Summary:
Fixes https://github.com/pytorch/pytorch/issues/29161.
I looked a bit at the code changes related to this and think I have all of the use cases of `DeprecatedTypeProperties` covered in the message, but suggestions from someone with more context on this would be very much appreciated :)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30281
Differential Revision: D18830818
Pulled By: ezyang
fbshipit-source-id: 1a7fcee15354ae09e6644577e7fa33bd26acfe20
Summary:
The original design of `torch::nn::utils::clip_grad_norm_` / `clip_grad_value_` takes input by non-const reference, which prevents users from passing rvalue reference as input into the functions. This PR changes the functions to take input by value, which matches the Python version's semantics, and also adheres to the C++ API convention that if a function modifies its input in-place, it should take that input by value.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30216
Differential Revision: D18632543
Pulled By: yf225
fbshipit-source-id: 97a09d6467f982fe9c8120f483a9c07fcf13699e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30146
This PR fixes naming for kl_div and binary_cross_entropy functional options, to be more consistent with the naming scheme of other functional options.
Test Plan: Imported from OSS
Differential Revision: D18618971
Pulled By: yf225
fbshipit-source-id: 2af62c1a0ace2cd0c36c2f1071639bf131d8fe61
Summary:
Hi yf225,
I have a few doubts related to implementation:
1) What tests do I have to write?
2) What does _load_state_from_dict does?
3) Do I need to override reset() function as I can not see it's utility?
4) InstanceNormOptions could be removed with BatchNormOptions, but I find that
`track_running_status` is not defined instead `stateful` is defined.
InstanceNorm{1,2,3}d https://github.com/pytorch/pytorch/issues/25883
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28790
Differential Revision: D18588666
Pulled By: yf225
fbshipit-source-id: bb9b81f01f62c3fc8765fa0ba0716768087ee155
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30112
Currently, we have torch::nn functionals that takes `input` as `Tensor&` in order to be able to in-place change `input`'s value. We likely shouldn't do this because it will prevent the following use case:
```cpp
F::elu(torch::tensor(1), F::ELUFuncOptions().inplace(true))
```
The solution is to change the type of `input` to `Tensor`, so that we can pass an rvalue into the functional.
Test Plan: Imported from OSS
Differential Revision: D18601580
Pulled By: yf225
fbshipit-source-id: 639a86eb62f6c986b0f20bf7e201983e83126e73
Summary:
Hi yf225 , I have added **NLLLoss and CrossEntropyLoss.**
```
Also, while using log_softmax in cross_entropy_loss, I am getting an error
../caffe2/../torch/csrc/api/include/torch/nn/functional/loss.h:537:63: error: no matching function for call to log_softmax(const at::Tensor&)’
const Tensor& log_softmax_input = torch::log_softmax(input);
aten/src/ATen/Functions.h:5551:22: note: candidate: at::Tensor at::log_softmax(const at::Tensor&, int64_t, c10::optional<c10::ScalarType>)
static inline Tensor log_softmax(const Tensor & self, int64_t dim, c10::optional<ScalarType> dtype) {
^~~~~~~~~~~
aten/src/ATen/Functions.h:5551:22: note: candidate expects 3 arguments, 1 provided
```
I think the other two parameters should be optional as in python frontend(shown in documentation here at https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.log_softmax ). Rest, there were no errors in build and tests have passed
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29812
Differential Revision: D18548249
Pulled By: yf225
fbshipit-source-id: 2ab350abd2a6f498d4dba2345f51ad87471f3038
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29653
I didn't remove is_variable from Tensor for BC reasons, but I did
remove as many uses as I could from the codebase.
at::impl::variable_excluded_from_dispatch got moved to TensorBody.h
so that it's more widely accessible.
This diff is NOT semantics preserving. Here are the major differences:
- In a number of native operator implementations, we tested that arguments
are not variable. I replaced these with asserts that variable is
excluded from dispatch. I actually don't think these asserts are really
necessary now (they should certainly be true, but it's hard to get
it wrong), but I've kept them for old time's sake. At least, they'll detect
if you call these functions before you've processed variable (indicating
a bug in your kernel.)
- There are a number of places where we do a per-tensor test for being a
variable, for better error reporting when someone commits Tensor/Variable
confusion. Although these tests are substantively the same as the
tests above, in these cases I decided to *delete* the test entirely.
