Summary:
Fixes [issue#67](https://github.com/MLH-Fellowship/pyre-check/issues/67)
This PR fixes the type checking errors in Pytorch torch/fx/node.py .
The variable types in 363:20 and 364:20 were declared to have type `List[str]` but were assigned a value of `None`. This caused an incompatitble variable type error. I changed the type from `List[str]` to `Optional[List[str]` . This therefore fixed the incompatitble variable type error.
Signed-off-by: Onyemowo Agbo
onionymous
0xedward
Pull Request resolved: https://github.com/pytorch/pytorch/pull/68124
Reviewed By: gmagogsfm
Differential Revision: D32322414
Pulled By: onionymous
fbshipit-source-id: be11bbbd463715ddf28a5ba78fb4adbf62878c80
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69032
I am removing it because, for packaging-related reasons, it's easier if
torch.fx is a pure Python module.
I don't think there is much reason to keep it: this functionality was
experimental, has no known users currently, and we didn't have a clear
path to turning it on by default due to regressions in tracing
performance. Also, it only was ever enabled for `rand` and friends.
Technically the removal of the `enable_cpatching` arguments on
`symbolic_trace` and `Tracer.__init__` are BC-breaking, but the
docstrings clearly state that the argument is experimental and BC is not
guaranteed, so I think it's fine.
Test Plan: Imported from OSS
Reviewed By: soulitzer
Differential Revision: D32706344
Pulled By: suo
fbshipit-source-id: 501648b5c3610ae71829b5e7db74e3b8c9e1a480
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67139
This diff enables setting breakpoint in the graph module's generated python code. See test plan for usage.
In order to support this functionality, and other similar functionalities to customize the generated code, a code transformer functionality is added to `fx.Graph`. This allows flexible customization of `fx.Graph`'s code gen behavior, in composable and functional ways. See test plan for its usage.
Test Plan:
### Use of `fx.experimental.debug.set_trace`
```
In [2]: from torch.fx.experimental.debug import set_trace
In [3]: set_trace(ttop)
Out[3]:
top(
(a): Sub()
)
In [4]: ttop(1)
> /data/users/kefeilu/fbsource33/fbcode/buck-out/dev/gen/caffe2/torch/fb/fx2trt/<eval_with_key>.10(6)forward()
(Pdb) l
1
2
3
4 def forward(self, x):
5 import pdb; pdb.set_trace()
6 -> a = self.a(x); x = None
7 getitem = a[0]
8 getitem_1 = a[0]; a = None
9 add = getitem + getitem_1; getitem = getitem_1 = None
10 return add
11
(Pdb)
```
### Use of `on_generate_code`
```
In [1]: def insert_pdb(body):
...: return ['import pdb; pdb.set_trace()\n', *body]
...:
In [8]: type(ttop)
Out[8]: torch.fx.graph_module.GraphModule.__new__.<locals>.GraphModuleImpl
In [10]: with ttop.graph.on_generate_code(lambda _: insert_pdb):
...: ttop.recompile()
...: print(f"== _on_generate_code should not be None: { ttop.graph._on_generate_code }")
...: print(ttop.code)
...:
== _on_generate_code should not be None: <function insert_pdb at 0x7fc9895ddd30>
def forward(self, x):
import pdb; pdb.set_trace()
a = self.a(x); x = None
getitem = a[0]
getitem_1 = a[0]; a = None
add = getitem + getitem_1; getitem = getitem_1 = None
return add
In [11]: ttop.graph._on_generate_code # restored to None
In [12]: ttop(1) # this should drop into pdb
> /data/users/kefeilu/fbsource33/fbcode/buck-out/dev/gen/caffe2/torch/fb/fx2trt/<eval_with_key>.6(6)forward()
(Pdb) l
1
2
3
4 def forward(self, x):
5 import pdb; pdb.set_trace()
6 -> a = self.a(x); x = None
7 getitem = a[0]
8 getitem_1 = a[0]; a = None
9 add = getitem + getitem_1; getitem = getitem_1 = None
10 return add
11
```
Reviewed By: jamesr66a
Differential Revision: D30736160
fbshipit-source-id: 9646867aae0461b5131dfd4ba9ee77a8c2ea9c93
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64787
This PR added support for lowering per channel quantization and dequantization operators
in fx2trt, this also extends TensorMeta with extra arguments corresponding to per channel quantized Tensors,
initially I was thinking of adding a qpram that can capture everything, but currently we still have some lowering support
for fbgemm ops (which has scale and zero_point in operator interface). I think we can move everything to qprams
after we deprecate lowering support for fbgemm ops in the future.
