NOTE: this PR removes `ScheduleFlexibleInterleaved1F1B`, let me know if theres any concerns.
`ScheduleFlexibleInterleaved1F1B` is a superset of `Interleaved1F1B` and uses most of the same implementation, but relaxes the condition that `n_microbatches % pp_size == 0`. This is refactors the implementation into `Interleaved1F1B` and then removes it since it is confusing to have both schedules with similar names. This also refactors the zero bubble logic to belong in the `ZeroBubble` schedule class.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/137783
Approved by: https://github.com/wconstab
The `AttrsDescriptor` class has been present in Triton for almost a year now (introduced [here](72c9833927)), so we should be able to rely on it existing. I am in the process of supporting the new `AttrsDescriptor` class and @jansel suggested I split changes to the existing class out separately to make sure nothing breaks removing the legacy attribute descriptor attributes.
Initially I attempted to remove the branching around detecting whether `AttrsDescriptor` exists but that breaks because PyTorch must build without Triton. So, I went back and updated for the naming introduced in the commit linked above, and also removed two unused attributes `divisible_by_8` and `ids_to_fold` which were removed in Feb 2024 (https://github.com/triton-lang/triton/pull/3122 and https://github.com/triton-lang/triton/pull/3080 respectively).
With these changes only the internal workings of the `AttrsDescriptor` class will differ between supported Triton versions, but the data stored will remain consistent.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/137458
Approved by: https://github.com/jansel
A proposal addressing Issue #1489: **Optimizer should track parameter names and not id.**
(also mentioned in here: [[RFC] Introducing FQNs/clarity eyeglasses to optim state_dict](https://dev-discuss.pytorch.org/t/rfc-introducing-fqns-clarity-to-optim-state-dict/1552)
## Summary
This PR introduces a backward-compatible enhancement where optimizers track parameter names instead of just their id.
Optimizers can be initialized with `named_parameters()` as:
```python
optimizer = optim.SGD(model.named_parameters(), lr=0.01, momentum=0.9)
```
This allows for greater clarity and ease when handling optimizers, as the parameters' names are preserved within the optimizer’s `state_dict` as:
```
state_dict =
{
'state': {
0: {'momentum_buffer': tensor(...), ...},
1: {'momentum_buffer': tensor(...), ...},
},
'param_groups': [
{
'lr': 0.01,
'weight_decay': 0,
...
'params': [0,1]
'param_names' ['layer.weight', 'layer.bias'] (optional)
}
]
}
```
Loading `state_dict` is not changed (backward-compatible) and the `param_names` key will be ignored.
## Key Features
#### Named Parameters in Optimizer Initialization:
Optimizers can accept the output of `model.named_parameters()` during initialization, allowing them to store parameter names directly.
#### Parameter Names in `state_dict`:
The parameter names are saved as a list in the optimizer’s `state_dict` with key `param_names`, alongside the `params` indices, ensuring seamless tracking of both names and parameters.
## Backward Compatibility
#### No Breaking Changes:
This change is fully backward-compatible. The added `param_names` key in the optimizer's `state_dict` is ignored when loading a state to the optimizer.
#### Customization with Hooks:
For more control, the loaded state_dict can be modified using a custom `register_load_state_dict_pre_hook`, providing flexibility for different design needs.
## Documentation Updates
Please refer to the documentation changes for more details on how this feature is implemented and how it can be used effectively.
## Solution Example:
A suggested solution to the problem mentioned in #1489, for the same parameters but in a different order.
The following `register_load_state_dict_pre_hook` should be added to the optimizer before loading to enable loading the state dict :
```python
def adapt_state_dict_ids(optimizer, state_dict):
# assuming a single param group.
current_state_group = optimizer.state_dict()['param_groups'][0]
loaded_state_group = state_dict['param_groups'][0]
# same number of params, same names, only different ordering
current_state_name_to_id_mapping = {} # mapping -- param_name: id
for i, name in enumerate(current_state_group['param_names']):
current_state_name_to_id_mapping[name] = current_state_group['params'][i]
# changing the ids of the loaded state dict to match the order of the given state dict.
for i, name in enumerate(current_state_group['param_names']):
loaded_state_group['params'][i] = current_state_name_to_id_mapping[name]
return state_dict
```
In this code, the loaded `state_dict` ids are adapted to match the order of the current optimizer `state_dict`.
Both the previous and the current optimizers are required to be initiated with `named_parameters()` to have the 'param_names' key in the dict.
### Note
This is my first contribution to PyTorch, and I wish to receive feedback or suggestions for improvement.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134107
Approved by: https://github.com/janeyx99
Co-authored-by: Jane (Yuan) Xu <31798555+janeyx99@users.noreply.github.com>
When enable tunableop, It is easy to have OOM since APP usually needs large video memory size, such as running a LLM for inference. So we need a offline mode to tune the GEMMs. This PR provide an offline mode for tunableOp:
- record untuned GEMMs to file.
- a python API named tune_gemm_in_file is added to read the untuned file and tune the GEMMs in file
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128813
Approved by: https://github.com/jeffdaily, https://github.com/hongxiayang, https://github.com/naromero77amd
Co-authored-by: Nikita Shulga <2453524+malfet@users.noreply.github.com>
Partially addresses https://github.com/pytorch/pytorch/issues/128150
When you have big sums of values, we end up computing long chains of
binary addition in our FX graph representation. Not only is this ugly,
it also is quadratic, as the sympy.Add constructor is O(N) in number
of arguments. Instead, ensure that we maintain the summation as a
single FX node so we can do the entire addition all in one go.
update_hint_regression benchmark, before and after:
```
update_hint_regression,compile_time_instruction_count,2648328980
update_hint_regression,compile_time_instruction_count,2563748678
```
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136429
Approved by: https://github.com/isuruf
This file didn't had an overall in a few years so long overdue. Most of the credit goes to @orionr for gathering all of this info.
The main rules we followed:
- No code contributor is removed, they're all placed as emeritus
- Breakdown too big categories to make this document useful to know who to ping
- No category where the code is still in the codebase is removed
- We did not rework the categories (for example to be closer to module: labels) and leave that for later
- All non-emeritus names are ordered by their number of comments on issues related to their topic
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136672
Approved by: https://github.com/eqy, https://github.com/ezyang, https://github.com/seemethere, https://github.com/malfet
Summary: Previous PR forgets to change two other places that also create `constants` and `signature`.
Test Plan:
Imported from GitHub, without a `Test Plan:` line.
{F1884584338}
Differential Revision: D63027728
Pulled By: Myrthan
Co-authored-by: Jokeren <robinho364@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136514
Approved by: https://github.com/jansel
Co-authored-by: Jokeren <robinho364@gmail.com>
Summary: Currently we process events in the regular allocation path and we call cudaEventQuery to check on the events and this path can take some locks in libcuda driver. Its not entirely needed to do process events in the allocation path, we could move this to a background thread and keep processing events regularly and put the freed block to the free list.
Differential Revision: D62396585
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135524
Approved by: https://github.com/zyan0
Summary:
This diff adds an option to round the non-split blocks in caching allocator so that they can be reused without causing lots of fragmentation for large memory segments.
For example, if we specify max_split memory size as 400MB, then all allocations more than 400MB will not be split. Lets say, we allocated some 1024MB blocks and these are cached in the allocator blocks. If we request a new 500MB block, we round it to nearest power-2-division, thats 512MB, we add default kLargeBuffer of 20MB, that will be 532MB and since 532MB is less than existing 1024MB block, the 1024MB will not be used for this allocation, instead a new 512MB block will be created. In this diff, we provide an option to cofigure the kLargeBuffer for rounding and expose as a configurable option, so 512MB + max_non_split_rounding_size and if thats greater than 1024MB, we will use te 1024MB and we wont create a new 512MB block using cudaMalloc. This option is added so that we can pre-allocate some large blocks so that we can reuse them as much as possible and we dont stall on calling cudaMalloc.
Differential Revision: D62758758
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136174
Approved by: https://github.com/zyan0
reland of https://github.com/pytorch/pytorch/pull/133113
I have to create a new PR because the previous reverted PR could not either be rebased, or imported successfully :(
----
Moving DTensor to be in the public namespace, to formally add the documentation page that includes all the public APIs. This includes:
* many path renames and path import fixes
* a dedicated doc page without too much content yet (adding in the next PRs)
* To preserve the BC for users still using the torch.distributed._tensor, I added a shim script to redirect old path calls to the new module
The BC preserving is evidented by the fact that all DTensor tests are still working without changing the public imports. So it's safe to land the changes
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134203
Approved by: https://github.com/tianyu-l
Refactor exporter errors to combine old errors and new errors for API consistency.
This PR also
1. Removes the `_C._check_onnx_proto(proto)` call in the old exporter. We don't need the ONNX checker because it is limited.
2. Removes the `OnnxExporterError` defined in the dynamo module. This class unnecessarily stores the onnx program object, making it very bulky. Instead, we revert to use the plain OnnxExporterError defined in the `errors` module and use it as the base class for all errors.
3. Continues to expose `OnnxExporterError` in `torch.onnx` and the rest of the errors in `torch.onnx.errors`.
4. Removes the `CheckerError` and `InvalidExportOptionsError` from `torch.onnx`. This is BC breaking but should have low impact.
5. I did not rename existing errors out of compatibility considerations, even though `ExporterError` would have been more succinct.
Fixes https://github.com/pytorch/pytorch/issues/135125
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135180
Approved by: https://github.com/titaiwangms
Refactor exporter errors to combine old errors and new errors for API consistency.
This PR also
1. Removes the `_C._check_onnx_proto(proto)` call in the old exporter. We don't need the ONNX checker because it is limited.
2. Removes the `OnnxExporterError` defined in the dynamo module. This class unnecessarily stores the onnx program object, making it very bulky. Instead, we revert to use the plain OnnxExporterError defined in the `errors` module and use it as the base class for all errors.