The reasoning is that in these cases, we didn't really care about
dispatch (also, see above; I'm not too sure we really need the dispatch
asserts), we cared about Tensor/Variable confusion. Since Tensor/Variable
confusion is impossible now, we don't need the tests. One of the key
factors which pushed me one way or another was whether or not a function
was doing per-tensor validation; if I kept the assert in such functions,
I'd repeatedly access the TLS. Even if we want to bring back the asserts,
they would have to go somewhere else.
Another similar idiom is the number of places we do !x.defined() ||
x.is_variable(); I treated this equivalently.
- nuclear_norm's computation of compute_uv is a bit weird, but I think
it's OK to just delete the is_variable case (I *suspect* that it is
always the case that self.is_variable(), but it doesn't really matter.)
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Differential Revision: D18496168
Pulled By: ezyang
fbshipit-source-id: 5a1ded931e0c10a6b758ba64a8380d34110e0c3e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29673
Following https://github.com/pytorch/pytorch/pull/29364 and https://github.com/pytorch/pytorch/pull/29404, this PR makes `F::EmbeddingFuncOptions` and `F::EmbeddingBagFuncOptions` separate classes from `torch::nn::EmbeddingOptions` and `torch::nn::EmbeddingBagOptions`, so that it's easier to enforce that arguments such as `num_embeddings` and `embedding_dim` are required for `torch::nn::EmbeddingOptions` and `torch::nn::EmbeddingBagOptions`.
Test Plan: Imported from OSS
Differential Revision: D18462540
Pulled By: yf225
fbshipit-source-id: f2abf431e48675b0a9d7f6f398cdb90ff9037c35
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29632
This PR is BC-breaking in the following way:
Previously, C++ `torch::tensor` with a floating-point literal with no suffix (e.g. `torch::tensor(1.1)`) or a (nested) braced-init-list of
floating-point literals with no suffix (e.g. `torch::tensor({{1.1, 2.2}})` produces a tensor with dtype `at::kDouble`. After this PR, it produces a tensor with dtype `torch::get_default_dtype()`, matching Python `torch.tensor` behavior.
Test Plan: Imported from OSS
Differential Revision: D18465819
Pulled By: yf225
fbshipit-source-id: 6834fe50335c677bc3832f2a5e9cf8d1ede9f665
Summary:
This PR changes the implementation of C++ Conv{1,2,3}d layers to exactly match the Python version, and add F::conv{1,2,3}d functionals. For more thorough testing, I will rely on the parity test mechanism which uses values from `common_nn.py` to generate the inputs and options that we are interested in testing.
This PR is BC-breaking in the following way:
In `Conv{1,2,3}dOptions`:
- `with_bias` is renamed to `bias`.
- `input_channels` is renamed to `in_channels`.
- `output_channels` is renamed to `out_channels`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28917
Differential Revision: D18471526
Pulled By: yf225
fbshipit-source-id: 7a33f60654ad93cc2e043245e7ff9e0ef9da15b3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29203
There is no more Variable/Tensor distinction, so fix the misleading name.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Differential Revision: D18353505
Pulled By: ezyang
fbshipit-source-id: dadc394d533ab7746f70bc186c6645441a784518
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29404
This PR makes all non-input arguments to functionals part of its options parameters, so that we won't break backward compatibility even if we add or reorder some of the non-input arguments to functionals in the future.
Test Plan: Imported from OSS
Differential Revision: D18378526
Pulled By: yf225
fbshipit-source-id: f5cf6bdfb844e75bf94fdee58c121e0955631b6e
Summary:
Fixes https://github.com/pytorch/pytorch/issues/17662
I'm not sure if `arange` needs to be in python_arg_parser at all, given the schemas in native_functions.yaml. In any case this at least fixes the dytpe mismatch.
In follow up PRs I will try to handle some of the other ops that do type inference at the python level, like randint.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/27629
Differential Revision: D17885939
Pulled By: eellison
fbshipit-source-id: f97a8bc722b7ab77de1c42a992e49a4a3175ad60
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29364
Currently, we use `torch::nn::*Options` both as module options and functional options. However, this makes it very hard to manage the parameters in `torch::nn::*Options`, because a module's constructor can take a different set of arguments than the module's equivalent functional (e.g. `torch.nn.BatchNorm1d` takes `num_features, eps=1e-5, momentum=0.1, affine=True,
track_running_stats=True`, while `F::batch_norm` takes `running_mean, running_var, weight=None, bias=None, training=False, momentum=0.1, eps=1e-5`).
