I was a lot skeptical about tensorflow 2.0 when I heard that it has been made like Keras. Thb, I absolutely hate keras as something always feels missing in the framework (maybe a lack of control). But the introduction of subclassing really exites me now. Its good that TF is a lot like Torch now, the flexibility looks amazing. Good work tensorflow!
Tensorflow 2.0 will make modelling so much expressive/artistic with the added touch of lower level control if needed , plus easier deployment? Wow! just wow! ! What an exciting time to exist.😂😂😂
Google programmers are so incredibly well spoken and good at presenting, either there must be some training internally or they select for this at interviews
The tutorials for beginners in the website do not explain most things (what are all the lines doing). I code in python for 15 years. I have never have a trouble understanding any library anything until now.
When is CSV data going to be done? It's still in the experimental section, and TF has been out 4 years. Secondly, there is no instructional coverage of CSV at all in the brand new Tensorflow Specialization on coursera by Andrew Ng and Google's trainer. The whole UCI ML dataset lib is CSV, in case you didn't notice yet somehow. Unlike software app developers, data scientists don't get to create whatever dataset we want just like app developers seem to do. We have to do work on the data that the organization has provided and in case data science is a new concept for you, that is largely CSV for example Higgs Boson dataset. But both Microsoft and Google still seem to treat data scientists as if software apps with unstructured vision/NLP are all that really matters, and the structured datasets (ie, CSV) are for science nerds and don't really matter much so lets let our summer interns take care of CSV support while we're on vacation. "It should be easy." Did Microsoft actually try to use the CNTKTextFormatReader (later simply CTF Reader)? DUDES go to Kaggle and try the BOSCH dataset (not Titanic enough of that) and get back to me stories about easy CSVs again, bro. Try to use your own tf.data.experimental.CsvDataset and get back to us.
Best introduction to Tensorflow 2.0 ever
Josh is a God of explanations in my Universe.
Thank you!
Very well explained ! I am impressed with this guy's articulation skills.
Awesome presentation! I was attracted to simplicity of PyTorch, but the new features of TF 2.0 look very promising.
I was a lot skeptical about tensorflow 2.0 when I heard that it has been made like Keras. Thb, I absolutely hate keras as something always feels missing in the framework (maybe a lack of control). But the introduction of subclassing really exites me now. Its good that TF is a lot like Torch now, the flexibility looks amazing. Good work tensorflow!
Tensorflow 2.0 will make modelling so much expressive/artistic with the added touch of lower level control if needed , plus easier deployment? Wow! just wow! ! What an exciting time to exist.😂😂😂
This is great, why did it take me so long to discover this?
Amazing video!!
WOW ! i am very excited to try ...these new features of TF 2.0
Great presentation, thanks Josh and the TF2 team.
Google programmers are so incredibly well spoken and good at presenting, either there must be some training internally or they select for this at interviews
Like this new way of writing custom stuff in the low level with tf.Module instead of keras
wow! super cool presentation! Thanks, Josh!!
That's a super cool presentation! I believe even someone who is not familiar with TF and ML will have the motivation to try it out after this talk.
Great explanations Josh, as usual! Thanks!
Amazing presentation, I really enjoyed it.
This is an awesome tutorial !
Thanks for an excellent presentation.
Great presentation. Are there slides available for this?
Absolutely fantastic talk, thank you
Wonderful presentation! Thanks so much.
Nice job, Josh!
تحياتي الخالصة
Nice presentation, thaks Josh!
really good
TensorFlow 2 is so amazing!!!
And I would use PyTorch 1.3
Me too
That was amazing Josh
Nice job team
Where can we find the slides please ?
good job!
Is keras 100 percent comparible with tensorflow 2.0?
Great talk, josh!
Excellent!
Superb
Anyone know where can I find the share collab projects?
The tutorials for beginners in the website do not explain most things (what are all the lines doing). I code in python for 15 years. I have never have a trouble understanding any library anything until now.
Wonderful
17:18 Can we use callbacks on train_on_batch? Any references for that?
Hello, are someone exciting with new version TF2? I saw a lot of new things release. Let's discuss about it
IMPRESSED
slides for this presentation: it has lots of useful links... drive.google.com/file/d/1YG3QX8syS-t8WyWOjGaGrtrsvSDqSTLx/view
Cannot install tensorflow as wrapt is not getting installed
HELP!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
When is CSV data going to be done? It's still in the experimental section, and TF has been out 4 years. Secondly, there is no instructional coverage of CSV at all in the brand new Tensorflow Specialization on coursera by Andrew Ng and Google's trainer. The whole UCI ML dataset lib is CSV, in case you didn't notice yet somehow. Unlike software app developers, data scientists don't get to create whatever dataset we want just like app developers seem to do. We have to do work on the data that the organization has provided and in case data science is a new concept for you, that is largely CSV for example Higgs Boson dataset. But both Microsoft and Google still seem to treat data scientists as if software apps with unstructured vision/NLP are all that really matters, and the structured datasets (ie, CSV) are for science nerds and don't really matter much so lets let our summer interns take care of CSV support while we're on vacation. "It should be easy." Did Microsoft actually try to use the CNTKTextFormatReader (later simply CTF Reader)? DUDES go to Kaggle and try the BOSCH dataset (not Titanic enough of that) and get back to me stories about easy CSVs again, bro. Try to use your own tf.data.experimental.CsvDataset and get back to us.
Nothing simple conceptually? What does that mean? 9:35
nice!
But early stopping shouldn't be used because it will lead orthogonalization
need help in prediction from trained & saved model, same as 36:36 ,,, any one expert here ?
Is it the google's server get down?
👍👍👍👍👍
import tensorflow as tf
I am learning
What are you doing here?
Excellent presentation. Thanks!
Superb