3. Continues to expose `OnnxExporterError` in `torch.onnx` and the rest of the errors in `torch.onnx.errors`.
4. Removes the `CheckerError` and `InvalidExportOptionsError` from `torch.onnx`. This is BC breaking but should have low impact.
5. I did not rename existing errors out of compatibility considerations, even though `ExporterError` would have been more succinct.
Fixes https://github.com/pytorch/pytorch/issues/135125
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135180
Approved by: https://github.com/titaiwangms
Hi,
I noticed the `unfold` operator was missing on MaskedTensor.
I tested that my change works when calling unfold and backward on a `MaskedTensor` but I didn't find the tests for the dispatch of such operation. Where is it?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125262
Approved by: https://github.com/cpuhrsch
## Semantic
The semantic is
(1) By default `torch.serialization.skip_data(materialize_fake_tensors=False)` will make `torch.save` skip writing storages (but reserve space for them in the checkpoint).
```python
import torch
import torch.nn as nn
sd = nn.Linear(3, 5).state_dict()
with torch.serialization.skip_data():
torch.save(sd, 'foo.pt')
print(torch.load('foo.pt', weights_only=True))
```
(2) With `torch.serialization.skip_data(materialize_fake_tensors=True)`If FakeTensor is passed to `torch.save` the pickler will treat these FakeTensors as being "materialized" space will be reserved in the checkpoint for the associated storage bytes, and when loading the type will be Tensor instead of FakeTensor)
```python
import torch
import torch.nn as nn
from torch._subclasses.fake_tensor import FakeTensorMode
with FakeTensorMode():
m = nn.Linear(3, 5, dtype=torch.float16, device='cuda')
sd = m.state_dict()
with torch.serialization.skip_data(materialize_fake_tensors=True):
torch.save(sd, 'bla.pt')
print(torch.load('bla.pt', weights_only=True))
# OrderedDict([('weight', tensor([[0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.]], device='cuda:0', dtype=torch.float16)), ('bias', tensor([0., 0., 0., 0., 0.], device='cuda:0', dtype=torch.float16))])
```
## Follow Ups
- [ ] `torch.load` semantic for skip_data context manager
- [ ] Mechanism for getting offsets of storages saved via this method (for writing in a separate pass)
Differential Revision: [D62238610](https://our.internmc.facebook.com/intern/diff/D62238610)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134504
Approved by: https://github.com/albanD
## Semantic
The semantic is
(1) By default `torch.serialization.skip_data(materialize_fake_tensors=False)` will make `torch.save` skip writing storages (but reserve space for them in the checkpoint).
```python
import torch
import torch.nn as nn
sd = nn.Linear(3, 5).state_dict()
with torch.serialization.skip_data():
torch.save(sd, 'foo.pt')
print(torch.load('foo.pt', weights_only=True))
```
(2) With `torch.serialization.skip_data(materialize_fake_tensors=True)`If FakeTensor is passed to `torch.save` the pickler will treat these FakeTensors as being "materialized" space will be reserved in the checkpoint for the associated storage bytes, and when loading the type will be Tensor instead of FakeTensor)
```python
import torch
import torch.nn as nn
from torch._subclasses.fake_tensor import FakeTensorMode
with FakeTensorMode():
m = nn.Linear(3, 5, dtype=torch.float16, device='cuda')
sd = m.state_dict()
with torch.serialization.skip_data(materialize_fake_tensors=True):
torch.save(sd, 'bla.pt')
print(torch.load('bla.pt', weights_only=True))
# OrderedDict([('weight', tensor([[0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.]], device='cuda:0', dtype=torch.float16)), ('bias', tensor([0., 0., 0., 0., 0.], device='cuda:0', dtype=torch.float16))])
```
## Follow Ups
- [ ] `torch.load` semantic for skip_data context manager
- [ ] Mechanism for getting offsets of storages saved via this method (for writing in a separate pass)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134504
Approved by: https://github.com/albanD
## Semantic
The semantic is
(1) By default `torch.serialization.skip_data(materialize_fake_tensors=False)` will make `torch.save` skip writing storages (but reserve space for them in the checkpoint).
```python
import torch
import torch.nn as nn
sd = nn.Linear(3, 5).state_dict()
with torch.serialization.skip_data():
torch.save(sd, 'foo.pt')
print(torch.load('foo.pt', weights_only=True))
```
(2) With `torch.serialization.skip_data(materialize_fake_tensors=True)`If FakeTensor is passed to `torch.save` the pickler will treat these FakeTensors as being "materialized" space will be reserved in the checkpoint for the associated storage bytes, and when loading the type will be Tensor instead of FakeTensor)
```python
import torch
import torch.nn as nn
from torch._subclasses.fake_tensor import FakeTensorMode
with FakeTensorMode():
m = nn.Linear(3, 5, dtype=torch.float16, device='cuda')
sd = m.state_dict()
with torch.serialization.skip_data(materialize_fake_tensors=True):
torch.save(sd, 'bla.pt')
print(torch.load('bla.pt', weights_only=True))
# OrderedDict([('weight', tensor([[0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.],
# [0., 0., 0.]], device='cuda:0', dtype=torch.float16)), ('bias', tensor([0., 0., 0., 0., 0.], device='cuda:0', dtype=torch.float16))])
```
## Follow Ups
- [ ] `torch.load` semantic for skip_data context manager
- [ ] Mechanism for getting offsets of storages saved via this method (for writing in a separate pass)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134504
Approved by: https://github.com/albanD
Hi,
I noticed the `unfold` operator was missing on MaskedTensor.
I tested that my change works when calling unfold and backward on a `MaskedTensor` but I didn't find the tests for the dispatch of such operation. Where is it?
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125262
Approved by: https://github.com/cpuhrsch
Summary: Recently https://github.com/pytorch/pytorch/pull/133620 added support for automatic dynamic shapes, where a new enum, `DIM`, was introduced to provide hints like `AUTO` and `STATIC`. This PR is a nominal change where we expose the hints via the existing public `Dim` API, and remove `DIM` from the public API. The main motivation is to avoid having users need to import too many things.
Test Plan: existing
Differential Revision: D61807361
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134484
Approved by: https://github.com/angelayi
Starter version of automatic dynamic shapes for export.
Creates enums `DIM.AUTO`, `DIM.STATIC`, allowing user to specify `AUTO` for dims in dynamic_shapes specs, meaning that corresponding dims are treated as dynamic, and relevant guards will do what's necessary (e.g. refine ValueRanges, set replacements based on equality, or even set static) without raising ConstraintViolationErrors. Basically allows the user to say, "a bunch of these dims can be dynamic, let export do model analysis and return the program with maximum possible dynamism, without complaining".
The usage for specifying `dynamic_shapes` is now:
```
AUTO -> dynamic by default, return whatever produce_guards() says, even if it's static
None/int/STATIC -> static
Dim/DerivedDim -> same as before - will complain if the min/max range is invalid, or if dims related to this are unspecified.
```
Caveat 1: specifying `AUTO` for a dim won't guarantee it'll be dynamic:
- specifying `AUTO` for a dim will return the maximum possible dynamism given your program and other specified constraints, but this can still mean you'll get a static program. For example, with the program below, x is specified dynamic, but it's equal to y, which is specified static, and with how we currently do things we won't promote y to dynamic, but will demote(?) x to static. So this can be surprising if you don't fully know your model, and/or missed one of your other inputs when specifying auto-dynamic shapes.
```
class Foo(torch.nn.Module):
def forward(self, x, y):
return x + y
inputs = (torch.randn(6), torch.randn(6))
export(Foo(), inputs, dynamic_shapes={"x": (DIM.AUTO,), "y": None})
```
Caveat 2: specifying `AUTO` and Dims in the same spec is still problematic:
- The way Dims/DerivedDims are currently handled is very strict. A Dim represents a symbol, and we require a user to specify the symbol for all dims governed by the symbol - that's why we've seen errors in the past like `The values of x must always be related to y by ...`, asking the user to specify the exact relation as in the program. We also require the specified min/max range to be a subset of the valid range from model analysis. All this doesn't compose well with specifying `AUTO` just yet - for example in the program below, ideal behavior could be to return a dynamic program, where `dx = x.size(0) = y.size(0)` has range (3,6). Unfortunately this crashes, and correct behavior is to specify `dx` for both inputs. So currently we raise a UserError and crash if both Dims + `AUTO` are present in the spec.
```
class Foo(torch.nn.Module):
def forward(self, x, y):
return x + y
inputs = (torch.randn(6), torch.randn(6))
export(Foo(), inputs, dynamic_shapes={"x": (DIM.AUTO,), "y": {0: Dim("dx", min=3, max=6)}}) # this doesn't work, because x & y and related
```
Implementation details:
This is done by setting `assume_static_by_default=False`, and doing a transform on the `dynamic_shapes` spec to preserve semantics. `assume_static_by_default=False` will treat unspecified dims or Nones as dynamic. This is the opposite of what `export.export()` currently does - unspecified Dims/Nones are treated as static. Historically this static-by-default behavior, where the user deals with fewer guards, has been desirable, and we would like to respect that in this implementation. So this internal spec transformation is added, `_transform_shapes_for_default_dynamic()`, does the spec conversion necessary to be compatbile with dynamic by default. Specifically, AUTOs are converted into Nones, and Nones/unspecified dims are filled in with explicitly static constraints.