This PR resolves the above problem by making `F::*FuncOptions` a different class from `torch::nn::*Options` when necessary (i.e. when a module's constructor takes a different set of arguments than the module's equivalent functional). In the rest of the cases where the module constructor takes the same set of arguments as the module's equivalent functional, `F::*FuncOptions` is an alias of `torch::nn::*Options`.
Also as part of this PR, we change all functional options to pass-by-value, to make the semantics consistent across all functionals.
Test Plan: Imported from OSS
Differential Revision: D18376977
Pulled By: yf225
fbshipit-source-id: 8d9c240d93bfd5af0165b6884fdc912476b1d06b
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28828
This updates torch::script::Module to more closely match the behavior
of nn.Module. In particular, it implements the (optionally recurisive)
iterators that retrieve submodules, parameters, and buffers and makes
their names match the python versions.
This also removes the individual accessors for Parameter, Module, Buffer, etc.
and replaces them with a single `attr` function which is equivalent to
writing `a.foo` in Python (`setattr` emulates `a.foo = v`).
As we build out the user-facing API for TorchScript values this will end
up matching how an attribute is accessed on general objects.
This PR preservers the python bindings for script::Module by emulating the
old API at the binding level. A followup will clean up the usage to more
directly match the C++ API.
Test Plan: Imported from OSS
Differential Revision: D18197611
Pulled By: zdevito
fbshipit-source-id: 7ee4dcbb258605d1c988314b05d938423f1ccee5
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/29066
This PR is BC-breaking in the following way:
Previously, C++ `torch::tensor` with an integer literal or a braced-init-list of
integer literals produces a tensor with dtype being the type of the integer literal(s). After this PR, it always produces a tensor of dtype `at::kLong` (aka. int64_t), matching Python `torch.tensor` behavior.
Test Plan: Imported from OSS
Differential Revision: D18307248
Pulled By: yf225
fbshipit-source-id: 7a8a2eefa113cbb238f23264843bdb3b77fec668
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28620
All Tensors are Variables now, they just happen to have requires_grad=False. Tensors ALWAYS have `VariableTensorId` in their type set.
When constructing this patch, I had to make decisions about what I would fix in this patch, and what I would leave for follow up PRs. Here is the cleanup that happens in this patch:
- The `is_variable` property is removed from TensorOptions. I removed this immediately because unlike Tensor::is_variable, TensorOptions::is_variable doesn't respect our VariableTensorId thread-local state. This means that there were a bunch of places where TensorOptions::is_variable was false, which is obviously bogus in the world when tensor and variable are merged. Instead of keeping the method as a function that always returns true, I just opted to remove it entirely (it's not public API.) All places we set `is_variable` are deleted.
- Knock on effect: there is no longer a separate DeprecatedTypeProperties for the variable and non-variable versions of type.
- Knock on effect: instead of asserting on TensorOptions::is_variable, instead we just test `at::impl::variable_is_excluded()`
- There is now only one copy of the cuDNN RNN dropout cache, not two (I'm not sure why we had two to begin with)
Some cleanup that doesn't happen in this patch:
- Eliminating unnecessary uses of `make_variable`
- Eliminating `Tensor::is_variable`
The most subtle part of this patch is retaining tracing behavior: the fact that everything is a Variable means that more code gets routed to VariableType than before; this can change traces. I identified two places where we didn't appropriately turn off VariableType, mostly factory functions:
- `torch.tensor` must turn off VariableType before invoking `at::empty` to construct the tensor, as it subsequently does direct data access
- `tensor_slow` (invoked when you pass a Python scalar to a tensor argument) must turn off VariableType before calling `scalar_to_tensor` so the scalar gets traced as constant, rather than as a call to `scalar_to_tensor`.
Honestly, these are all giant hacks, and should be replaced with a more specialized guard that just toggles tracing.
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Test Plan: Imported from OSS
Reviewed By: dreiss
Differential Revision: D18171156
Pulled By: ezyang
fbshipit-source-id: 5b6a045beba37492647e350190f495114e86504d
Summary:
Adds C++ API clip_grad_value_ for torch::nn:utils module.
Also, fix the for indent level error in the original test/test_nn.py.
Issue: https://github.com/pytorch/pytorch/issues/25883
Reviewer: yf225
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28736
Differential Revision: D18263807
Pulled By: yf225
fbshipit-source-id: 29282450bd2099df16925e1d0edd3d933f6eeb9b
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/28523
New features:
1. Previously, `torch::tensor({true, false, true})` throws `"tensor_cpu" not implemented for 'Bool'`. After this PR, it produces the correct bool tensor, matching the Python API behavior.
