Summary:
We add states in the constant folding process for AOTInductor.
Basically, there's 3 states, which is
(1) None: The state when no constants are loaded and uninitialized.
(2) Initialized: The state when constants are loaded, but not yet
folded.
(3) Folded: The state where the model is fully ready with folded
constants.
Note that even if constant folding is not enabled, we still only run
when state is FOLDED, this is okay because without constant folding, the
transition from INITIALIZED to FOLDED is just a pass-throught.
Test Plan:
python test/inductor/test_aot_inductor.py -k test_constant_folding_with_update
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D73002538](https://our.internmc.facebook.com/intern/diff/D73002538)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/151273
Approved by: https://github.com/jingsh, https://github.com/desertfire
This PR implements _allgather_base, reduce_scatter, and _reduce_scatter_base in the MPI backend (ProcessGroupMPI), enabling support for Fully Sharded Data Parallel (FSDP) in environments that use MPI for distributed communication.
### Context
As noted in https://github.com/pytorch/pytorch/issues/85628, FSDP currently supports only the NCCL backend. Due to this limitation, FSDP cannot run on legacy HPC environments or clusters that rely on MPI.
By implementing just these three collective operations, we can enable FSDP to work with the MPI backend. These collectives are implemented in a similar manner to existing operations such as allgather.
### Testing
We validated this PR using pytorch/build/bin/ProcessGroupMPITest with OpenMPI, and all tests passed successfully.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150162
Approved by: https://github.com/H-Huang
While fixing the memory leak in https://github.com/pytorch/pytorch/pull/145757, we accidentally close the socket for the case when nread == 0 and thought it is the case when connection is closed. This is not true. According to libuv doc: https://docs.libuv.org/en/v1.x/stream.html#c.uv_read_cb.
> nread might be 0, which does not indicate an error or EOF. This is equivalent to EAGAIN or EWOULDBLOCK under read(2).
We found this bug when debugging a broken pipe issue when users first call a set and then wait for all keys right afterwards on 128 ranks. This might also cause other broken pipe issues we have seen in the prod jobs recently.
Added a unit test to test this case.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150987
Approved by: https://github.com/d4l3k, https://github.com/XilunWu
Summary:
We add the functionality to allow users to directly pass in a at::Tensor
into AOTInductor, that would be used as the constant.
This user managed buffer skips the copying step in AOTInductor, and let
users to directly manage the memory usage themselve.
Test Plan:
LD_LIBRARY_PATH=/data/users/$USER/pytorch/build/lib
/data/users/$USER/pytorch/build/bin/test_aoti_inference
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D72589514](https://our.internmc.facebook.com/intern/diff/D72589514)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150276
Approved by: https://github.com/chenyang78, https://github.com/desertfire
Summary:
Profiler side of memory snapshot.
1. Add API to actually do snapshot when client interface is called
2. Add ifdefs to builds so that kineto hooks snapshot correctly.
Design Philosophy: There is one interesting part of this implementation and it is during export. For export we are callign the python impl of the export rather than CPP even though we are already in CPP. This is because it is better to simply have one path of export rather than 2. Personally, I want there to be parity between auto-trace and on-demand so it if we can limit the side paths then we will have an easier time maintaining this relationship
Test Plan: {F1976563426}
Reviewed By: sanrise
Differential Revision: D70733247
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150559
Approved by: https://github.com/sanrise
operations
Summary:
Fix the test for memory tracking. This PR does:
(1) Add tracking before and after for all memory-related operations.
Make sure the operation do indeed captures memory both in CUDA and
torch's CUDACachAllocator Make sure the operation do indeed captures
consumed memory both in CUDA and torch's CUDACachAllocator.
(2) Keep track of memory being reserved by CUDACacheAllocator in
torch and it's relationship with global CUDA memory consumption.
Test Plan:
This PR is adding tests.
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150269
Approved by: https://github.com/jingsh, https://github.com/chenyang78, https://github.com/desertfire
Relanding #148590 due to merge conflict.
This PR has multiple changes to `ProcessGroupNCCL` (which unfortunately are related):
1. When async_op=False, we directly launch the collective on "current" stream, instead of a trampoline stream and join back.
- Resolves#147729
- Resolves#146881
- Also saves two event syncs (which have overhead in case of HIP) and one pybind when we call `work.wait()` in distributed_c10d.py on behalf of user.