For example, this would look like, for a 3-d tensor: `{0: DIM.AUTO, 1: None, 2: Dim("dx")} -> {0: None, 1: 32, 2: Dim("dx")}`
This does seem overly complicated, but it's done to preserve dynamic shapes semantics for `torch._dynamo.export()`, which already uses `assume_static_by_default=False`, and follows the same process for generating shape constraints , via `_process_dynamic_shapes`. There the semantics are:
```
None/unspecified: dynamic by default
Dim/DerivedDim: also a strict assertion
```
If we don't care about BC for `_dynamo.export(dynamic_shapes)`, then we can just modify semantics for `_process_dynamic_shapes()` and change all the relevant tests in `test/dynamo/test_export.py`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133620
Approved by: https://github.com/avikchaudhuri
Zero bubble can be expressed through `ScheduleFlexibleInterleaved1F1B` by setting `enable_zero_bubble=True`. But instead of having to include this flag in schedule initialization we should create a separate ZeroBubbleSchedule and also transition `Interleaved1F1B` to derive from `ScheduleFlexibleInterleaved1F1B`. Then we dont need to expose `ScheduleFlexibleInterleaved1F1B` since the naming is not obvious
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133467
Approved by: https://github.com/wconstab
ghstack dependencies: #132691
This PR switches to cuDSS library and has the same purpose of #127692, which is to add Sparse CSR tensor support to linalg.solve.
Fixes#69538
Minimum example of usage:
```
import torch
if __name__ == '__main__':
spd = torch.rand(4, 3)
A = spd.T @ spd
b = torch.rand(3).to(torch.float64).cuda()
A = A.to_sparse_csr().to(torch.float64).cuda()
x = torch.linalg.solve(A, b)
print((A @ x - b).norm())
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129856
Approved by: https://github.com/amjames, https://github.com/lezcano, https://github.com/huydhn
Co-authored-by: Zihang Fang <zhfang1108@gmail.com>
Co-authored-by: Huy Do <huydhn@gmail.com>
Summary:
This PR adds in cuSPARSELt as a backend to PyTorch.
It is now possible to see if cuSPARSELt is available and the version if
it is with
```
torch.backends.cusparselt.is_available()
torch.backends.cusparselt.version()
```
Test Plan:
```
python test/test_sparse_semi_structured.py -k test_cusparselt_backend
```
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128534
Approved by: https://github.com/cpuhrsch, https://github.com/eqy, https://github.com/syed-ahmed
Add decorator `torch.compiler.substitute_in_graph` to register polyfill for unsupported C++ function to avoid graph break. This API provides an official way to add support for dynamo for third-party C extensions. Also, it can be used to simplify our implementation for `torch._dynamo.polyfill`.
5ee070266f/torch/_dynamo/variables/builtin.py (L97-L107)
Example:
```python
>>> import operator
>>> operator.indexOf([1, 2, 3, 4, 5], 3)
2
>>> torch.compile(operator.indexOf, fullgraph=True)([1, 2, 3, 4, 5], 3)
Unsupported: ...
>>> @torch.compiler.substitute_in_graph(operator.indexOf)
... def indexOf(sequence, x):
... for i, item in enumerate(sequence):
... if item is x or item == x:
... return i
... raise ValueError("sequence.index(x): x not in sequence")
>>> torch.compile(operator.indexOf, fullgraph=True)([1, 2, 3, 4, 5], 3)
2
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133712
Approved by: https://github.com/jansel
Link various classes and functions of the `optim.swa.util` to make doc content accessible from the `torch.optim` doc.
Currently, if you click the link,
https://pytorch.org/docs/stable/optim.html#module-torch.optim.swa_utils it goes to a blank, bottom of the page section of `torch.optim`.
Also,
`torch.optim.swa_utils.AveragedModel` and `torch.optim.swa_utils.SWALR` classes as well as `torch.optim.swa_utils.update_bn()` and `optim.swa_utils.get_ema_multi_avg_fn` are not linked to doc.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133393
Approved by: https://github.com/janeyx99
Add decorator `torch.compiler.substitute_in_graph` to register polyfill for unsupported C++ function to avoid graph break. This API provides an official way to add support for dynamo for third-party C extensions. Also, it can be used to simplify our implementation for `torch._dynamo.polyfill`.
5ee070266f/torch/_dynamo/variables/builtin.py (L97-L107)
Example:
```python
>>> import operator
>>> operator.indexOf([1, 2, 3, 4, 5], 3)
2
>>> torch.compile(operator.indexOf, fullgraph=True)([1, 2, 3, 4, 5], 3)
Unsupported: ...
>>> @torch.compiler.substitute_in_graph(operator.indexOf)
... def indexOf(sequence, x):
... for i, item in enumerate(sequence):
... if item is x or item == x:
... return i
... raise ValueError("sequence.index(x): x not in sequence")
>>> torch.compile(operator.indexOf, fullgraph=True)([1, 2, 3, 4, 5], 3)
2
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133712
Approved by: https://github.com/jansel
Add decorator `torch.compiler.substitute_in_graph` to register polyfill for unsupported C++ function to avoid graph break. This API provides an official way to add support for dynamo for third-party C extensions. Also, it can be used to simplify our implementation for `torch._dynamo.polyfill`.
5ee070266f/torch/_dynamo/variables/builtin.py (L97-L107)
Example:
```python
>>> import operator
>>> operator.indexOf([1, 2, 3, 4, 5], 3)
2
>>> torch.compile(operator.indexOf, fullgraph=True)([1, 2, 3, 4, 5], 3)
Unsupported: ...
>>> @torch.compiler.substitute_in_graph(operator.indexOf)
... def indexOf(sequence, x):
... for i, item in enumerate(sequence):
... if item is x or item == x:
... return i
... raise ValueError("sequence.index(x): x not in sequence")
>>> torch.compile(operator.indexOf, fullgraph=True)([1, 2, 3, 4, 5], 3)
2
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133712
Approved by: https://github.com/jansel
Moving DTensor to be in the public namespace, to formally add the
documentation page that includes all the public APIs. This includes:
* many path renames and path import fixes
* a dedicated doc page without too much content yet (adding in the next
PRs)
* To preserve the BC for users still using the `torch.distributed._tensor`,
I added a shim script to redirect old path calls to the new module
The BC preserving is evidented by the fact that all DTensor tests are still
working without changing the public imports. So it's safe to land the
changes
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133113
Approved by: https://github.com/XilunWu
ghstack dependencies: #133305, #133306
The `LRScheduler` class provides methods to adjusts the learning rate during optimization (as updated in this PR). Also, as a note, all the classes of lr_scheduluer are already provided in the `How to adjust learning rate` section.
Fixes#127884
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133243
Approved by: https://github.com/janeyx99
The goal of this PR is to avoid stack overflow when we create extremely long chains of thunks, and then evaluate them (e.g., as occurs if you sum(long list of symint)). The basic idea behind this PR is to only thunkify proxies if they're being created in places where they may or may not be used--crucially, symint operations that occur in user code we are tracing are eagerly placed into the graph, even if they may eventually be dead.
I annotated the PR with explanation of changes.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132421
Approved by: https://github.com/Skylion007, https://github.com/zou3519
ghstack dependencies: #132674, #132675
Instead of having a separate context variable for SymDispatchMode, we
now simply delegate to the current active proxy tensor mode when we
need to trace a SymInt. We maintain a separate `__sym_dispatch__` magic
method as the calling convention is different than `__torch_dispatch__`.
Consolidating the modes in this ways means that we can consistently
disable both of these modes in tandem simply by removing the mode
from the proxy mode infra slot.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132674
Approved by: https://github.com/zou3519, https://github.com/bdhirsh
There are still some differences between CUDA and non-CUDA custom devices when
construct FSDP because CUDA is selected as the default device. For example,
when construct FSDP from CPU model and device_id is not passed, device_handle
will choose CUDA as default device. This PR will autoselect the real device
as the default device.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127609
Approved by: https://github.com/awgu
The goal of this PR is to avoid stack overflow when we create extremely long chains of thunks, and then evaluate them (e.g., as occurs if you sum(long list of symint)). The basic idea behind this PR is to only thunkify proxies if they're being created in places where they may or may not be used--crucially, symint operations that occur in user code we are tracing are eagerly placed into the graph, even if they may eventually be dead.
I annotated the PR with explanation of changes.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132421
Approved by: https://github.com/Skylion007, https://github.com/zou3519
ghstack dependencies: #132674, #132675
Instead of having a separate context variable for SymDispatchMode, we
now simply delegate to the current active proxy tensor mode when we
need to trace a SymInt. We maintain a separate `__sym_dispatch__` magic
method as the calling convention is different than `__torch_dispatch__`.
Consolidating the modes in this ways means that we can consistently
disable both of these modes in tandem simply by removing the mode
from the proxy mode infra slot.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132674
Approved by: https://github.com/zou3519, https://github.com/bdhirsh
Add similar semantics for creating a buffer object similar to creating a parameter. This is done by introducing a new Buffer class that can be used for type disambiguation. The underlying functionality of registering a buffer remains the same as the register_buffer method has not been changed. The persistent parameter in the Buffer type is to indicate whether a buffer object should be persistent or not. Other non-test changes have to do with getting the new Buffer type recognized by inductor and dynamo. Remaining changes are test changes to make sure that the Buffer type can be used as a drop in replacement for register_buffer as it just leads to register_buffer being called. The addition of this new functionality still allows for normal tensors to be used as buffers so these changes are intended to be backwards compatible.
Fixes#35735
Co-authored-by: Mikayla Gawarecki <mikaylagawarecki@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125971
Approved by: https://github.com/albanD, https://github.com/anijain2305, https://github.com/mlazos
#109581
At this point, the vanilla implementation (the default) is good.
