Summary:
jit.ScriptModule deletes all the actual attributes but still uses the nn.Module implementation.
Since I don't know how to add this new set() to the ScriptModule, it is simpler to just raise a nice error for now.
I also reverted the logic so that an empty set() (which is always the case in a ScriptModule) means that everything is persistent.
cc zdevito should we open an issue to add this to the ScriptModule?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/38131
Differential Revision: D21502183
Pulled By: albanD
fbshipit-source-id: 96f83098d9a2a9156e8af5bf5bd3526dd0fefc98
Summary:
Fixes https://github.com/pytorch/pytorch/issues/37182
The `zero_grad` wrapper from `_replicate_for_data_parallel` can't be pickled. So instead, I set an attribute `_is_replica = True` and check for this in `Module.zero_grad`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/37307
Differential Revision: D21246119
Pulled By: mrshenli
fbshipit-source-id: 4755786d48a20bc247570ba672de9dd526914ce1
Summary:
As described in https://github.com/pytorch/pytorch/issues/33934, the current attribute error in `nn.Module`'s properties are wrong.
```python
from torch import nn
class MyModule(nn.Module):
property
def something(self):
hey = self.unknown_function()
return hey
model = MyModule()
print(model.something)
```
This raises `AttributeError: 'MyModule' object has no attribute 'something'` when what we want is `AttributeError: MyModule instance has no attribute 'unknown_function'`.
This fixes this issue and will make properties much easier to debug !
Pull Request resolved: https://github.com/pytorch/pytorch/pull/34324
Differential Revision: D20645563
Pulled By: ezyang
fbshipit-source-id: 130f861851bdbef43803569a5ce9e24d2b942179
Summary:
Fixes https://github.com/pytorch/pytorch/issues/31768, second attempt of https://github.com/pytorch/pytorch/issues/32870
DataParallel creates replicas of the original `nn.Module` with the parameters duplicated onto the destination devices. Calling `backwards` will propagate gradients onto the original module parameters but calling `zero_grad` on the replica module doesn't clear the gradients from the parent module. However, any replica using backwards was broken anyway since the replica's parameters are not leaf nodes in autograd. So, we should issue a warning.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/33064
Differential Revision: D19790178
Pulled By: albanD
fbshipit-source-id: 886f36640acef4834a6fa57a26ce16b42ff0e9ad
Summary:
Fixes https://github.com/pytorch/pytorch/issues/31768
`DataParallel` creates replicas of the original `nn.Module` with the parameters duplicated onto the destination devices. Calling `backwards` will propagate gradients onto the original module parameters but calling `zero_grad` on the replica module doesn't clear the gradients from the parent module,
~breaking any model that uses `backward`-`zero_grad` in its `forward`. I fix this by patching the replica module so that `zero_grad` clears grads on the parent as well.~
However, any replica using backwards was broken anyway since the replica's parameters are not leaf nodes in autograd. So, we should raise a warning.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/32870
Differential Revision: D19730209
Pulled By: ezyang
fbshipit-source-id: cb9b2cb0c2e0aca688ce0ff3e56b40fbd2aa3c66
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/27850
Many of these are real problems in the documentation (i.e., link or
bullet point doesn't display correctly).
Test Plan: - built and viewed the documentation for each change locally.
Differential Revision: D17908123
Pulled By: zou3519
fbshipit-source-id: 65c92a352c89b90fb6b508c388b0874233a3817a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/27399
This was devised in a time when we didn't have module attributes. They
are essentially just tensor lists, so represent them that way. This has
the additional benefit of making the RNN forward pass faster because we
effectively cache the flattened weights.
The only complication part is that someone may come along and do:
```
my_rnn_mod.w_ih_l0 = torch.nn.Parameter(...)
```
This means we need to override setattr to keep the flattened weights
cache up to date.
Test Plan: Imported from OSS
Differential Revision: D17785658
Pulled By: suo
fbshipit-source-id: 7789cd1d0d4922bfd5eba1716976442fbf150766
Summary:
Most of this was old cruft left over from special handling of `training` before we had a `bool` type. This makes all modules have a `training` attribute that is true by default and removes all other special handling.