2. Entirely remove `record_stream` and use CPU-side stashing for managing tensor lifetime against recycling.
- Resolves#147168
3. Remove tensor life management when async_op=False; only use it when async_op=True.
4. To guard against user not calling `work.wait()`, we ask watchdog to unstash tensors after detecting completion of collectives, to prevent us from holding reference to tensors forever. This is a safety net, rather than a service guarantee, see discussion [here](https://github.com/pytorch/pytorch/issues/147168#issuecomment-2660142460).
5. Profile in async_op=False mode would look different -- collective kernels would show up in the same line and compute kernels.
Joint work with @cenzhaometa who wants to remove the event sync overhead.
Squashed contents:
* [ptd][nccl] use current-stream as nccl-stream under async=False mode (#147820)
PTD current workflow:
- PTD creates its own dedicated `ncclStream` for comm operation
- it will first add a dependency on current-stream (typically the compute stream) to ensure tensors are ready before invoking collective
such stream synchronization become expensive in Inference world (cpu overhead: 70us vs GPU kernel time: 160us).
This diff:
- async=False [default], will use current-stream as nccl-stream and avoid the stream-sync overhead
- async=True, will retain existing logic: create new nccl-stream, let it wait on current-stream to ensure tensors are ready
- pass down async from c10d down to NCCL-PG
this helps shave off 50% CPU overhead **(70us -> 35us)**, which reduce total CPU/GPU from **230us to 195us by 15%**
* [PGNCCL] Make avoid-record-stream default
* [c10d] Add asyncOp argument to Ops
* Change python side wait
* Pass asyncOp at ProcessGroup level
* Watchdog unstashing tensors as a safety net
* Stash tensors for reduce_scatter_v and all_gather_v
Pull Request approved: https://github.com/pytorch/pytorch/pull/149753
* [c10d] Move unstashing from watchdog to main thread
Pull Request approved: https://github.com/pytorch/pytorch/pull/150079
* [PGNCCL][BE] Merge mutex into TensorShelf for encapsulation
Pull Request approved: https://github.com/pytorch/pytorch/pull/150130
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150398
Approved by: https://github.com/atalman
Summary: Add extract_constant_map that allows users to inspect the constants being used by AOTInductor
Test Plan:
`python test/inductor/test_aot_inductor.py -k extract_constants_map`
`LD_LIBRARY_PATH=/data/users/$USER/pytorch/build/lib /data/users/$USER/pytorch/build/bin/test_aoti_inference`
Differential Revision: D72020400
Pull Request resolved: https://github.com/pytorch/pytorch/pull/150163
Approved by: https://github.com/chenyang78
internally.
Summary:
This diff allows freeing the usage of folded constants that's created by
AOTInductor through CUDACachingAllocator instead of the constant blob
from cudaMalloc directly.
Test Plan:
LD_LIBRARY_PATH=/data/users/$USER/pytorch/build/lib
/home/$USER/local/pytorch/build/bin/test_aoti_inference
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149825
Approved by: https://github.com/chenyang78, https://github.com/desertfire, https://github.com/jingsh
Summary:
We might free the active buffer if we free the buffer twice.
Test Plan:
```
LD_LIBRARY_PATH=/data/users/$USER/pytorch/build/lib
/home/$USER/local/pytorch/build/bin/test_aoti_inference
```
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149810
Approved by: https://github.com/chenyang78
Similar to #140425, we are making the implementation usable via header-only code sharing.
Review note: #62546 by @yanbing-j removed expm1 usage from this path. I don't know why and expm1 should be more efficient, so I've put it back. Please let me know if there is a good reason I shouldn't.
Testing: existing correctness tests should cover.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149673
Approved by: https://github.com/cyyever, https://github.com/Skylion007
Summary:
We need to properly fakify torchbind objects, including the ones in graph module attributes, so the resgitered fake implementation works properly.
- _fakify_script_objects in `compile_fx`
- Allow fake torchbind objects in `torchbind_constants`
Remove `node.meta["unbacked_bindings"]` for `aot_compile` in `compile_fx`. Otherwise `ShapeProp` will fail when trying to resolve the `unbacked_bindings` of `with_effect` tokens.
Update `sigrid_transforms_test` to use the latest `torch._inductor.aot_compile` API.
Add a test for `Fakify torchbind objects in compile_fx and add tests for SigridTransformsInstanceTorchBind` in `e2e_test`.