Docs: https://docs-preview.pytorch.org/pytorch/pytorch/129905/generated/torch.optim.Adafactor.html#torch.optim.Adafactor
Specifically, the impl in this PR, which attempts to replicate the paper,
```
optim = torch.optim.Adafactor([weight])
```
is close enough to https://pytorch-optimizers.readthedocs.io/en/latest/optimizer/#pytorch_optimizer.AdaFactor
```
optim_c = AdaFactor([weight], betas=(0, 0.999), scale_parameter=False)
```
is close enough to https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/Adafactor
```
optim = keras.optimizers.Adafactor(learning_rate=0.01)
```
The three results respectively for the same randomly generated weights:
```
# ours
tensor([[ 0.3807594, -0.3912092],
[ 0.0762539, 0.5377805],
[ 0.2459473, 0.4662207]])
# pytorch-optimizer
tensor([[ 0.3807592, -0.3912172],
[ 0.0762507, 0.5377818],
[ 0.2459457, 0.4662213]])
# keras
array([[ 0.38076326, -0.39121315],
[ 0.0762547 , 0.5377859 ],
[ 0.24594972, 0.46622536]], dtype=float32)
```
This gives me confidence to move forward in speeding up the implementation now that a baseline has been established. If you're curious about differences:
* keras assigns step_size (rho_t in their code) to `min(lr, 1 / sqrt(step)` whereas the OG impl uses a hardcoded 0.01 instead of lr. We do the same thing as keras, but our lr default is 0.01.
* We differ from the pytorch-optimizers default in that our default will not track momentum (thus `beta1=0`) and we do not apply parameter scaling.
<details>
Keras collab: https://colab.research.google.com/drive/1i3xF8ChL7TWKJGV_5v_5nMhXKnYmQQ06?usp=sharing
My script repro:
```
import torch
from pytorch_optimizer import AdaFactor
torch.set_printoptions(precision=7)
weight = torch.tensor([[ 0.37697506, -0.39500135],
[ 0.07246649, 0.53399765],
[ 0.24216151, 0.46243715]], dtype=torch.float32)
# bias = torch.tensor([0, 0], dtype=torch.float32)
weight.grad = torch.tensor([[-0.5940447, -0.7743838],
[-0.5940447, -0.7743838],
[-0.5940447, -0.7743838]], dtype=torch.float32)
# bias.grad = torch.tensor([-2.5027974, 1.5422692], dtype=torch.float32)
weight_c = weight.clone()
weight_c.grad = weight.grad.clone()
optim = torch.optim.Adafactor([weight])
optim.step()
print(weight)
optim_c = AdaFactor([weight_c], betas=(0, 0.999), scale_parameter=False)
optim_c.step()
print(weight_c)
```
<details>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129905
Approved by: https://github.com/albanD
Fixes#130284Fixes#130653
- Add `torch.library.register_vmap` to custom ops
- Add `register_vmap` for operators in ops in custom_op_db.
- Make `torch.autograd.Function` support kwarg-only kwargs for vmap
- test operators in op_db with `tests/test_vmap`.
- change `test_vmap` to allow custom `out_dim` and allow "None" in `out_dim` when testing.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130589
Approved by: https://github.com/zou3519
Fixes#130284Fixes#130653
- Add `torch.library.register_vmap` to custom ops
- Add `register_vmap` for operators in ops in custom_op_db.
- Make `torch.autograd.Function` support kwarg-only kwargs for vmap
- test operators in op_db with `tests/test_vmap`.
- change `test_vmap` to allow custom `out_dim` and allow "None" in `out_dim` when testing.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130589
Approved by: https://github.com/zou3519
Summary:
Add three top level APIs for numeric debugger in pt2e flow that can log intermediate output in the model
and calculate summary for metric comparisons between nodes in two graphs
* `prepare_for_propagation_comparison`
* `extract_results_from_loggers`
* `compare_results`
Test Plan:
python test/test_quantization.py -k test_prepare_for_propagation_comparison
python test/test_quantization.py -k test_extract_results_from_loggers
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130643
Approved by: https://github.com/dulinriley, https://github.com/tarun292
# Motivation
Before this PR, device construction was `cuda` type when only a device index was given. It also returns the `PrivateUser1` type if a `PrivateUser1` type is registered.
```bash
>>> import torch
>>> device = torch.device(0)
>>> device.type
'cuda'
>>> a = torch.tensor([1, 2])
>>> b = a.to(0)
>>> b
tensor([1, 2], device='cuda:0')
```
It works well on CUDA GPU. But it will raise unexpected information and error running on XPU.
```bash
>>> import torch
>>> device = torch.device(0)
>>> device.type
'cuda'
>>> a = torch.tensor([1, 2])
>>> b = a.to(0)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/xxx/pytorch/torch/cuda/__init__.py", line 302, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled
```
With this PR, refine the logic to use the currently available device type instead.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129119
Approved by: https://github.com/albanD, https://github.com/gujinghui, https://github.com/EikanWang
ghstack dependencies: #129463, #129205, #129363
Looks like one of the first failures seen is `test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda` when `test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda` passes.
What seems interesting here is that the `torch.compile` version fails while the eager version passes. Not sure what the difference would be here...
Nevertheless, is there a recommended mechanism to skip cuDNN SDPA as a backend for this test? CC @drisspg
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125343
Approved by: https://github.com/Skylion007
Changes by apply order:
1. Replace all `".."` and `os.pardir` usage with `os.path.dirname(...)`.
2. Replace nested `os.path.dirname(os.path.dirname(...))` call with `str(Path(...).parent.parent)`.
3. Reorder `.absolute()` ~/ `.resolve()`~ and `.parent`: always resolve the path first.
`.parent{...}.absolute()` -> `.absolute().parent{...}`
4. Replace chained `.parent x N` with `.parents[${N - 1}]`: the code is easier to read (see 5.)
`.parent.parent.parent.parent` -> `.parents[3]`
5. ~Replace `.parents[${N - 1}]` with `.parents[${N} - 1]`: the code is easier to read and does not introduce any runtime overhead.~
~`.parents[3]` -> `.parents[4 - 1]`~
6. ~Replace `.parents[2 - 1]` with `.parent.parent`: because the code is shorter and easier to read.~
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129374
Approved by: https://github.com/justinchuby, https://github.com/malfet
Looks like one of the first failures seen is `test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda` when `test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda` passes.
What seems interesting here is that the `torch.compile` version fails while the eager version passes. Not sure what the difference would be here...
Nevertheless, is there a recommended mechanism to skip cuDNN SDPA as a backend for this test? CC @drisspg
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125343
Approved by: https://github.com/Skylion007
Changes by apply order:
1. Replace all `".."` and `os.pardir` usage with `os.path.dirname(...)`.
2. Replace nested `os.path.dirname(os.path.dirname(...))` call with `str(Path(...).parent.parent)`.
3. Reorder `.absolute()` ~/ `.resolve()`~ and `.parent`: always resolve the path first.
`.parent{...}.absolute()` -> `.absolute().parent{...}`
4. Replace chained `.parent x N` with `.parents[${N - 1}]`: the code is easier to read (see 5.)
`.parent.parent.parent.parent` -> `.parents[3]`
5. ~Replace `.parents[${N - 1}]` with `.parents[${N} - 1]`: the code is easier to read and does not introduce any runtime overhead.~
~`.parents[3]` -> `.parents[4 - 1]`~
6. ~Replace `.parents[2 - 1]` with `.parent.parent`: because the code is shorter and easier to read.~
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129374
Approved by: https://github.com/justinchuby, https://github.com/malfet
When caching is enabled, an internal model fails with
```
assert_size_stride(bmm_9, (17, s0, 512), (54784, 512, 1))
AssertionError: expected size 17==17, stride 57344==54784 at dim=0
```
looking at this model, the exact problem is when the cache is hit on the forward graph, the generated code for backward fails since the strides of the outputs of forward, passed to backward as inputs, are not what we expected.
This PR changes the evaluation logic so that we defer evaluation of output stride exprs to load path as opposed to eagerly doing it on save path.
I have not been able to come up with a unit test repro for this problem.
Differential Revision: [D58796503](https://our.internmc.facebook.com/intern/diff/D58796503)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128997
Approved by: https://github.com/ezyang
Summary:
WARNING: This API is highly unstable and will be subject to change in the future.
Add a protoype to "decompose" an ExportedProgram into a joint graph form, so that we can compute the gradients on this graph.
Test Plan: buck test mode/opt caffe2/torch/fb/export:test_experimental
Differential Revision: D55657917
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128847
Approved by: https://github.com/tugsbayasgalan
I'd like to discuss the criteria that we regard an implementation as stable. If there is no existing standard, my initial proposal would be a 6 month period after the commit to regard it as stable. As a result, now Adam and AdamW on CUDA would be considered as stable, while the rest are of beta.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129006
Approved by: https://github.com/malfet
### bc-breaking for existing users of the private API:
- Existing policy functions must now change their return value to be [CheckpointPolicy](c0b40ab42e/torch/utils/checkpoint.py (L1204-L1230)) Enum instead of bool.
- To restore previous behavior, return `PREFER_RECOMPUTE` instead of `False` and `{PREFER,MUST}_SAVE` instead of `True` depending whether you prefer the compiler to override your policy.
- Policy function now accepts a `ctx` object instead of `mode` for its first argument.
- To restore previous behavior, `mode = "recompute" if ctx.is_recompute else "forward"`.
- Existing calls to `_pt2_selective_checkpoint_context_fn_gen` must be renamed to `create_selective_checkpoint_contexts `. The way you use the API remains the same. It would've been nice to do something different (not make the user have to use functools.partial?), but this was the easiest to compile (idk if this should actually be a constraint).
Related doc: https://docs.google.com/document/d/1BKyizkZPdri9mHqdDOLAUpkI7SbbKfLHRFVVpK9ZWqo/edit
Memory considerations:
- As with the existing SAC, cached values are cleared upon first use.
- We error if the user wishes to backward a second time on a region forwarded with SAC enabled.
In-place:
- We use version counting to enforce that if any cached tensor has been mutated. In-place operations not mutating cached tensors are allowed.
- `allow_cache_entry_mutation=True` can be passed to disable this check (useful in the case of auto AC where the user is cleverly also saves the output of the in-place)
Randomness, views
- Currently in this PR, we don't do anything special for randomness or views, the author of the policy function is expected to handle them properly. (Would it would be beneficial to error? - we either want to save all or recompute all random tensors)
Tensor object preservation
- ~We guarantee that if a tensor does not requires grad, and it is saved, then what you get out is the same tensor object.~ UPDATE: We guarantee that if a tensor is of non-differentiable dtype AND it is not a view, and it is saved, then what you get out is the same tensor object. This is a nice guarantee for nested tensors which care about the object identity of of the offsets tensor.
