oral documentation for beginners is quite complex."
"I've had issues with stability when I use a lot of data and try out different combinations of modeling techniques."
"I would love to have a user interface like a programming interface. You need to have a set of menus where you can put things together in a graphical interface. The complete automation of the integration of the modules would also be interesting. It’s more like plumbing as opposed to a fully automated environment."
"TensorFlow deep learning takes a lot of computation power. The more systems you can use, the easier it is. That's a good ability, if you can make a system run immediately at the same time on the same task, it's much faster rather than you having one system running which is slower. Running systems in parallel is a complex situation, but it can improve. There is a lot of work involved."
"It would be cool if TensorFlow could make it easier for companies like us to program for running it across different hyperscalers."
"There are a lot of problems, such as integrating our custom code. In my experience model tuning has been a bit difficult to edit and tune the graph model for best performance. We have to go into the model but we do not have a model viewer for quick access."
"Personally, I find it to be a bit too much AI-oriented."
"TensorFlow Lite only outputs to C."
"It would be nice if the solution was in Hungarian. I would like more Hungarian NAT models."
"It doesn't allow for fast the proto-typing. So usually when we do proto-typing we will start with PyTorch and then once we have a good model that we trust, we convert it into TensorFlow. So definitely, TensorFlow is not very flexible."
PyTorch is ranked 10th in AI Development Platforms with 6 reviews while TensorFlow is ranked 4th in AI Development Platforms with 16 reviews. PyTorch is rated 8.6, while TensorFlow is rated 9.0. The top reviewer of PyTorch writes "Offers good backward compatible and simple to use". On the other hand, the top reviewer of TensorFlow writes "Effective deep learning, free to use, and highly stable". PyTorch is most compared with OpenVINO, MXNet, Microsoft Azure Machine Learning Studio, Caffe and Google Vertex AI, whereas TensorFlow is most compared with Microsoft Azure Machine Learning Studio, Google Vertex AI, OpenVINO, Hugging Face and Azure OpenAI. See our PyTorch vs. TensorFlow report.
See our list of best AI Development Platforms vendors.
We monitor all AI Development Platforms reviews to prevent fraudulent reviews and keep review quality high. We do not post reviews by company employees or direct competitors. We validate each review for authenticity via cross-reference with LinkedIn, and personal follow-up with the reviewer when necessary.