Fixes#26884
](https://our.intern.facebook.com/intern/diff/17728129/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/27109
Pulled By: driazati
Differential Revision: D17728129
fbshipit-source-id: 8ddc9fbb07a953dd05529538bfdd01ed88b5cb57
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/23779
Mangling is two underscores, not one :(. We want this method to be
private so that inheritors who define a `__construct` do not interfere
with Module initialization
Test Plan: Imported from OSS
Differential Revision: D16645156
Pulled By: suo
fbshipit-source-id: b9060cb35bfaa0391ff200b63fb78b1ac15fee39
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/23680
Now when initializing a ScriptModule during the torch.jit.load()
process, there is already a cpp module backing the thing. That means
that setting training will overwrite whatever the initialized
ScriptModule had.
This PR splits apart the common "set up internal state" part of the
Module __init__ and calls that from ScriptModule.__init__ and
Module.__init__, leaving the "nn.Module-specific" part (setting
`self.training`) for the nn.Module __init__
Test Plan: Imported from OSS
Differential Revision: D16606959
Pulled By: suo
fbshipit-source-id: f7ea6b36551ff4e4472b7685f65731d5cfab87fd
Summary:
We introduced RTTI in recent change: https://github.com/pytorch/pytorch/pull/21613
For internal mobile build we don't enable '-frtti' yet. This diff is trying to replace
RTTI with alternative approach.
According to dzhulgakov we could compare two tensors' type_id directly in most cases -
which is more strict than comparing TensorImpl subclass type as TensorImpl -> type_id
mapping is 1-to-n but it's more proper for this use case.
The only two cases where we can relax direct type comparison (for legacy reason) are:
1. CPUTensor <-> CUDATensor;
2. SparseCPUTensor <-> SparseCUDATensor;
Pull Request resolved: https://github.com/pytorch/pytorch/pull/22773
Differential Revision: D16277696
Pulled By: ljk53
fbshipit-source-id: 043e264fbacc37b7a11af2046983c70ddb62a599
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/22285
Previously forward hooks are expected to return None, this PR adds the support to overwrite input and output in `forward_pre_hook` and `forward_hook`, this is used to implement inserting quant/dequant function calls around forward functions.
Differential Revision: D16022491
fbshipit-source-id: 02340080745f22c8ea8a2f80c2c08e3a88e37253
Summary:
# Motivation
We allow to override JIT module serialization with `__getstate__/__setstate__` in order to cover cases where parameters are not serializable. Use cases include: MKLDNN integration: a388c78350/torch/utils/mkldnn.py (L18-L26)
and also fbgemm prepacked format integration for quantized tensors.
However many Eager scripts use `torch.save(module.state_dict())` form of serialization. There are several ways to make it work:
* make packed_weight itself pickleable (e.g. by binding `__getstate__/__setstate__` on C++ UDT level)
* change: we’d need to allow module buffers to be of arbitrary, non-Tensor types
* pro: no change to state_dict behavior
* cons: might not be directly inspectable by user calling .state_dict(), especially if packed weights represent several tensors fused together
* make packed_weight being proper Tensor layout
* pro: no change to state_dict or buffers behavior
* cons: adding new tensor layouts is pretty costly today
* cons: doesn’t work if multiple tensors are packed in one interleaved representation
* *[this approach]* allow Modules to override state_dict and return regular tensors
* pro: most flexible and hackable
* pro: maintains semantic meaning of statedict as all data necessary to represent module’s state
* cons: complicates state_dict logic
* cons: potential code duplication between `__getstate__/__setstate__`
Based on discussions with zdevito and gchanan we decided to pick latter approach. Rationale: this behavior is fully opt-in and will impact only modules that need it. For those modules the requirement listed above won't be true. But we do preserve requirement that all elements of state_dict are tensors. (https://fburl.com/qgybrug4 for internal discussion)
In the future we might also implement one of the approaches above but those are more involved.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/21933
Differential Revision: D15937678
Pulled By: dzhulgakov
fbshipit-source-id: 3cb5d1a8304d04def7aabc0969d0a2e7be182367
Summary:
https://github.com/pytorch/pytorch/pull/17072 breaks `model.to(xla_device)`, because moving `model` to XLA device involves changing its parameters' TensorImpl type, and the current implementation of `nn.Module.to()` doesn't support changing module parameters' TensorImpl type:
```python
# 6dc445e1a8/torch/nn/modules/module.py (L192-L208)
def _apply(self, fn):
...