Policy function
- Enum values are `{MUST,PREFER}_{SAVE,RECOMPUTE}` (bikeshed welcome). Alternatively there was `{SAVE,RECOMPUTE}_{NON_,}OVERRIDABLE`. The former was preferred bc it seemed clearer that two `MUST` clashing should error, versus it is ambiguous whether two `NON_OVERRIDABLE` being stacked should silently ignore or error.
- The usage of Enum today. There actually is NO API to stack SAC policies today. The only thing the Enum should matter for in the near term is the compiler. The stacking SAC policy would be useful if someone wants to implement something like simple FSDP, but it is not perfect because with a policy of `PREFER_SAVE` you are actually saving more than autograd would save normally (would be fixed with AC v3).
- The number of times we call the policy_fn is something that should be documented as part of public API. We call the policy function for all ops except ~~detach~~ UPDATE : metadata ops listed in `torch.utils.checkpoint.SAC_IGNORED_OPS`) because these ops may be called a different number of times by AC itself between forward and recompute.
- The policy function can be a stateful object (we do NOT make separate copies of this object for forward/recompute, the user is expected to handle that via is_recompute see below).
Tensors guaranteed to be the same tensor as-is
- Policy function signature takes ctx object as its first argument. The ctx function is an object encapsulating info that may be useful to the user, it currently only holds "is_recompute". Adding this indirection gives us flexibility to add more attrs later if necessary.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125795
Approved by: https://github.com/Chillee, https://github.com/fmassa
Fixes#127908
## Description
Created docs to document the torch.cuda.cudart function to solve the issue #127908.
I tried to stick to the [guidelines to document a function](https://github.com/pytorch/pytorch/wiki/Docstring-Guidelines#documenting-a-function) but I was not sure if there is a consensus on how to handle the docs of a function that calls an internal function. So I went ahead and tried what the function will raise, etc. from the user endpoint and documented it (i.e. I am giving what actually _lazy_init() will raise).
Updated PR from #128298 since I made quite a big mistake in my branch. I apologize for the newbie mistake.
### Summary of Changes
- Added docs for torch.cuda.cudart
- Added the cudart function in the autosummary of docs/source/cuda.rst
## Checklist
- [X] The issue that is being fixed is referred in the description
- [X] Only one issue is addressed in this pull request
- [X] Labels from the issue that this PR is fixing are added to this pull request
- [X] No unnecesary issues are included into this pull request
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128741
Approved by: https://github.com/msaroufim
The following are all constrained under the ONNX exporter project scope.
- `personal_of_interest.rst`
- Moving folks no longer working on the project to emeritus.
- Adding @justinchuby, @titaiwangms, @shubhambhokare1 and @xadupre,
who have all made countless contributions to this project.
- `CODEOWNERS`
- Removing folks no longer working on the project.
- Updating new owners who will now be notified with PRs related to
the specific file paths.
- `merge_rules.yaml`
- Removing folks no longer working on the project.
🫡
Co-authored-by: Justin Chu <justinchuby@users.noreply.github.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126364
Approved by: https://github.com/titaiwangms, https://github.com/justinchuby, https://github.com/albanD
Related doc: https://docs.google.com/document/d/1BKyizkZPdri9mHqdDOLAUpkI7SbbKfLHRFVVpK9ZWqo/edit
Memory considerations:
- As with the existing SAC, cached values are cleared upon first use.
- We error if the user wishes to backward a second time on a region forwarded with SAC enabled.
In-place:
- We use version counting to enforce that if any cached tensor has been mutated. In-place operations not mutating cached tensors are allowed.
- `allow_cache_entry_mutation=True` can be passed to disable this check (useful in the case of auto AC where the user is cleverly also saves the output of the in-place)
Randomness, views
- Currently in this PR, we don't do anything special for randomness or views, the author of the policy function is expected to handle them properly. (Would it would be beneficial to error? - we either want to save all or recompute all random tensors)
Tensor object preservation
- We guarantee that if a tensor does not requires grad, and it is saved, then what you get out is the same tensor object. If the tensor does require grad, we must detach to avoid creating a reference cycle. This is a nice guarantee for nested tensors which care about the object identity of of the offsets tensor.
Policy function
- Enum values are `{MUST,PREFER}_{SAVE,RECOMPUTE}` (bikeshed welcome). Alternatively there was `{SAVE,RECOMPUTE}_{NON_,}OVERRIDABLE`. The former was preferred bc it seemed clearer that two `MUST` clashing should error, versus it is ambiguous whether two `NON_OVERRIDABLE` being stacked should silently ignore or error.
- The usage of Enum today. There actually is NO API to stack SAC policies today. The only thing the Enum should matter for in the near term is the compiler. The stacking SAC policy would be useful if someone wants to implement something like simple FSDP, but it is not perfect because with a policy of `PREFER_SAVE` you are actually saving more than autograd would save normally (would be fixed with AC v3).
- The number of times we call the policy_fn is something documented part of public API. We call the policy function for all ops except detach because detach is itself called a different number of times by AC between forward and recompute.
- The policy function can be a stateful object (we do NOT make separate copies of this object for forward/recompute, the user is expected to handle that via is_recompute see below).
Tensors guaranteed to be the same tensor as-is
- Policy function signature takes ctx object as its first argument. The ctx function is an object encapsulating info that may be useful to the user, it currently only holds "is_recompute". Adding this indirection gives us flexibility to add more attrs later if necessary.
"bc-breaking" for existing users of the private API:
- Existing policy functions must now change their return value to use the Enum.
- Existing calls to `_pt2_selective_checkpoint_context_fn_gen` must be renamed to `gen_selective_checkpoint_context_fn`. The way you use the API remains the same. It would've been nice to do something different (not make the user have to use functools.partial?), but this was the easiest to compile (idk if this should actually be a constraint).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125795
Approved by: https://github.com/Chillee, https://github.com/fmassa
Looks like one of the first failures seen is `test_causal_variants_compile_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda` when `test_causal_variants_causal_variant_CausalVariant_LOWER_RIGHT_shape0_cuda` passes.
What seems interesting here is that the `torch.compile` version fails while the eager version passes. Not sure what the difference would be here...
Nevertheless, is there a recommended mechanism to skip cuDNN SDPA as a backend for this test? CC @drisspg
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125343
Approved by: https://github.com/Skylion007
In case user modified stage module out of place, such as
mod = DDP(mod)
mod = torch.compile(mod)
They need a stage builder else than `pipe.build_stage()`.
This PR provides an API to do so:
```
def build_stage(
stage_module,
stage_index,
pipe.info(),
...
)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128273
Approved by: https://github.com/wconstab
----
- Bring PipelineStage/Schedule more front-and-center
- provide details on how to manually construct PipelineStage
- move tracer example and manual example below so the high-level flow
(e2e) is closer to the top
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128236
Approved by: https://github.com/H-Huang
ghstack dependencies: #128201, #128228
Changed the API of `pipeline()` to take microbatch instead of full batch as example args.
Main purpose is to:
- make this API more atomic;
- decouple tracing frontend from runtime info like `num_chunks`.
Side effects:
- Creates opportunity for varying `num_chunks` of schedules with the same `pipe` object.
- User has to create example microbatch input.
- Chunk spec stuff are now all moved to runtime side.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128163
Approved by: https://github.com/H-Huang
Renaming ManualPipelineStage to remove the "Manual" part. I needed to replace the existing `PipelineStage` which takes in the `pipe` argument, so I have renamed that to `TracerPipelineStage`. @kwen2501 will remove this entirely in favor of adding a util to `Pipe` to just create the stage directly.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128157
Approved by: https://github.com/wconstab
Summary:
Part of the work helping export's automatic dynamic shapes / dynamic shapes refining based on suggested fixes.
Introduces a util function refine_dynamic_shapes_from_suggested_fixes() that takes the error message from a ConstraintViolationError message containing suggested dynamic shapes fixes, along with the original dynamic shapes spec, and returns the new spec. Written so that the suggested fixes from export can be directly parsed and used.
Example usage for the automatic dynamic shapes workflow:
```
# export, fail, parse & refine suggested fixes, re-export
try:
export(model, inps, dynamic_shapes=dynamic_shapes)
except torch._dynamo.exc.UserError as exc:
new_shapes = refine_dynamic_shapes_from_suggested_fixes(exc.msg, dynamic_shapes)
export(model, inps, dynamic_shapes=new_shapes)
```
For examples of behavior, see the added test and docstring. Will take suggestions for renaming the function to something else 😅
Test Plan: test_export tests
Differential Revision: D57409142
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127436
Approved by: https://github.com/avikchaudhuri
Fixes#126367.
## Description
Fixed a broken link in the pytorch/docs/source/torch.compiler_faq.rst doc and deleted a few words that were extra according to the issue tagged above.
## Checklist
- [X] The issue that is being fixed is referred in the description
- [X] Only one issue is addressed in this pull request
- [X] Labels from the issue that this PR is fixing are added to this pull request
- [X] No unnecesary issues are included into this pull request
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127938
Approved by: https://github.com/msaroufim
For a masked `tl.load` operation, the Triton language specifies that values masked out (i.e. where the mask evaluates to false) are undefined in the output of the load. Triton provides an optional `other` parameter which, when included, provides an explicit value to use for masked out values from the load. If the output from a masked load without the `other` parameter is used in a conditional, unexpected behavior can occur.