Test Plan:
```
buck run //caffe2/torch/fb/sparsenn:sigrid_test -- -r test_transform_torch_bind
buck run //sigmoid/inference/test:e2e_test_cpu -- -r SigridTransforms
buck2 run mode/dev-nosan sigmoid/inference/ts_migration:pt2i_readiness_main -- --model_id 545017754 --test_suite ads_all --mode test_preproc
```
Differential Revision: D70013257
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149529
Approved by: https://github.com/angelayi
The default value for `run_single_threaded` was wrongly specified in the .cpp file instead of the header, breaking C++-side instantiation of `AOTIModelPackageLoader` with no arguments. This PR fixes this and adds a test for the use case of running with `AOTIModelPackageLoader` instead of `AOTIModelContainerRunner` on the C++ side.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/149082
Approved by: https://github.com/desertfire
This PR has multiple changes to `ProcessGroupNCCL` (which unfortunately are related):
1. When async_op=False, we directly launch the collective on "current" stream, instead of a trampoline stream and join back.
- Resolves#147729
- Resolves#146881
- Also saves two event syncs (which have overhead in case of HIP) and one pybind when we call `work.wait()` in distributed_c10d.py on behalf of user.
2. Entirely remove `record_stream` and use CPU-side stashing for managing tensor lifetime against recycling.
- Resolves#147168
3. Remove tensor life management when async_op=False; only use it when async_op=True.
4. To guard against user not calling `work.wait()`, we ask watchdog to unstash tensors after detecting completion of collectives, to prevent us from holding reference to tensors forever. This is a safety net, rather than a service guarantee, see discussion [here](https://github.com/pytorch/pytorch/issues/147168#issuecomment-2660142460).
5. Profile in async_op=False mode would look different -- collective kernels would show up in the same line and compute kernels.
Joint work with @cenzhaometa who wants to remove the event sync overhead.
Cc: @ngimel @awgu @Aidyn-A @skyw @wconstab @leonardo0lyj
Differential Revision: [D70937982](https://our.internmc.facebook.com/intern/diff/D70937982)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148590
Approved by: https://github.com/eqy, https://github.com/Aidyn-A, https://github.com/fduwjj
This PR has multiple changes to `ProcessGroupNCCL` (which unfortunately are related):
1. When async_op=False, we directly launch the collective on "current" stream, instead of a trampoline stream and join back.
- Resolves#147729
- Resolves#146881
- Also saves two event syncs (which have overhead in case of HIP) and one pybind when we call `work.wait()` in distributed_c10d.py on behalf of user.
2. Entirely remove `record_stream` and use CPU-side stashing for managing tensor lifetime against recycling.
- Resolves#147168
3. Remove tensor life management when async_op=False; only use it when async_op=True.
4. To guard against user not calling `work.wait()`, we ask watchdog to unstash tensors after detecting completion of collectives, to prevent us from holding reference to tensors forever. This is a safety net, rather than a service guarantee, see discussion [here](https://github.com/pytorch/pytorch/issues/147168#issuecomment-2660142460).
5. Profile in async_op=False mode would look different -- collective kernels would show up in the same line and compute kernels.
Joint work with @cenzhaometa who wants to remove the event sync overhead.
Cc: @ngimel @awgu @Aidyn-A @skyw @wconstab @leonardo0lyj
Differential Revision: [D70835197](https://our.internmc.facebook.com/intern/diff/D70835197)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148590
Approved by: https://github.com/eqy, https://github.com/Aidyn-A, https://github.com/fduwjj
This PR has multiple changes to `ProcessGroupNCCL` (which unfortunately are related):
1. When async_op=False, we directly launch the collective on "current" stream, instead of a trampoline stream and join back.
- Resolves#147729
- Resolves#146881
- Also saves two event syncs (which have overhead in case of HIP) and one pybind when we call `work.wait()` in distributed_c10d.py on behalf of user.
2. Entirely remove `record_stream` and use CPU-side stashing for managing tensor lifetime against recycling.
- Resolves#147168
3. Remove tensor life management when async_op=False; only use it when async_op=True.
4. To guard against user not calling `work.wait()`, we ask watchdog to unstash tensors after detecting completion of collectives, to prevent us from holding reference to tensors forever. This is a safety net, rather than a service guarantee, see discussion [here](https://github.com/pytorch/pytorch/issues/147168#issuecomment-2660142460).
5. Profile in async_op=False mode would look different -- collective kernels would show up in the same line and compute kernels.
Joint work with @cenzhaometa who wants to remove the event sync overhead.