for param in self._parameters.values():
if param is not None:
# Tensors stored in modules are graph leaves, and we don't
# want to create copy nodes, so we have to unpack the data.
param.data = fn(param.data) # NOTE: this doesn't allow changing `param.data`'s TensorImpl type
if param._grad is not None:
param._grad.data = fn(param._grad.data) # NOTE: this doesn't allow changing `param._grad.data`'s TensorImpl type
...
```
yf225 TODO: fix the description here when we finish the implementation
To fix this problem, we introduce a new API `model.to_()` that always assign new tensors to the parameters (thus supporting changing the parameters to any TensorImpl type), and also bump the version counter of the original parameters correctly so that they are invalidated in any autograd graph they participate in.
We also add warning to the current `model.to()` API to inform users about the upcoming behavior change of `model.to()`: in future releases, it would create and return a new model instead of in-place updating the current model.
This unblocks adding XLA to our CI test suite, which also allows XLA to catch up with other changes in our codebase, notably the c10 dispatcher.
[xla ci]
cc. resistor ailzhang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/21613
Differential Revision: D15895387
Pulled By: yf225
fbshipit-source-id: b79f230fb06019122a37fdf0711bf2130a016fe6
Summary:
When we pass `fn` to `nn.Module._apply()` and `fn` is an in-place operation, the correct behavior should also include bumping the parameters' and their gradients' version counters. This PR fixes the old incorrect behavior and makes sure the new behavior is right.
Note that this PR is BC-breaking in the following way:
Previously, passing an in-place operation to `nn.Module._apply()` does not bump the module's parameters' and their gradients' version counters. After this PR, the module's parameters' and their gradients' version counters will be correctly bumped by the in-place operation, which will invalidate them in any autograd graph they previously participate in.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/21865
Differential Revision: D15881952
Pulled By: yf225
fbshipit-source-id: 62f9244a4283a110147e9f20145ff232a5579fbd
Summary:
Resubmit #20698 which got messed up.
Idea is that when PyTorch is used in a custom build environment (e.g. Facebook), it's useful to track usage of various APIs centrally. This PR introduces a simple very lightweight mechanism to do so - only first invocation of a trigger point would be logged. This is significantly more lightweight than #18235 and thus we can allow to put logging in e.g. TensorImpl.
Also adds an initial list of trigger points. Trigger points are added in such a way that no static initialization triggers them, i.e. just linking with libtorch.so will not cause any logging. Further suggestions of what to log are welcomed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20745
Differential Revision: D15429196
Pulled By: dzhulgakov
fbshipit-source-id: a5e41a709a65b7ebccc6b95f93854e583cf20aca
Summary:
load_state_dict includes a recursive inner function `load` that captures
Tensors through the close-over variable `state_dict`. Because it's
recursive, it also captures itself leading to a reference cycle.
This breaks the reference cycle so that any Tensors in state_dict can be
collected immediately instead of waiting until the next GC cycle.
Alternatively, we could have passed `state_dict` and `metadata` as
arguments to load to prevent capture of Tensors. (That would still
result in cyclic garbage, but not any cyclic garbage of Tensors).