Despite the language specification, all Triton backends currently in use by PyTorch Inductor (NVIDIA, AMD, and Intel) 0-initialize masked loads if `other` is not present (we recently changed the Intel backend behavior to match NVIDIA and AMD because that's what our users expect, even if we are not following the Triton spec to the tee). This PR attempts to "future-proof" Inductor for new backends (or perhaps changes in the current backends? - we did not see any performance change from 0-initializing in the Intel XPU backend but one could imagine compiler optimizations to remove paths that depend on undefined) to add an explicit `other` in instances where later conditionals depend on the `tl.load` output. I also removed an exception to `other` behavior for boolean loads, which was put in place for a Triton bug that should be fixed. I added `other` to the getting started documentation as a clue that masked load behavior requires explicit initialization if, even though I don't expect `undef` values to cause the example code to fail if the underlying output is not 0-initialized. Finally, I added other to the `make_load` function in `select_algorithm.py`, though I wasn't able to determine if that function was actually being called.
Fixes#126535
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127311
Approved by: https://github.com/jansel
We create a new landing page for PyTorch custom ops (suggested by
jansel). All of our error messages will link here, and I'll work with
the docs team to see if we can boost SEO for this page.
NB: the landing page links some non-searchable webpages. Two of those
(the Python custom ops tutorial and C++ custom ops tutorial) will turn
into actual webpages when PyTorch 2.4 comes around. I'll make the third one
(the Custom Operators Manual) once it stabilizes (we continously add new
things to it and the length means that we might want to create a custom
website for it to make the presentation more ingestable).
Test Plan:
- view docs preview.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127400
Approved by: https://github.com/jansel
ghstack dependencies: #127291, #127292
The `usort` config in `pyproject.toml` has no effect due to a typo. Fixing the typo make `usort` do more and generate the changes in the PR. Except `pyproject.toml`, all changes are generated by `lintrunner -a --take UFMT --all-files`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127126
Approved by: https://github.com/kit1980
The `usort` config in `pyproject.toml` has no effect due to a typo. Fixing the typo make `usort` do more and generate the changes in the PR. Except `pyproject.toml`, all changes are generated by `lintrunner -a --take UFMT --all-files`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127126
Approved by: https://github.com/kit1980
ghstack dependencies: #127122, #127123, #127124, #127125
The `usort` config in `pyproject.toml` has no effect due to a typo. Fixing the typo make `usort` do more and generate the changes in the PR. Except `pyproject.toml`, all changes are generated by `lintrunner -a --take UFMT --all-files`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127124
Approved by: https://github.com/Skylion007
ghstack dependencies: #127122, #127123
# Motivation
## for `torch.amp.GradScaler`,
- `torch.cpu.amp.GradScaler(args...)` is completely equivalent to `torch. amp.GradScaler("cpu", args...)`.
- `torch.cuda.amp.GradScaler(args...)` is completely equivalent to `torch.amp.GradScaler("cuda", args...)`.
So, we intend to depreate them and **strongly recommend** developer to use `torch.amp.GradScaler`.
## for `custom_fwd` and `custom_bwd`,
this is a good solution to make the custom function run with or without effect even in an autocast-enabled region and can be shared by other backends, like CPU and XPU.
So we generalize it to be device-agnostic and put them int `torch/amp/autocast_mode.py` and re-expose to `torch.amp.custom_fwd` and `torch.amp.custom_bwd`. Meanwhile, we deprecate `torch.cuda.amp.custom_fwd` and `torch.cuda.amp.custom_bwd`.
# Additional Context
Add UT to cover the deprecated warning.
No need for more UTs to cover the functionality of `torch.amp.custom_f/bwd`, the existing UTs that previously covered the functionality of `torch.cuda.amp.custom_f/bwd` can cover them.
To facilitate the review, we separate these code changes to two PRs. The first PR cover `torch.amp.GradScaler`. The follow-up covers `custom_fwd` and `custom_bwd`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126527
Approved by: https://github.com/jgong5, https://github.com/gujinghui, https://github.com/janeyx99, https://github.com/EikanWang
Summary:
1. Define explicit `use_agent_store` on rdzv handlers. Handlers that set is true can share the store.
2. Instead of agent coordinating master_add/master_port values, the logic is now encapsulated by a *rdzv_handler* where `RendezvousInfo` will have `RendezvousStoreInfo` object that handlers must return.
- Depending on the implementation they can either:
- point to existing store (and expected to `use_agent_store` as true - point 1). Client code will rely on `TORCHELASTIC_USE_AGENT_STORE` env variable to know if the store is shared.
- build args that `torch.distributed.init_process_group` can bootstrap by creating new store.
Additional points:
- When TCPStore is shared, it should be wrapped in PrefixStore to qualify/scope namespace for other usecases.
- `next_rendezvous` signature changed to return instance of `RendezvousInfo` instead of a (store, rank, world_size) tuple for extensibility purposes.
Why:
- Reduce moving parts
- easier to swap implementation
- improve tractability
- addressing perf/debug-ability will benefit all usecases
-
Test Plan: CI
Differential Revision: D57055235
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125743
Approved by: https://github.com/d4l3k
Rule is enforced by #126103.
The rule:
- If `torch.a.b` defines a public class `C` (i.e. to be exposed in torch API namespace), then `torch.a.b` must be a public path, i.e. no `_`.
- `torch.a.b` should ideally have an `__all__` that defines what should be imported from this file when it is imported.
- All other definitions in `torch.a.b` that you don't want to expose should have a `_` prefix.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126812
Approved by: https://github.com/wconstab
The padded dense -> jagged conversion op has the signature:
```
_fbgemm_dense_to_jagged_forward(Tensor dense, Tensor[] offsets, SymInt? total_L=None) -> Tensor
```
when `total_L` is not specified, the meta registration has a data-dependent output shape (based on `offsets[0][-1]`). Returning an unbacked SymInt here should work in theory, but traceable wrapper subclass support is missing in later code to handle deferred runtime asserts. This PR fixes this.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126198
Approved by: https://github.com/ezyang
#### Conditions for allowlisting tensor subclasses
We allow tensor subclasses types that
(1) Do not override `__setstate__`, `__getattr__`, `__setattr__`, `__get__`, `__set__` or `__getattribute__` of `torch.Tensor` (`torch.Tensor` does not have a definition of `__getattr__`, `__get__` or `__set__` so we check that these are `None`)
(2) Use the generic `tp_alloc`
(3) Are in a module that *has been imported by the user*
to be pushed onto the stack as strings by `GLOBAL` instructions, while storing the type in a dict
The strings will be converted to the classes as appropriate when executing `REBUILD` with `_rebuild_from_type_v2`
*Note that we use `inspect.getattr_static(sys.modules[module], name)` to get the class/function as this method claims to have no code execution.
The rationale for the 3 conditions above is as follows:
The rebuild func provided by `Tensor.__reduce_ex__` is `torch._tensor._rebuild_from_type_v2`, which is defined as such (note the call to `getattr`, `Tensor.__setstate__` and the call to `as_subclass` as well as the call to `_set_obj_state` which calls `setattr`)
4e66aaa010/torch/_tensor.py (L57-L71)
`as_subclass` is implemented with a call to `THPVariable_NewWithVar`
that will eventually call `tp_alloc` here
4e66aaa010/torch/csrc/autograd/python_variable.cpp (L2053)
The `func` arg to `_rebuild_from_type_v2` for wrapper subclasses is `Tensor.rebuild_wrapper_subclass`, which will similarly call into `THPVariable_NewWithVar` and hit the above `tp_alloc`
**Note that we do not call `tp_init` or `tp_new` (i.e. `cls.__init__` or `cls.__new__`) when unpickling**
### How do we check something is a tensor subclass/constraints around imports
In order to check whether `bla` is a tensor subclass in the bytecode `GLOBAL module.name`, we need to do an `issubclass` check, which entails converting the global string to the appropriate type. We *do not* arbitrarily import modules but will perform this check as long as the given subclass (given by `module.name`) has already been imported by the user (i.e. `module in sys.modules` and `issubclass(getattr(sys[modules], name), torch.Tensor)`
This PR also allowlisted `torch._utils._rebuild_wrapper_subclass` and `torch.device` (used by `_rebuild_wrapper_subclass`)
### API for allow listing
This PR also added `torch.serialization.{add/get/clear}_safe_globals` that enables user to allowlist globals they have deemed safe and manipulate this list (for example they could allowlist a tensor subclass with a custom `__setstate__` if they have checked that this is safe).
Next steps:
- Add testing and allowlist required classes for all in-core tensor subclasses (e.g. `DTensor`, `FakeTensor` etc.)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124331
Approved by: https://github.com/albanD
This does a few things that were originally a few PRs but I am on a new machine and don't have ghstack.
If it is too problematic to review, I can re-split, just let me know.
This does:
- Cleanup context manager use in test_flop_counter
- Remove need for mod argument in FlopCounterMode, warning about it
- Re-implement a Module tracker from scratch using global forward Module use and multi_grad_hook (we cannot use global backward Module hook because they don't look for nested Tensor and they're custom Function based instead of multi_grad_hook).
- Update FlopCouterMode to use the new ModuleTracker. All the existing test suite passes as-is (only changes there are new tests and refactoring mentioned above)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125352
Approved by: https://github.com/mikaylagawarecki
`torch.utils.benchmark.Compare` is not directly exposed in torch.utils.benchmark documentation.
I think this is a valuable resource to add since it can help people embracing the torch benchmark way of doing things, and help people building documentation towards it.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125009
Approved by: https://github.com/mikaylagawarecki
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom):
* #124944
* #124939
* __->__ #122965
Differential Revision: [D55493240](https://our.internmc.facebook.com/intern/diff/D55493240/)
*This PR is now ready for merge and is not an RFC*
Major choices are:
-- the introduction of the AsyncStager protocol
-- removed `executor` from param.
-- leave async as a separate method (for now)
This proposal seeks to add extension points to dcp.async_save, allowing users to:
- Specify a specific staging method when calling async_save
- Allow a vehicle for also making the staging method async, to allow for cases where we may want to overlap with the training loop (e.g., overlap d2h with and only synchronize at the optim.step)
- Potentially specify the execution method for doing async_save in parallel. For example some users may prefer a subprocess over a thread to avoid GIL issues.