Test Plan:
Test for per channel weight:
```
python torch/fx/experimental/fx2trt/example/quantized_resnet_test.py
```
change BC compatibility test expect for TensorMeta
```
python test/test_fx.py TestFXAPIBackwardCompatibility.test_class_member_back_compat --accept
```
Imported from OSS
Reviewed By: jfix71, mrshenli, 842974287
Differential Revision: D30879848
fbshipit-source-id: 76c3804bb1d9343183ae53d9f02c1a3bf6c79e1c
Summary:
This PR is created to replace https://github.com/pytorch/pytorch/pull/53180 PR stack, which has all the review discussions. Reason for needing a replacement is due to a messy Sandcastle issue.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64234
Reviewed By: gmagogsfm
Differential Revision: D30656444
Pulled By: ansley
fbshipit-source-id: 77536c8bcc88162e2c72636026ca3c16891d669a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/64261
Note that this does not preserve byte-for-byte compatibility with
existing names.
Test Plan:
* Rely on CI to catch gross errors.
* Merge after release cut to catch subtle issues.
Reviewed By: albanD
Differential Revision: D30700647
Pulled By: dagitses
fbshipit-source-id: 7b02f34b8fae3041240cc78fbc6bcae498c3acd4
Summary:
Previously `torch.jit.trace` relies on AutoGrad hooks to infer name of tensors in computation, including those of function/method arguments. This often doesn't work out because:
- These names often do not exist
- Tracer uses argument name of first tensor operation on each tensor as inferred argument names. These tensor operations have programmatically-generated names like `argument_1`
This PR extracts argument names directly from Python functions and pass them down to tracer, which then assigns them to correct graph inputs. This way, we always have the correct argument names captured in IR.
This is useful for both debugging and supporting using `InterfaceType` to represent traced modules.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/51775
Reviewed By: izdeby
Differential Revision: D26273105
Pulled By: gmagogsfm
fbshipit-source-id: 934a385041137dc3731bb6fa8657b11532fed9e5
Summary:
Fixes https://github.com/pytorch/pytorch/issues/598
This is BC-breaking as we now explicitly don't call the hook when there are not Tensors at the top level of the output.
This feature was not working anyways as the returned grad_input/grad_output were wrong (not respecting the output structure and wrong inputs for multi-Node Module).
This is also BC-breaking as we now report the correct gradients for `nn.Module`s that contain multiple autograd `Node`s while we use to return bad results before.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/46163
Reviewed By: ailzhang, mruberry
Differential Revision: D24894180
Pulled By: albanD
fbshipit-source-id: e1b5d193d2818eb2f51e2a2722c7405c8bd13c2b
Summary:
Raise and assert used to have a hard-coded error message "Exception". User provided error message was ignored. This PR adds support to represent user's error message in TorchScript.
This breaks backward compatibility because now we actually need to script the user's error message, which can potentially contain unscriptable expressions. Such programs can break when scripting, but saved models can still continue to work.
Increased an op count in test_mobile_optimizer.py because now we need aten::format to form the actual exception message.
This is built upon an WIP PR: https://github.com/pytorch/pytorch/pull/34112 by driazati
Pull Request resolved: https://github.com/pytorch/pytorch/pull/41907
Reviewed By: ngimel
Differential Revision: D22778301
Pulled By: gmagogsfm
fbshipit-source-id: 2b94f0db4ae9fe70c4cd03f4048e519ea96323ad
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40718
Currently only constant except tensor must be inlined during serialization.
Tensor are stored in the contant table. This patch generalizes this capability
to any IValue. This is particularly useful for non ASCII string literal that
cannot be inlined.
Test Plan: Imported from OSS
Differential Revision: D22298169
Pulled By: bzinodev
fbshipit-source-id: 88cc59af9cc45e426ca8002175593b9e431f4bac
Summary: Add support for populating domain_discrete field in TensorBoard add_hparams API
Test Plan: Unit test test_hparams_domain_discrete
Reviewed By: edward-io
Differential Revision: D22227939
fbshipit-source-id: d2f0cd8e5632cbcc578466ff3cd587ee74f847af
Summary:
https://github.com/pytorch/pytorch/pull/40129 fixed the error responsible for the first revert, but exposed another error in the same test.