See:
https://github.com/pytorch/pytorch/issues/20199#issuecomment-491089004
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20397
Differential Revision: D15414834
Pulled By: colesbury
fbshipit-source-id: 4c2275a08b2d8043deb3779db28be03bda15872d
Summary:
Added the ">>>" python interpreter sign(three greater than symbols), so that the edited lines will appear as code, not comments/output, in the documentation. Normally, the interpreter would display "..." when expecting a block, but I'm not sure how this would work on the pytorch docs website. It seems that in other code examples the ">>>" sign is used as well, therefore I used with too.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/19347
Differential Revision: D14986154
Pulled By: soumith
fbshipit-source-id: 8f4d07d71ff7777b46c459837f350eb0a1f17e84
Summary:
return missing_keys and unexpected_keys from load_state_dict so the user can handle them when strict mode is off; also removed an unused variable
Pull Request resolved: https://github.com/pytorch/pytorch/pull/18668
Differential Revision: D14782073
Pulled By: ezyang
fbshipit-source-id: ab3b855eb77bb7422594d971988067e86eef20f2
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/17618
Base on the code, we only add key to `missing_keys` and `unexpected_keys` if `$strict` is `True`. The documentation is confusing.
This diff also fix one FLAKE8 warning.
Reviewed By: ailzhang
Differential Revision: D14280593
fbshipit-source-id: d368f5596bdf74ff62ee4d28d79120f5af91e0a3
Summary:
Previously this would fail with the error message:
```
ValueError: Auto nesting doesn't know how to process an input object of type dict. Accepted types: Tensors, or lists/tuples of them
```
Turns out we're not using the line that causes this error (or a side effect of that line), so removing it fixes the issue. Also cleaned up some related dead code (cc apaszke to make sure the code isn't useful in some way)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16616
Differential Revision: D13908352
Pulled By: suo
fbshipit-source-id: 27094f1f4ea0af215b901f7ed3520e94fbc587b3
Summary:
without this "if", code below will throw error " Linear' object has no attribute '_buffers' "
And with this if, error would be "cannot assign buffer before Module.\_\_init\_\_() call", which I think it's more accurate, just like register_parameter.
```
import math
import torch
from torch.nn.parameter import Parameter
from torch.nn import functional as F
from torch.nn import Module
class Linear(Module):
def __init__(self, in_features, out_features, bias=True):
self.in_features = in_features
self.out_features = out_features
self.register_buffer('test', torch.Tensor(out_features, in_features))
self.weight = Parameter(torch.Tensor(out_features, in_features))
if bias:
self.bias = Parameter(torch.Tensor(out_features))
else:
self.register_parameter('bias', None)
super(Linear, self).__init__()
self.reset_parameters()
def reset_parameters(self):
stdv = 1. / math.sqrt(self.weight.size(1))
self.weight.data.uniform_(-stdv, stdv)
if self.bias is not None:
self.bias.data.uniform_(-stdv, stdv)
def forward(self, input):
return F.linear(input, self.weight, self.bias)
def extra_repr(self):
return 'in_features={}, out_features={}, bias={}'.format(
self.in_features, self.out_features, self.bias is not None
)
linear = Linear(3,4)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/16110
Differential Revision: D13715839
Pulled By: soumith
fbshipit-source-id: c300eff0a8655aade448354cf489a592f7db722a
Summary:
Fixes an issue that arose from https://github.com/pytorch/pytorch/pull/13481 where `.shared_memory()` couldn't be called. Effectively undoes all changes to `nn.Module` from that PR and solve the relevant problem in a different way (the goal was to be able to call `._apply()` on the Python wrapper for a C++ module).
soumith
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15305
Differential Revision: D13493937
Pulled By: goldsborough
fbshipit-source-id: 4cb8687f90fc8709a536c5e7eacd0dc8edf6f750
Summary:
This PR enables C++ frontend modules to be bound into Python and added as submodules of Python modules. For this, I added lots of pybind11 bindings for the `torch::nn::Module` class, and modified the `torch.nn.Module` class in Python to have a new Metaclass that makes `isinstance(m, torch.nn.Module)` return true when `m` is a C++ frontend module. The methods and fields of C++ modules are bound in such a way that they work seamlessly as submodules of Python modules for most operations (one exception I know of: calling `.to()` ends up calling `.apply()` on each submodule with a Python lambda, which cannot be used in C++ -- this may require small changes on Python side).