A totally reasonable alternative to this entire proposal is to expect users who want this level of customization
to write their own custom async save methods. Here's an example which addresses the issues mentioned
in PR comments.
```
def custom_async_save(...):
# this step accomplishes staging and includes the usual 'planning' calls (issue 1)
buffered_writer = CpuBufferedWriter() # this is stateful, contains a copy of state_dict
dcp.save(state_dict, storage_writer=buffered_writer)
final_storage_writer = FileSystemWriter()
mp.spawn( # issue2 is gone, do whatever you want here
dcp.save, # or some custom sub-process method which calls dcp.save under the hood
buffered_writer.state_dict, # lot's of way's to do this, not really the most important part
checkpoint_id=checkpoint_id,
storage_writer=storage_writer,
planner=planner,
process_group=process_group, # this actually wouldn't work, but again not the pt.
)
# leaving out the rest of the details for managing your extra special subprocess.
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122965
Approved by: https://github.com/daulet-askarov
Fixes#124528
Going over the options for our MapAllocator and what they do, I don't think any other of them need to be piped up to `torch.load`
4f29103749/aten/src/ATen/MapAllocator.h (L8-L16)
~However, I wonder if this `MmapVisibility(Enum)` is a good way to represent "or-ing" together of `mmap` flags if we want to extend it in the future. I looked over the flags for [`mmap(2)`](https://man7.org/linux/man-pages/man2/mmap.2.html), and could not immediately see how most of them would be useful for `torch.load` (would maybe `MAP_LOCKED` (like `mlock`) or `MAP_HUGE` ever be worthwhile?)~
Using the flags provided by the python `mmap` library so that we can extend the allowed flags and pipe them down to the cpp `mmap` call if there is a need for other flags in the future
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124889
Approved by: https://github.com/albanD
This PR introduces a new way of building `dynamic_shapes` for export. The idea is to build up a mapping from input tensors to the dynamic shapes that should be assigned to their corresponding fake tensors.
This mapping is automatically converted to the current form of `dynamic_shapes`, which must exactly match the structure of inputs. We do this by using pytree utils.
With the current `dynamic_shapes`, we had to be careful about user-defined classes that are registered with pytree, since such classes are not necessarily polymorphic containers; they may be fine containing tensors, but not dynamic shapes. Thus we had decided to allow input instances of such classes to be associated with dynamic shapes in flattened form. This decision needs to be mirrored in this PR as well. To make it easier to keep these code paths in sync, we refactor the current recursive procedure for associating inputs with dynamic shapes to use the same pytree utils. This needs minor fixes to a few tests where `dynamic_shapes` were not exactly matching the structure of inputs.
Differential Revision: D56551992
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124898
Approved by: https://github.com/zhxchen17
Summary:
This makes barrier and rank operations linear instead of quadratic with the number of workers. This drastically improves performance for rendezvous when running with over 1000 hosts.
This uses 2 approaches for different areas:
* local rank assignment: each worker does 1 set and 1 get, local ranks are assigned on the rank 0 host in a O(n) operation which reduces total store operations to be linear with number of workers.
* exit_barrier: use a counter and a final flag so each worker has to do max 1 set, 1 get and 1 add.
At 4000 hosts we see torchelastic be able to run in as little as 10 seconds down from 373 seconds.
Test Plan:
This is testing using many small tests running on a remote cluster.
{D56549942}
```
torchx run --scheduler mast -- --image=torchelastic_benchmark --j=4000x1
```
Differential Revision: D56605193
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124982
Approved by: https://github.com/kiukchung, https://github.com/kurman
MTIA device has its own Module in PyTorch now.
torch.mtia has following APIs similar to other backends. The lazy_init is also supported.
```
__all__ = [
"init",
"is_available",
"synchronize",
"device_count",
"current_device",
"current_stream",
"default_stream",
"set_stream",
"stream",
"device",
]
```
------------
For device management. We expand AccleratorHooksInterface to support generic device management and it can be used in both C++ and PyThon.
```
def _accelerator_hooks_device_count() -> _int: ...
def _accelerator_hooks_set_current_device(device_index: _int) -> None: ...
def _accelerator_hooks_get_current_device() -> _int : ...
def _accelerator_hooks_exchange_device(device_index: _int) -> _int : ...
def _accelerator_hooks_maybe_exchange_device(device_index: _int) -> _int : ...
```
---------
Adding get_device_module API to retrieve device modules for different device types.
```
def get_device_module(device: Optional[Union[torch.device, str]] = None)
```
---------
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123612
Approved by: https://github.com/albanD
ghstack dependencies: #123611
Previously, unbacked SymInts would gradually get larger and larger as we kept rebinding them. Now, we do the replacement to preserve the old symbol.
Actually doing this is a bit tricky. Here’s the order things happen when retracing data dependent:
1. Run fake tensor prop: allocate new unbacked SymInt
2. Run proxy tensor mode, calculate bindings and associate them with FX node
3. Run PropagateUnbackedSymInts, rename unbacked bindings to their old ones so they are consistent
So the problem is when we calculate bindings in step (2), we don't know
what the original names are yet, we only find out later at (3). But by
the time (3) runs, we've already stuffed some new bindings in
meta["unbacked_bindings"] and we don't know how to update them! To fix
this, I introduce resolve_unbacked_bindings which post facto applies any
of the renamings we discovered in (3).
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124782
Approved by: https://github.com/lezcano
ghstack dependencies: #124310, #124314, #124316, #124394, #124739
This PR has a lot of "draw the rest of the fucking owl" energy. Here's how to break it down.
1. **torch/_inductor/graph.py** - We start by tightening unbacked symbol invariants. Specifically, as we lower FX nodes, we check whether or not every unbacked_binding recorded on the FX node meta, actually ends up getting bound (according to get_unbacked_symbol_defs) in all the buffers generated by the lowering. Hopefully this invariant is self evident. This leads to a lot of failures.
2. **torch/_inductor/ir.py** - Problem 1: There is softness in how Inductor computes defs of unbacked symbols in IR node. Previously, we tried to infer it by looking at the output sizes/strides/etc and see if new unbacked symbols popped up that we hadn't seen in the inputs. I don't know exactly what was buggy about the old code, but sometimes we would fail to notice an unbacked symbol had been bound, or rebind an unbacked symbol multiple times. Fortunately, thanks to the earlier PRs in our stack, we now have a nice list of unbacked symbol bindings from FX, so we now just store it directly on ExternKernel and use it directly to report defs. This has to be done twice: once for FallbackKernel (e.g., nonzero) and once for DynamicScalar (e.g., item) (see also **torch/_inductor/lowering.py**, **torch/_inductor/codegen/wrapper.py** and **torch/_inductor/codegen/cpp_wrapper_cpu.py** for the lowering and codegen changes for item)
* **process_kernel** - Sidequest! It turns out that Inductor lowering can reallocate unbacked symbols. This happens specifically when we repropagate fake tensors through the operator in `process_kernel`. This repropagation process is necessary because Inductor may have changed the strides of input tensors, and it must now recompute the strides so that it can continue to appropriately plan the rest of the lowering process. This is fine: we just make sure we do the rebind unbacked + compute_unbacked_bindings dance we've been doing previously in the PR stack. But instead of putting unbacked_bindings on a new FX node, they go straight into our unbacked_bindings on the Inductor IR node.
* **codegen_unbacked_symbol_defs** - Sidequest! FallbackKernel lowering is done in two steps. First, you emit the FallbackKernel buffer. Then, you emit MultiOutput buffers which actually give access to the individual outputs of FallbackKernel, which may have been multi-output. There is a design decision here: does the FallbackKernel bind the unbacked symbols, or the MultiOutput buffer? Historically, we put the binding on MultiOutput buffer, because it's more convenient: the FallbackKernel buffer is fake, in fact, it doesn't even get a name in C++ codegen. But it's kind of inconsistent with the keypath model that we've been tracking unbacked bindings with: if you have a multi-output node, you'd expect a keypath like `[0].size()[0]` representing the first output's first dimension size. That suggests that it's the FallbackKernel that should define the things. So that was my first implementation. Unfortunately, the C++ codegen is too cursed and I could not understand how to make it work in that case. So now we just unsoundly assume you cannot have multi-output data dependent output, and do the codegen in MultiOutput. There are some comments explaining exactly what we are improperly assuming.
3. **_rename_unbacked_to** in **torch/fx/experimental/symbolic_shapes.py** - Previously, when we renamed unbacked symbols, we clobbered any facts we previously knew about them. So for example, if we had a replacement `u0 -> s0` but then we renamed u0 to u1, we would now setup the replacement `u0 -> u1`, clobbering the old replacement. This apparently didn't matter in earlier PRs in the stack, but with Inductor now on the ball, there were some tests that indicated this was a problem. The solution is easy: if u0 had a preexisting replacement, reapply it to u1. However...
* **torch/_functorch/_aot_autograd/collect_metadata_analysis.py** - When we run forward analysis, this triggers fake tensor repropagation and fresh allocations. Previously, we just cleared out the pending symbols when finished the analysis. But with the change above, this would also migrate replacements to the new symbols... which are now dead. So now we explicitly suppress generation of these symbols with `ignore_fresh_unbacked_symbols` so that no rebinding happens at all.
* **torch/_dynamo/eval_frame.py** - same deal; I just searched for all sites we called clear() on pending
4. The last step is fixing the long tail of extra problems that show up, now that unbacked_bindings are load bearing into Inductor
* **torch/_dynamo/eval_frame.py** - Some of the exports are making copies of nodes without repropagating fake tensors, so in this case, it is important to also copy the `unbacked_bindings` (apparently this didn't matter before without the Inductor changes)
* **torch/_export/pass_base.py** - I discover that this is doing fake tensor repropagation via a test suite failure. Do the same playbook as AOTAutograd: PropagateUnbackedSymInts too! Actually, they also have implemented their own tracer as well, so do the same playbook as proxy_tensor: record unbacked_bindings on the newly traced nodes. UGH code duplication.