This PR is intended as the "master copy" for merge, and it runs on full CI.
Two other PRs (restricted to run on a small subset of CI) supporting debugging DDP failures/hangs with multiple devices per process (`test_c10d.py:DistributedDataParallelTest.test_grad_layout_1devicemodule_2replicaperprocess`).
- https://github.com/pytorch/pytorch/pull/40290 tries the test with purely rowmajor contiguous params on an untouched master. In other words https://github.com/pytorch/pytorch/pull/40290 contains none of this PR's diffs aside from the test itself.
- https://github.com/pytorch/pytorch/pull/40178, for comparison, tries the test with this PR's diffs.
Both fail the same way, indicating failure is unrelated to this PR's other diffs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40358
Differential Revision: D22165785
Pulled By: albanD
fbshipit-source-id: ac7cdd79af5c080ab74341671392dca8e717554e
Summary:
Currently, whether `AccumulateGrad` [steals](67cb018462/torch/csrc/autograd/functions/accumulate_grad.h (L42)) or [clones](67cb018462/torch/csrc/autograd/functions/accumulate_grad.h (L80)) an incoming gradient, the gradient ends up rowmajor contiguous, regardless of its param's layout. If the param's layout is channels last, or otherwise not rowmajor contigous, later kernels that apply gradients to params are forced into an uncoalesced memory access pattern for either the param or the gradient. This may not sound like a big deal but for any binary op on large tensors it's a >3X increase in gmem traffic => 3X slowdown.
The present PR changes `AccumulateGrad` to prefer, where possible, stashing gradients that match their params' layouts (["Gradient Layout Contract"](https://github.com/pytorch/pytorch/pull/34904/files#diff-ef1a56d24f66b280dcdb401502d6a796R29-R38)).
Allowing `AccumulateGrad` to stash non-rowmajor-contiguous grads means DDP allreduces and DP reduces must allow non-rowmajor-contiguous grads. This PR extends DDP and DP to allow gradients with non-rowmajor-contiguous strides as long as their layout is nonoverlapping and dense.
For good measure, I include changes that allow all five nccl primitives (allreduce, reduce, broadcast, allgather, reducescatter) to act on non-rowmajor-contiguous tensors (again as long as each input's layout is nonoverlapping and dense, and as long as all tensors participating in a given collective have the same layout). The primitive comm changes aren't necessary to enable the DDP changes, but I wasn't sure this would end up true until I had written both sets of changes. I think primitive comm enablement is reasonable to keep in the PR, especially since the code for it is simple.
Channels last params will be a major beneficiary of this PR, but I don't see it as channels-last-specific fix. The spirit is layout matching in general:
- Grads should be stashed with memory layouts matching their params.
- Src and dst tensors on opposite ends of collectives should have matching dense layouts.
This PR also updates autograd docs to describe potential BC-breaking changes below.
## BC notes
ngimel albanD gchanan
#### BC-breaking
In the common case where the user lets AccumulateGrad decide grad layouts, strides for grads of dense but non-rowmajor-contiguous params will change. Any user code that was accustomed to `view(-1)`ing these grads will break.
Also, the circumstances under which a grad can be stolen directly from the backward function that created it, as opposed to deep-copied by AccumulateGrad, have changed. In most cases we expect silent performance improvement, because we expect channels-last-aware backward kernels will create channels last gradients for channels last params. Now those can be stolen, whereas before this PR they were cloned and made rowmajor contiguous. IMO this is a mild BC breakage. Param backward hooks still see grads come in with whatever format the backward kernel gave them. The only BC breakage potential I see is if user code relies somehow on a grad in a hook having or not having the same deep memory as the eventual `param.grad`. Any such users hopefully know they're off the edge of the map and understand how to update their expectations.
#### BC escape hatches
At alband's recommendation, this PR's changes to AccumulateGrad do not alter the pre-PR code's decisions about whether grad is accumulated in or out of place. Accumulations of new grads onto an existing `.grad` attribute were (usually) in-place before this PR and remain in-place after this PR, keeping the existing `.grad`'s layout. After this PR, if the user wants to force accumulation into a grad with a particular layout, they can preset `param.grad` to a zeroed tensor with the desired strides or call `grad.contiguous(desired format)`. This likely won't be as performant as letting AccumulateGrad establish grad layouts by cloning or stealing grads with contract-compliant strides, but at least users have a control point.
