[HN Gopher] What's New in TensorFlow 2.10?
       ___________________________________________________________________
        
       What's New in TensorFlow 2.10?
        
       Author : RafelMri
       Score  : 34 points
       Date   : 2022-09-06 18:43 UTC (4 hours ago)
        
 (HTM) web link (blog.tensorflow.org)
 (TXT) w3m dump (blog.tensorflow.org)
        
       | amilios wrote:
       | Is there any point on which Tensorflow actually wins out against
       | Pytorch? My sense is that Pytorch is easier for almost every if
       | not every single use case. Not to mention the 1 => 2 transition
       | gave me serious Python 2 => 3 vibes with the switch to default
       | eager execution etc. Anyone still using Tensorflow care to weigh
       | in?
        
         | jphoward wrote:
         | I use pytorch the vast majority of the time, but I think there
         | are 2 reasons tensorflow remains competitive.
         | 
         | First, is TF's deployment has always been ahead of Pytorch's
         | (although the gap is closing, especially as Onnx becomes more
         | popular).
         | 
         | However, the more important reason is how amazingly good value
         | TPUs are in terms of their RAM and FLOPS. Although pytorch has
         | XLA support, it just doesn't work as well as TF on TPU pods.
         | 
         | In Kaggle competitions, when the input data can fit on a
         | consumer GPU everyone uses pytorch. When it doesn't, everyone
         | uses the free TPUs on the Kaggle platform and reverts to
         | tensorflow/keras.
        
         | emehex wrote:
         | I think TF is still popular (with beginners, at least) because
         | of the keras API (far easier to get started!)
        
         | jpeter wrote:
         | And google started using Jax instead of Tensorflow
        
         | suresk wrote:
         | I seem to be in a small group that doesn't have a huge
         | preference between the two, maybe I'm not doing that much
         | advanced stuff?
         | 
         | If I'm just prototyping something myself, I usually reach for
         | TF first, mostly because Keras feels like the "right" level of
         | abstraction for most stuff. I used to prototype things with
         | fastai/pytorch more, but newer developers didn't like how much
         | was hidden behind multiple levels of *kwargs and some of the
         | dataloader stuff could get tricky if you tried to do anything
         | non-standard. I haven't tried PyTorch Lightning.
         | 
         | Besides the deployment story, which is pretty big and others
         | have touched on, there are some minor things that feel nice in
         | the TF/Keras ecosystem:
         | 
         | - Part of deployment, I guess, but I like that more of the
         | preprocessing can happen in the model, vs as a separate step.
         | The less transforming of data that has to be done in the
         | serving code, the fewer possibilities for things to get out of
         | sync and introduce bad data at runtime.
         | 
         | - Keras being able to infer input sizes in layers is nice for
         | avoiding a bunch of bookkeeping code to calculate layer sizes.
         | 
         | That said, I feel like maybe the TF ecosystem has more sharp
         | edges? I've encountered more than a few of them lately as I've
         | been doing work on some recommender models using tfrs. I've
         | also run into things like tensorboard logging not working with
         | entire classes of layers and causing training to crash.
         | 
         | I'm curious - what are some things about PyTorch that make it
         | better for almost every single use case?
        
         | fxtentacle wrote:
         | If you want to actually deploy AI models, TFLite is still your
         | best option. And for exporting there, you first need things to
         | run well inside regular TF.
         | 
         | Also, I often find myself using the TF data pre-processing
         | pipeline even from inside PyTorch, because it's just so much
         | easier to get excellent processing performance with TF. Not
         | sure why, but PyTorch is in many cases "unnecessarily" single-
         | threaded.
        
           | learndeeply wrote:
           | > If you want to actually deploy AI models
           | 
           | On mobile*.
           | 
           | > I often find myself using the TF data pre-processing
           | pipeline even from inside PyTorch
           | 
           | the tf.data pipeline is quite nice, has some neat auto-tuning
           | features.
        
       | Sin2x wrote:
       | Slowly becoming irrelevant: https://paperswithcode.com/trends
        
         | jorgemf wrote:
         | For that statement you would need to show industry graphs, not
         | only research
        
           | Sin2x wrote:
           | Industry does not exist in a vacuum, it follows research.
           | 
           | Additionally, see Google trends: https://trends.google.com/tr
           | ends/explore?date=today%205-y&ge...
        
       ___________________________________________________________________
       (page generated 2022-09-06 23:02 UTC)