Cc: @ngimel @awgu @Aidyn-A @skyw @wconstab @leonardo0lyj
Differential Revision: [D70835197](https://our.internmc.facebook.com/intern/diff/D70835197)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/148590
Approved by: https://github.com/eqy, https://github.com/Aidyn-A, https://github.com/fduwjj
This PR introduces two new methods to the LazyGraphExecutor class:
- ClearComputationCache(): Allows clearing the entire computation cache.
- RemoveFromComputationCache(hash): Enables removal of specific cache entries based on their hash.
The main objective is to expose cache management functionality for debugging cache hits and misses across different computations. For instance:
- Reset the cache state in tests, allowing reuse of the same computation client to evaluate cache logic consistently.
- Selectively remove cache entries to analyze the impact on subsequent computations.
- Improve observability into the cache behavior, aiding in the investigation of cache-related issues or optimizations.
On the XLA lazy graph executor, we want to run a series of tests that modify some parts of the HLO module proto of the computation, and we need a means to ensure that the hash is agnostic to some elements (OpMetadata in the XLA proto data). Hence, it would be easy to parameterize the test, clear the cache and validate that the resulting hash is the same between runs. Otherwise, we'd need to hardcode the resulting serialized hash.
Simultaneously, **another motivation**, is that users could also clear some computation hashes for an added flexibility in their applications, by introducing their own custom strategies for maintaining the cache (without relying on the default LRU).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144489
Approved by: https://github.com/wconstab
Summary:
This PR is basically a replacement of
https://github.com/pytorch/pytorch/pull/140087, which caused some perf
drop due to frequent TCPStore check in watchdog thread. The fix is to move the
tcpstore check in monitoring thread
If unhealthy, the user should be able to get the type of errors, e.g.,
timeout,nccl error or remote error.
This API is applied to PG level, compared to the
work.get_future_result() API which is applied to Work Level.
Error detection at PG level is much more convenient for users to handle
the PG failure as a whole, e.g, restarting the PG.
Error handling at the work level is still useful for users to attach
work specific context and debug the RC of the specific failing
work/collective
Note it is critical for all ranks in the PG to be notified about an
error as soon as it occurs, so we introduce an errorType of
REMOTE_ERROR, which is 'broadcasted' from a src rank (which detects a
local error) to all other ranks in the PG, the broadcast is done through
TCPStore currently
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144498
Approved by: https://github.com/kwen2501
Reported in https://github.com/pytorch/pytorch/issues/143470, we have a dangling references in `CudaEventCache`. So we want to fix it.
1. We add a unit test to repro the issue mentioned in the issue.
2. Instead of converting variables to shared pointers as suggested in the issue, we then make the cache itself a shared pointer. So if the thread creates the cache dies before all events get recycled, the cache is still there until the last CudaEvent get deleted. (thanks for the suggestion from @kwen2501 )
Pull Request resolved: https://github.com/pytorch/pytorch/pull/144496
Approved by: https://github.com/kwen2501
Summary:
LLVM-15 has a warning `-Wunused-variable` which we treat as an error because it's so often diagnostic of a code issue. Unused variables can compromise readability or, worse, performance.
This diff either (a) removes an unused variable and, possibly, it's associated code or (b) qualifies the variable with `[[maybe_unused]]`.
- If you approve of this diff, please use the "Accept & Ship" button :-)
Test Plan: Sandcastle
Reviewed By: palmje
Pull Request resolved: https://github.com/pytorch/pytorch/pull/143517
Approved by: https://github.com/mhorowitz
Summary: Provide a standalone path to compile and run a ExportedProgram in C.