* **torch/_subclasses/fake_tensor.py**, **torch/_subclasses/fake_impls.py** (with call site updates at **torch/_functorch/_aot_autograd/traced_function_transforms.py** and **torch/fx/passes/fake_tensor_prop.py**) - What's this new epoch thing? I noticed that sometimes I would be retracing, call nonzero() on a fake tensor, and not allocate a new unbacked symbol. This is actually bad, because if I don't get a new unbacked symbol, I don't know there's a binding site, and `unbacked_bindings` is now missing a binding. The reason for this is memoization: if I reuse the exact same fake tensor on my retrace, it will already have an unbacked symint memoized on it and we will short circuit allocation. Well, that's no good. So I associate the memos with a fake tensor epoch, and every time you start a new fake tensor propagation from scratch, you bump the epoch so that I clear all the memos.
* **torch/_inductor/scheduler.py** - I notice in unit tests that V.current_node is not always set when we call process_kernel. So I save it into the IR node and restore it when we are running `get_estimated_runtime`.
* **torch/fx/experimental/symbolic_shapes.py** - A few things
* **rebind_unbacked** (re **_tensor_version**). Ordinarily, when you have an unbacked SymInt, you persistently hvae it all the way to the end of the program. `_tensor_version` violates this: this generates an unbacked SymInt (for reasons I don't quite understand?) and then gets rid of it later. This triggered an assert violation. I think this op is kind of misusing unbacked SymInt, but I didn't know how to refactor it, so it gets a special case.
* **rebind_unbacked** (re **Simplify SymBool binding**). Ugh, SymBool, what a pain in the butt. I have an assert that you can only rebind unbacked symbol to another unbacked symbol. This assert fails when a boolean is involved, because the result of running keypath on the result is not `u1`, it's `sympy.Piecewise(... sympy.Eq(u1, 1) ...)`. This is actually just `u1`, but Sympy doesn't know it because it doesn't know that `u1` value range is `[0, 1]`. So we manually implement the simplification needed to get the assert to pass.
* **compute_unbacked_bindings** (re **This is pretty fragile**). There is a really funny disaster involving memoization and Inductor process kernel. Ordinarily when I retrace, if there was a memo hit in the old trace, there will be a memo hit in the new trace. However, Inductor process kernel breaks this, because it recreates fake tensor inputs to the operator call from scratch (since they might have different strides), and obviously these tensor inputs don't have the memo from the old one. I tried a little bit to try to manually transplant the memo to the new fake tensor but it seemed hopeless, so I just let the fresh symbol ride, allocating a new unbacked symbol. However, in one of our tests, we rely on knowing that the first nonzero call is equal to the second (memoized) nonzero call. The equality test looked pretty easy to discharge, so I just went ahead and added a deferred runtime assert to this effect and it worked.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124394
Approved by: https://github.com/jansel
ghstack dependencies: #124310, #124314, #124316
This is a partial revert of https://github.com/pytorch/pytorch/pull/124059
Like in #124297, profiling has revealed that testing equality on *every* output is kind of expensive. So we only test equality when we know there is an unbacked binding. This is the same playbook as the previous PR, just on FakeTensorProp instead of PropagateUnbackedSymInts. Note that we also need to populate `unbacked_bindings` in proxy_tensor.py, since we're generating an entirely new graph in that case.
We now have enough propagation that we're able to trigger a bug related to divisibility replacement. In https://github.com/pytorch/pytorch/pull/113165 we allowed to replace `u0` with `u1 * c` for some constant c, when we have determined that u0 is divisible by c. However, where does the binding for u1 come from? What we will have in practice is that there is some node that is supposed to have bound u1, but which actually is getting a `u1 * c` in its output. So, to get u1, we must divide out c. Fortunately, under the divisibility condition, this is always possible (but remember, we must test divisibility at runtime!)
Because we have tightened up asserts, it is now an error to allocate unbacked SymInts and then fail to track them under unbacked_bindings. In torch/_dynamo/eval_frame.py and torch/_functorch/_aot_autograd/collect_metadata_analysis.py there are examples of benign cases where we repropagated fake tensors but then immediately threw away the results. In these cases, it's not appropriate to rebind, since we're still using the old FX graph that has all of the old symbols. So we just manually clear it. It is possible that other cases will need to be updated, so this PR is "risky" from the perspective of hitting fbcode.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124310
Approved by: https://github.com/lezcano
Summary:
Adding function to log additional debug information before killing the expired watchdog timers.
Additional information like stack trace can be added in the debug function using worker process IDs from expired timers.
Test Plan: buck test mode/opt caffe2/test/distributed/elastic/timer:file_based_timer_test
Differential Revision: D56044153
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123883
Approved by: https://github.com/kurman
MTIA device has its own Module in PyTorch now.
torch.mtia has following APIs similar to other backends. The lazy_init is also supported.
```
__all__ = [
"init",
"is_available",
"synchronize",
"device_count",
"current_device",
"current_stream",
"default_stream",
"set_stream",
"stream",
"device",
]
```
------------
For device management. We expand AccleratorHooksInterface to support generic device management and it can be used in both C++ and PyThon.
```
def _accelerator_hooks_device_count() -> _int: ...
def _accelerator_hooks_set_current_device(device_index: _int) -> None: ...
def _accelerator_hooks_get_current_device() -> _int : ...
def _accelerator_hooks_exchange_device(device_index: _int) -> _int : ...
def _accelerator_hooks_maybe_exchange_device(device_index: _int) -> _int : ...
```
---------
Adding get_device_module API to retrieve device modules for different device types.
```
def get_device_module(device: Optional[Union[torch.device, str]] = None)
```
---------
Differential Revision: [D56443356](https://our.internmc.facebook.com/intern/diff/D56443356)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123612
Approved by: https://github.com/albanD
ghstack dependencies: #123611
This subsumes https://github.com/pytorch/pytorch/pull/124069
In the original PR, my idea was that when we run PropagateUnbackedSymInts, we check that the sizes before and after are exactly the same. This ended up turning up lots of bugs that I didn't feel like fixing. Separately, Ivan let me know that this pass was quite expensive in terms of compile time, since we spent a lot of time thinking about the equalities.
To kill two birds with one stone, we now only check for equality precisely when an unbacked SymInt was bound (thanks to the previous PR in this stack, we now have this information). Specifically, we look to see if `meta["unbacked_bindings"]` is set on the old node, and if it is, we assert the old value is equal to the new value from the repropagation. Note that the pytree key is used to actually extract the new value from the example value, as it may be nested inside an, e.g., tensor size.
We do something a bit naughty at the end: we use `defer_runtime_assert` to actually teach ShapeEnv about the equality. This is implementationally equivalent to what we used to do, but we're going to change this later soon.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124297
Approved by: https://github.com/lezcano
ghstack dependencies: #124290
The important comment:
```
# Whenever we allocate a fresh unbacked Symbol, we add it to this
# pending list. Unbacked symbol allocation can occur at unpredictable
# points during meta tensor propagation, but at some point, the we
# have to know what the binding site for an unbacked symbol is, and
# this is computed when we actually place the node in the graph. The
# important thing is that we always actually handle every unaccounted
# for unbacked symbol, so this list helps us keep track of them and
# then make sure they are all accounted for.
#
# We could potentially give rise to errors earlier by lexically
# scoping when we do propagation, and only allowing unbacked symbols
# to be allocated at this point in time. However this is inconvenient
# to do in Dynamo, because fake tensor propagation is far from when we
# analyze binding sites (set_example_value), so we do it in a more
# mutatey way.
#
# NB: fresh unbacked symbols NEVER get substitutions applied to them,
# they are binding sites!
```
The compute_unbacked_bindings is the other half of the equation: the thing that actually consumes the pending_fresh_unbacked_symbols and does something with them. Important comment:
```
After having run fake tensor propagation and producing example_value
result, traverse example_value looking for freshly bound unbacked
symbols and record their paths for later. It is an error if
we have allocated an unbacked SymInt but it cannot be found in
example_value. (NB: this means if you have a multi-output
function, you must call this on the tuple of tensor output, you
cannot wait!)
```
For example, if I return a tensor with size `[u0, u1]`, and u1 is a fresh unbacked SymInt, then I'll have `{u1: KeyPath(".size(1)")}`, telling me I can get u1 by running `size(1)` on the result of this node. u0 is not fresh (it probably flowed in as an argument), so I don't generate a binding for it.
I eventually intend to propagate this information all the way to Inductor lowering, where extra metadata about unbacked symbol binding will be canonically used for codegen, instead of trying to infer it from defs/uses.
Signed-off-by: Edward Z. Yang <ezyang@meta.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124290
Approved by: https://github.com/lezcano
This PR:
- exposes torch.testing._internal.optests.opcheck as
torch.library.opcheck
- Adds support for CustomOpDef (aka functions decorated with
torch.library.custom_op) to opcheck.
Test Plan:
- Updated tests
- We validated opcheck's design internally.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124496
Approved by: https://github.com/williamwen42
Following the example of PyTorch supporting a preferred Linalg library (cusolver or magma), this PR introduces a preferred blas library selector of either cublas or cublaslt for CUDA and hipblas or hipblaslt for ROCm via normal hipification of sources.
The default blas implementation remains cublas or hipblas. cublaslt or hipblaslt can be enabled using environment variable TORCH_BLAS_PREFER_CUBLASLT=1 (or TORCH_BLAS_PREFER_HIPBLASLT=1 as an alias) or by calling `torch.backends.cuda.preferred_blas_library(backend="cublaslt")` or as an alias `backend="hipblaslt"`.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122106
Approved by: https://github.com/lezcano
