uncertainty-baselines VS probability

Code for https://arxiv.org/abs/2106.04015 found: https://github.com/google/uncertainty-baselines

I know Reddit likes dumping on TensorFlow but it's actually really easy. Set a policy and upcast your final logits to float32.

TensorFlow-Probability

If you are interested in probabilistic programming, causal modeling and bayesian graphical modeling, I recommend checking out Tensorflow Probability (https://www.tensorflow.org/probability).

As others have said, R for academia, Python for industry. However, i'd also throw Stan into the mix, along with other PPL frameworks like Tensorflow Probability and Pyro. The latter two will require you to learn Python first, though.

This tutorial explains what is probabilistic programming & provides a review of 5 frameworks (PPLs) using an example taken from Chapter 4 of Statistical Rethinking by Dr. Richard McElreath. Frameworks (PPLs) reviewed are - Stan (https://mc-stan.org/) PyMC3 (https://docs.pymc.io/) Tensorflow Probability (https://www.tensorflow.org/probability) Pyro/NumPyro (https://pyro.ai/) Turing.jl (https://turing.ml/stable/) I also provide the basic review of a great library called arviz (https://arviz-devs.github.io/arviz/), which can be used for all the above-mentioned PPLs to do Exploratory Data Analysis of Bayesian Models. Here is the link to the notebook in which I have implemented the example model using the above Frameworks/PPLs https://colab.research.google.com/drive/1zgR2b0j2waGi1ppnIe1rw7emkbBXtMqF?usp=sharing

2021-03-20 20:09:56.451490: E tensorflow/core/platform/hadoop/hadoopfile_system.cc:132] HadoopFileSystem load error: libhdfs.so: cannot open shared object file: No such file or directory WARNING:tensorflow:From /home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/ops/distributions/distribution.py:265: ReparameterizationType.init_ (from tensorflow.python.ops.distributions.distribution) is deprecated and will be removed after 2019-01-01. Instructions for updating: The TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use tfp.distributions instead of tf.distributions. WARNING:tensorflow:From /home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflowcore/python/ops/distributions/bernoulli.py:169: RegisterKL.init_ (from tensorflow.python.ops.distributions.kullbackleibler) is deprecated and will be removed after 2019-01-01. Instructions for updating: The TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use tfp.distributions instead of tf.distributions. ERROR thonny.backend: PROBLEM WITH THONNY'S BACK-END Traceback (most recent call last): File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 1240, in wrapper result = method(self, args, *kwargs) File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 1227, in wrapper return method(self, args, *kwargs) File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 1297, in _execute_prepared_user_code exec(statements, global_vars) File "/home/pi/Desktop/security/security_system_v2.py", line 10, in import tensorflow as tf File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow/init.py", line 98, in from tensorflow_core import * File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/init.py", line 28, in from tensorflow.python import pywrap_tensorflow # pylint: disable=unused-import File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "", line 1019, in _handle_fromlist File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow/init.py", line 50, in __getattr_ module = self.load() File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow/init.py", line 44, in _load module = _importlib.import_module(self.name) File "/usr/lib/python3.7/importlib/init.py", line 127, in import_module return _bootstrap._gcd_import(name[level:], package, level) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/init.py", line 88, in from tensorflow.python import keras File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/keras/init.py", line 26, in from tensorflow.python.keras import activations File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/keras/init.py", line 26, in from tensorflow.python.keras import activations File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/keras/activations.py", line 23, in from tensorflow.python.keras.utils.generic_utils import deserialize_keras_object File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/keras/utils/init.py", line 34, in from tensorflow.python.keras.utils.io_utils import HDF5Matrix File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/tensorflow_core/python/keras/utils/io_utils.py", line 30, in import h5py File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "/home/pi/.virtualenvs/cv/lib/python3.7/site-packages/h5py/init_.py", line 25, in from . import _errors File "/usr/lib/python3/dist-packages/thonny/plugins/cpython/cpython_backend.py", line 314, in _custom_import module = self._original_import(args, *kw) File "h5py/_errors.pyx", line 1, in init h5py._errors ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 44 from C header, got 40 from PyObject

You implemented much of your model yourself. I have come across this GItHub issue recently where people reported a large slowdown when they were running their self-built models eagerly. It seems to be somewhat similar to yours. One answer (this one: https://github.com/tensorflow/probability/issues/629#issuecomment-551875968) seems to allow people to achieve speedups.