So, true. However if you have good knowledge of building NN then a framework/library should not be a barrier as these are just tools and tools change over time.
I'd suggest learning TensorFlow first because it's backed by Google and has a certification behind it which will lead you to google's Data Engineer cert and Machine Learning Cert. Zerotomastery and Coursera have really good prep courses for tensorflow certification and pytorch for that matter
i am currently learning pytorch. I checked the job requirements, they ask you both. I think when I get proficient with PyTorch, I will be able to pick up Tensorflow as well.
I tried to install about 10 times and failed. Until, for some reason i dont know why, It randomly started working. Maybe my computer needed some rest before import the lib.
Perhaps you're trying to install the GPU version on windows. I learned they no longer support GPUs on Windows native. I had to install an older version
Just keep learning. When I'm not on a contract I'm learning and training. Doing PyTorch myself. I have C#.NET, Java, Python, Angular and React. When I started I did COBOL, DB2, CICS, JCL on mainframes.
Did you practice maths too as well? Like linear algebra, calculus, etc.. like, do I have to practice these topics in math well for good in machine learning?
Sure do! Building models is super easy, thanks to all of these great Python libraries. But when things don’t work, having an underlying understanding of the nuts and bolts (aka the math) behind the models will make your life a ton easier.
That may depend on what field you works at since machine learning is very broad. But generally speaking, most companies use popular frameworks like PyTorch, TF 2, Jax, ONNX. These frameworks solve low-level problems and provide some optimizations which would be a pain in ass to implement from the scratch
I am a beginner wanting to learn computer vision, I did some basic projects in open cv, yolo. Now I want to go in-depth in either tensor flow/Pytorch what should i choose ?
This content is insightful. A matching book would be my suggestion if you're invested in this topic. "From Bytes to Consciousness: A Comprehensive Guide to Artificial Intelligence" by Stuart Mills
I find pytorch self manages memory out of the box better than tensorflow. Pytorch is like an automatic car. Tensorflow is more configurable but more fiddly.
For real i think pytorch gives u advanced controls for you to operate with... in tensorflow i pretty much hate its graph execution and build up algorithm
Honestly I have no clue how people get hired as Data Scientists with not at least a Masters. It is pretty much impossible here getting hired even with a Masters. Most expect a PhD
I encountered an issue where, upon attempting to execute the code li=[1,2,3,4,5,6] followed by print(li), the Python interpreter unexpectedly terminated without providing any error message or indication of the problem. Can you please tell me what happened??
If you know math potato potato what is the difference they are just the same thing doing the same math. After I learn pytorch I learn tensorflow in just 5 hours lol. Just the name of the function change nothing difference.
@AnushkaChathuranga-cw7tcNo it’s not. Learning to build a NN from scratch is invaluable knowledge. That means if one knows the core building blocks of NN, one can also know what to improve if a NN starts acting weird (Presumably if you built it using tf or PyTorch). Debugging your NN is hard if you don’t know how NN’s work.
Exactly and this way of thinking is true to any occupation that requires problem solving, it is absolutely necessary to understand the problems you need to tackle on a lower level to be competent and successful
First of all i am big fan of you.. I started my journey with you when you uploded your Tensorflow object detection setup video. I am Arbaj S. looking for job in data science I am perfectly fit for this role because I have following skills, 1. Advance machine learning (supervised, unsupervised) 2. Advance Computer vision(object detection, yolo, CNN, RNN, ANN, fast RCNN, and many more) 3. Natural Language Processing (Basic to advance), transformers, GPT, BERT, LSTM 4. Time series & Recommender system (learning) 5. Statistics and maths 6. PYTHON for coding 7. Tableau for visualization 8. SQL database 9. Spark(learning) 10. Azure Data science(basic) I studied this skills since 2020 parallel with my graduation, I dig into data and get a content out of it. I collect, process, manage, visualise and interpret data to accomplish this, I focus on the value of the data and then manage it in a desirable way to visualise and interpret in a better way. If u find these skills are enough to get job please guide me ..... Arbajsofficial@gmail.com +917058988589 Please please please BIG FAN Spiderman😅
That's the most worthless advice for someone who wants to learn it for fun. What potential fucking employer? I'm not trying to get any work in ML. I want to know what the differences are, what I can expect, which one feels more similar to another language I might already know, etc.
I'm an ML engineer, my tip is learning both, just take one of them and go deeper, and then try to replicate the same stuff in the other In general debugging TF is harder (even now that it is basically keras, specifically when you execute stuff in graph mode) On the other hand Pytorch is more pythonic (that's why it is preferred for research) but it is not used as much as TF in industry In resume, if you have the possibility to work with pytorch go for it, but always learn TF no matter what