2. Tensors with zero-size dimensions are now supported, e.g. `torch::tensor({{}, {}})` produces a tensor with sizes `{2, 0}`, matching the Python API behavior.
BC-breaking bug fixes:
1. Previously, `torch::tensor({{1}, {2}})` produces a tensor of sizes `{2}`. After this PR, it produces a tensor of sizes `{2, 1}`, matching the Python API behavior.
2. Fixed semantics of `torch::tensor(1.1)`: it now returns a 0-dim tensor instead of a 1-dim tensor, matching the Python API behavior.
3. Previously, when passed a non-dtype `TensorOptions` to the `torch::tensor` constructor, it always produces a tensor of dtype `float`. After this PR, it produces tensor of different dtypes based on the dtype of the braced-init-list, matching the behavior of the no-options case.
```cpp
// Previously:
torch::tensor({1, 2, 3}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> float
torch::tensor({{1, 2, 3}}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> float
torch::tensor({1., 2., 3.}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> float
torch::tensor({{1., 2., 3.}}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> float
// Now:
torch::tensor({1, 2, 3}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> int
torch::tensor({{1, 2, 3}}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> int
torch::tensor({1., 2., 3.}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> double
torch::tensor({{1., 2., 3.}}, torch::TensorOptions(/*non-dtype-options*/)).dtype() -> double
// As comparison, currently:
torch::tensor({1, 2, 3}).dtype() -> int
torch::tensor({{1, 2, 3}}).dtype() -> int
torch::tensor({1., 2., 3.}).dtype() -> double
torch::tensor({{1., 2., 3.}}).dtype() -> double
```
Notes:
1. From now on, the behavior of `at::tensor(scalar_value)` (which produces a 1-dim tensor) would be different from `torch::tensor(scalar_value)` (which produces a 0-dim tensor). I will fix the behavior of `at::tensor(scalar_value)` in a follow-up PR.
2. From now on, the behavior of `at::tensor({1, 2, 3}, torch::TensorOptions(/*non-dtype-options*/))` (which produces a `float` tensor) would be different from `torch::tensor({1, 2, 3}, torch::TensorOptions(/*non-dtype-options*/))` (which produces a an `int` tensor). I will fix this behavior of `at::tensor` constructor in a follow-up PR.
Context for the changes in this PR:
The motivation comes from fixing the "`torch::tensor({{1}, {2}})` gives tensor of wrong sizes" bug - in order to fix it, I have to move the handling of `at::ArrayRef` and `std::vector` into `InitListTensor` (see below on why we need to do this) and renamed `InitListTensor` to `TensorDataContainer`. After such changes, support for bool values comes out of the box without extra effort, and support for tensors with zero-size dimensions only requires adding a default constructor for `TensorDataContainer`, so I added those two in this PR.
For the semantic change of `torch::tensor(1.1)`, it's actually more effort to preserve the original wrong behavior (i.e. we need to check the sizes of the tensor converted from `TensorDataContainer` and reshape any scalar tensor to a 1-D tensor). I think preserving the original wrong behavior doesn't give us much value, and since the above changes naturally fix the problem, we should just start using the right behavior instead.
For the "constructor with non-dtype options behavior" fix, the code looks simpler and easier to reason about with the fix, so I included it in this PR.
--------
Why we need to move the handling of `at::ArrayRef` and `std::vector` into `TensorDataContainer`:
`torch::tensor({{1}, {2}})` can match this function overload:
`torch::tensor(at::ArrayRef<int> values)`, because `{1}` and `{2}` can be treated as
a list-initialization of an `int` value. However, this will produce a Tensor with sizes `{2}`,
but we actually want a Tensor with sizes `{2, 1}`. In order to avoid matching this function overload,
we removed the function overload and moved the ability to convert `at::ArrayRef<T>`
(and similarly `std::vector<T>`) into `TensorDataContainer`, and since for braced-init-list the
`TensorDataContainer(std::initializer_list<TensorDataContainer>)` constructor is always preferred over all other constructors, it will take the `std::initializer_list` path, and all is good.
Test Plan: Imported from OSS
Differential Revision: D18234625
Pulled By: yf225
fbshipit-source-id: 0f3f6912e82e2117d2103e31b74e7e97baaa8693