One limitation (present before this PR and unchanged by this PR): Presetting `param.grad` does not ensure in-place accumulation all the time. For example, if `create_graph=True`, or if incoming `new_grad` is dense and existing `variable_grad` is sparse, accumulation occurs out of place, and the out-of-place result may not match the existing grad's strides.
----------------------------
I also noticed some potential DDP improvements that I considered out of scope but want to mention for visibility:
1. make sure Reducer's ops sync with AccumulateGrad streams
2. ~to reduce CPU overhead and incur fewer kernel launches, lazily create flat `contents` tensors by a single `cat` kernel only when a bucket is full, instead of `copy_`ing grads into `contents` individually as soon as they are received.~ PR includes a [minor change](https://github.com/pytorch/pytorch/pull/34904/files#diff-c269190a925a4b0df49eda8a8f6c5bd3R312-R315) to divide grads while copying them into flat buffers, instead of copying them in, then dividing separately. Without cat+div fusion, div-while-copying is the best we can do.
3. https://github.com/pytorch/pytorch/issues/38942
Pull Request resolved: https://github.com/pytorch/pytorch/pull/34904
Differential Revision: D20496044
Pulled By: albanD
fbshipit-source-id: 248d680f4b1bf77b0a986451844ec6e254469217
Summary:
The root cause of incorrect rendering is that numbers are treated as a string if the data type is not specified. Therefore the data is sort based on the first digit.
closes https://github.com/pytorch/pytorch/issues/29906
cc orionr sanekmelnikov
Pull Request resolved: https://github.com/pytorch/pytorch/pull/31544
Differential Revision: D21105403
Pulled By: natalialunova
fbshipit-source-id: a676ff5ab94c5bdb653615d43219604e54747e56
Summary:
Fixes https://github.com/pytorch/pytorch/issues/34964
Sparse cuda add was implemented by just concatenating the indices and values for the tensor. If called repeatedly in a tight loop this will let `nnz` grow unbounded. In the worst case of `x.add_(x)` it grows exponentially.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/36030
Differential Revision: D20873504
Pulled By: zou3519
fbshipit-source-id: d90ed8dda0c89571fb89e358757b5dde299513df
Summary:
Resubmit of https://github.com/pytorch/pytorch/pull/35424, only this time I run optimizations in the right order so the PR description is actually true.
This speeds up the inlining pass of FairSeq model from 180s -> 13s, and MaskRCNN model from 5s -> 1.5s.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/35562
Differential Revision: D20738922
Pulled By: eellison
fbshipit-source-id: 1439cf9d1f0bc780e2d64a744694f8b3b7ba4b70
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30445
Create distributed and rpc directories under caffe/test for better management
of unit tests.
Differential Revision: D18702786
fbshipit-source-id: e9daeed0cfb846ef68806f6decfcb57c0e0e3606
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30244
This makes several small changes to the tensorboard graph parsing methods to address the recent changes to the PyTorch JIT trace/graph.
- Inline graph to get information for all nodes
- Assign and propagate scope names to GetAttr nodes
- Prune all useless GetAttr nodes (any with a ClassType output type - tensors and primitives are kept)
- Create output nodes so output tensor shape can be examined
Reviewed By: sanekmelnikov
Differential Revision: D18556323
fbshipit-source-id: b73a809bacfa554c3fe9c4ae3563525f57539874
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:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25059
This fixes the cases where a type annotated with optional cannot
be conditionally assigned to none:
```
x : Optional[int] = 4
if ...:
x = None
```
Test Plan: Imported from OSS
Differential Revision: D16975166
Pulled By: zdevito
fbshipit-source-id: 5a7a81224d08b9447e1f4d957fcd882091e02f32
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/24989
This fixes the cases where a type annotated with optional cannot
be conditionally assigned to none:
```
x : Optional[int] = 4
if ...:
x = None
```
Test Plan: Imported from OSS
Differential Revision: D16949314
Pulled By: zdevito
fbshipit-source-id: 7f63d88b30a3f5b024c2a539aa74967c9202af00
