Torch Helpers, a set of utilities for PyTorch Machine Learning Workflow
What does this do? It does:
-
Automatically setting the
cudaflag so that you don't have to do that manually.import torch_helpers as h h.varify(torch.randn(5, 2)) # <== Automatically setting cuda depending on if torch.cuda.is_availabe() # And you can override this! a = h.varify(torch.randn(5, 2), cuda=False) b = h.varify(torch.randn(5, 2), cuda=True) # The defaut is "auto" b = h.varify(torch.randn(5, 2), cuda="auto")
Todo
Done
-
h.varify, returns <torch.Variable> requires grad. -
h.const, returns <torch.Variable> withrequires_grad=False -
h.volatile, returns <torch.Variable> withvolatile=True. -
h.one_hot, encodes one_hot vector -
h.sample_probs, applies a mask to a categorical distribution, and passes gradient back through the distribution. -
h.cast, casts variables and tensors to a different type. For exampleint -> float.
Installation
pip install torch_helpersUsage
Here is a snippet from the test:
import torch
import numpy as np
import torch_helpers as h
def test_requires_grad():
assert h.requires_grad(requires_grad=True) is True
assert h.requires_grad(requires_grad=False) is False
assert h.requires_grad(volatile=True) is False
assert h.requires_grad(volatile=False) is False
assert h.requires_grad(requires_grad=True, volatile=True) is False
assert h.requires_grad(requires_grad=False, volatile=True) is False
assert h.requires_grad(requires_grad=False, volatile=False) is False
assert h.requires_grad(requires_grad=True, volatile=False) is True
def test_varify():
# varify takes integer range objects
x = range(0, 3)
v = h.varify(x, 'int') # setting a `Float` tensor results in RunTimeError
# takes float arrays
x = np.arange(0.0, 3.0)
v = h.varify(x)
# takes torch tensors
x = torch.randn(4, 1)
t = h.varify(x)
t = h.varify(x, volatile=True)
t = h.const(x, volatile=True)
assert t.requires_grad is False and t.volatile is True
# You can override the requires_grad flag in constants.
# This is useful when you want to have a constant by default, but
# would like to switch to var when a requires_grad flag is passed in.
t = h.const(x, requires_grad=True)
assert t.requires_grad is True
# h.one_hot gives you one_hot encoding
def test_one_hot():
acts = h.const([1, 2], dtype='int')
n = 3
oh = h.one_hot(acts, n)
h.assert_equal(oh.data, h.tensorify([[0., 1., 0.], [0., 0., 1.]]),
message="one_hot gives incorrect output {}".format(oh))
# For RL tasks, `h.sample_probs` allows you to back-prop through action probability
def test_mask():
probs = h.varify([[0.1, 0.2, 0.7], [0.4, 0.5, 0.1]])
acts = h.const([1, 2], dtype='int')
sampled_probs = h.sample_probs(probs, acts)
sampled_probs.sum().backward()
dp = probs.grad.data.numpy()
assert dp[0, 1] is not None and dp[1, 2] is not None, 'key entries of probs grad should be non-zero'To Develop
git clone https://github.com/episodeyang/torch_helpers.git
cd torch_helpers
make devTo test, run
make testThis make dev command should build the wheel and install it in your current python environment. Take a look at the ./Makefile for details.
To publish, first update the version number, then do:
make publish