I've added quite a bunch of tests to verify the bindings and equality with Python. I think I should also try out adding a C++ module as part of some large PyTorch module, like a WLM or something, and see if everything works smoothly.
The next step for inter-op across our system is ScriptModule <-> C++ Frontend Module inter-op. I think this will then also allow using C++ frontend modules from TorchScript.
apaszke zdevito
CC dzhulgakov
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13481
Differential Revision: D12981996
Pulled By: goldsborough
fbshipit-source-id: 147370d3596ebb0e94c82cec92993a148fee50a7
Summary:
Add to the Tensor doc info about `.device`, `.is_cuda`, `.requires_grad`, `.is_leaf` and `.grad`.
Update the `register_backward_hook` doc with a warning stating that it does not work in all cases.
Add support in the `_add_docstr` function to add docstring to attributes.
There is an explicit cast here but I am not sure how to handle it properly. The thing is that the doc field for getsetdescr is written as being a const char * (as all other doc fields in descriptors objects) in cpython online documentation. But in the code, it is the only one that is not const.
I assumed here that it is a bug in the code because it does not follow the doc and the convention of the others descriptors and so I cast out the const.
EDIT: the online doc I was looking at is for 3.7 and in that version both the code and the doc are const. For older versions, both are non const.
Please let me know if this should not be done. And if it should be done if there is a cleaner way to do it !
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14339
Differential Revision: D13243266
Pulled By: ezyang
fbshipit-source-id: 75b7838f7cd6c8dc72b0c61950e7a971baefaeeb
Summary:
Problems with SN and DP after #12671 :
1. in eval mode, `weight_orig` is not getting correct gradient #12737 .
Fix: keep `v` vector around as a buffer and always calculate `W = W_orig / (u @ W_orig @ v)` even in eval.
2. in training mode, the `weight` buffer of the parallelized module is never updated, if someone touches `weight_orig` and/or `weight` and makes them not sharing storage. So in `eval` the weight used is wrong.
Fix: Make `weight` not a buffer anymore and always calculate it as above.
3. #12671 changed SN to update `u` in-place to make DP work correctly, but then it breaks backward through two forwards (e.g., the common GAN loss `D(real) - D(fake)`) because the vectors needed to backprop the 1st forward is changed in the 2nd forward.
Fix: This PR clones `u` and `v` before using them.
To maintain BC, I added a hook interface for producing and loading state_dict. This is ugly and we should really have better interface for spectral_norm. But for the purpose to fix this issue, I make this patch. Even if we have a better interface, BC mechanism for legacy loading legacy state_dict still needs to be done.
cc The controller you requested could not be found. crcrpar
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13350
Differential Revision: D12931044
Pulled By: SsnL
fbshipit-source-id: 8be6f934eaa62414d76d2c644dedd7e1b7eb31ef
Summary:
I spent a couple of minutes trying to understand which shape corresponds to checkpoint and which one to the model
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12870
Differential Revision: D10466600
Pulled By: SsnL
fbshipit-source-id: 3b68530b1b756462a2acd59e3a033ff633567a6b
Summary:
In the state dict loading code, it would print the error message referring to the shape of the loaded parameters and the parameters in the initialised model with the formatting in the wrong order. Swapped them round to fix.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/11200
Differential Revision: D9631160
Pulled By: SsnL
fbshipit-source-id: 03d9446303bd417fef67027b10d7a27de06486be
Summary:
This commit adds the ``buffers()`` and ``named_buffers()`` methods as
analogues of ``parameters()`` and ``named_parameters()``.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/10554
Reviewed By: SsnL
Differential Revision: D9367762
Pulled By: jma127
fbshipit-source-id: f2042e46a7e833dce40cb41681dbd80d7885c74e
