revert: Allow 0-dimensional shape tensors#972

Merged

matthewfeickert merged 31 commits intomasterfrom revert/allow-non-empty-shape

Jul 25, 2020

+108-61

Commits on Jul 25, 2020

Revert non-empty tensor for NumPy backend

matthewfeickert
authored and
kratsg
committed
Revert non-empty tensor for JAX backend

matthewfeickert
authored and
kratsg
committed
Revert non-empty tensor for PyTorch backend

matthewfeickert
authored and
kratsg
committed
Revert non-empty tensor for TensorFlow backend

matthewfeickert
authored and
kratsg
committed
Add docstring example for astensor

matthewfeickert
authored and
kratsg
committed
Ensure tensor of float has empty shape

matthewfeickert
authored and
kratsg
committed
THIS WILL FAIL BUT SAVE WORK

matthewfeickert
authored and
kratsg
committed
Use tf.broadcast_to

matthewfeickert
authored and
kratsg
committed
Unsqueeze PyTorch 0-d tensors

matthewfeickert
authored and
kratsg
committed
Have test_tensor understand floats

matthewfeickert
authored and
kratsg
committed
qmu is a 0-d tensor

matthewfeickert
authored and
kratsg
committed
Allow 0-d result tensor

matthewfeickert
authored and
kratsg
committed
Use torch.Tensor.view for faster op

matthewfeickert
authored and
kratsg
committed
Use Lukas's pattern as seems nicer

matthewfeickert
authored and
kratsg
committed
q_mu should be a float

matthewfeickert
authored and
kratsg
committed
Ensure float is returned

matthewfeickert
authored and
kratsg
committed
Add atleast_1d method

matthewfeickert
authored and
kratsg
committed
Enforce scalar return for qmu

matthewfeickert
authored and
kratsg
committed
Use list comprehensions

matthewfeickert
authored and
kratsg
committed
Fix typo n docstrings

matthewfeickert
authored and
kratsg
committed
Add clarifyng note

matthewfeickert
authored and
kratsg
committed
Correct docs formatting

matthewfeickert
authored and
kratsg
committed
Return list only if multiple tensors given

matthewfeickert
authored and
kratsg
committed
Show that a single entry gives a tensor back

matthewfeickert
authored and
kratsg
committed
Add stub of tests for test_atelast_1d

matthewfeickert
authored and
kratsg
committed
Restore harmonization in backends

matthewfeickert
authored and
kratsg
committed
Factor out adding atelast_1d test to another feat PR

matthewfeickert
authored and
kratsg
committed
Fix docstrings for mle

matthewfeickert
authored and
kratsg
committed
Ensure returned objective function is a scalar

matthewfeickert
authored and
kratsg
committed
returned objective is still tensor, just 0-d so repr's as tensor

matthewfeickert
authored and
kratsg
committed
Add mle.fit tests for shape of returns

matthewfeickert
authored and
kratsg
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

revert: Allow 0-dimensional shape tensors#972

Commits on Jul 25, 2020

Revert non-empty tensor for NumPy backend

Revert non-empty tensor for JAX backend

Revert non-empty tensor for PyTorch backend

Revert non-empty tensor for TensorFlow backend

Add docstring example for astensor

Ensure tensor of float has empty shape

THIS WILL FAIL BUT SAVE WORK

Use tf.broadcast_to

Unsqueeze PyTorch 0-d tensors

Have test_tensor understand floats

qmu is a 0-d tensor

Allow 0-d result tensor

Use torch.Tensor.view for faster op

Use Lukas's pattern as seems nicer

q_mu should be a float

Ensure float is returned

Add atleast_1d method

Enforce scalar return for qmu

Use list comprehensions

Fix typo n docstrings

Add clarifyng note

Correct docs formatting

Return list only if multiple tensors given

Show that a single entry gives a tensor back

Add stub of tests for test_atelast_1d

Restore harmonization in backends

Factor out adding atelast_1d test to another feat PR

Fix docstrings for mle

Ensure returned objective function is a scalar

returned objective is still tensor, just 0-d so repr's as tensor

Add mle.fit tests for shape of returns