Test Plan:
(1) Generate a compiled model from ExportedProgram
```
python generate_lowered_cpu.py --input-path /tmp/$USER/ep.pt --output-path /tmp/$USER/final.pt
```
(2) Compile a standalone test runner
```
TORCH_ROOT_DIR=/data/users/$USER/pytorch sh standalone_compile.sh standalone_test.cpp standalone_test.out
```
(3) Run test for the compiled model in step (1)
```
LD_LIBRARY_PATH=/data/users/$USER/pytorch/build/lib ./standalone_test.out /tmp/$USER/final.pt
```
Differential Revision: D66872380
Pull Request resolved: https://github.com/pytorch/pytorch/pull/142327
Approved by: https://github.com/hl475
Using EC2 G6 instance, based on NVIDIA L4, added to scale config in https://github.com/pytorch/test-infra/pull/5376
To enable more balanced sharding, had to push 148ae19935
Added `@xfailIfSM89` to the following tests:
- test_fp8_pattern_2
- test_original_aten_preserved_split_addmm
- test_sparse_semi_structured_scaled_mm
- test_sparse_semi_structured_scaled_mm_fp8
- test_sparse_fp8fp8_mm
Increased tolerance to 2e-4 for `RNNTest.BidirectionalMultilayerGRU_CPU_vs_CUDA`
Skipped following inductor tests (that either flaky OOMs or timeouts):
- test_reduction_fn_std_float64
- test_reduction_fn_var_mean_float64
- test_multi_output_unbacked_custom_op
Pull Request resolved: https://github.com/pytorch/pytorch/pull/140305
Approved by: https://github.com/wdvr, https://github.com/ZainRizvi
We added `CudaEventCache` in https://github.com/pytorch/pytorch/pull/133727 and this is a feature which tries to reuse CudaEvent so that we don't call destroy of CudaEvent which causes hang in the past. We had a bunch of tests and testing on TorchTitan and internal workload already. So far no errors or crash are found at the moment so we decide to roll out to all OSS users. For internal workload, this PR would not affect it because of some internal gating.
Also we observed some multi-device use cases in OSS, so that we want to bring back multi-device support originally proposed in https://github.com/pytorch/pytorch/pull/122732/files.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/140975
Approved by: https://github.com/eqy, https://github.com/kwen2501
Summary: Tighten the AOTIModelContainerRunner::run interface to take a const vector of at::Tensor, which 1) makes it clear that the runner will not modify the input tensor vector; 2) runner will be able to take a temp vector of tensors as the input.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/139955
Approved by: https://github.com/chenyang78
- Refactored traceback code into `work.printTraceback()`. cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o @shuqiangzhang
- Refactored desync debug code into `class DesyncDebugger`.
- Moved occurrences of `futureWorkResult_->markCompleted` into `checkAndSetException` and `checkTimeout`, respectively. cc @shuqiangzhang
- Modularized dump signal broadcast code into `ProcessGroupNCCL::broadcastDumpSignal`. cc @fduwjj @c-p-i-o
Pull Request resolved: https://github.com/pytorch/pytorch/pull/139834
Approved by: https://github.com/shuqiangzhang
This change fixes the RUNPATH of installed c++ tests so that the linker can find the shared libraries they depend on.
For example, currently:
```bash
venv/lib/python3.10/site-packages/torch $ ./bin/test_lazy
./bin/test_lazy: error while loading shared libraries: libtorch.so: cannot open shared object file: No such file or directory
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136627
Approved by: https://github.com/malfet
Summary:
Blocking wait mode is not widely used, probably useful in debugging.
in blockingWait mode, we don't need to enable the watchdog thread to
check the timeout or nccl error because the main thread would throw an
exception if error happens and it is obvious to user which work fails
and its user's responsibility to handle the exception.
Test Plan:
CI
Pull Request resolved: https://github.com/pytorch/pytorch/pull/138001
Approved by: https://github.com/fduwjj, https://github.com/c-p-i-o
ghstack dependencies: #137799
synchronization
Summary:
Barrier is essentially intended to block CPU thread (instead of GPU
streams). Before we used 2 stream synchronizations (1. current stream
blocked by nccl stream end event, 2. CPU thread blocked on current
stream). This is unnecessary as we already have CPU thread blocking
logic in wait(). Also, adding barrier specific code block in the general
GPU synchronize() API is intrusive and confusing.
This PR cleans this.
Test Plan:
CI
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/137516
Approved by: https://github.com/fduwjj, https://github.com/kwen2501
Fixes#127920
This commit addresses a build failure occurring with GCC 12 and above due to the -Werror=nonnull flag. The error manifests in the test_api target.
**Issue:**
When building with GCC 12+, the following error occurs:
```
error: argument 1 null where non-null expected [-Werror=nonnull]
431 | __builtin_memmove(__result, __first, sizeof(_Tp) * _Num);
| ~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
This change ensures that:
1. The flag is only added for GCC 12 or higher
2. The flag is only added if it's supported by the compiler
3. The flag is added specifically to the test_api target, not globally
By disabling this specific error, we allow the build to proceed while maintaining other compiler warnings.
**Test Plan:**
- Verified successful build with GCC 12 and above
- Ensured no regression in builds with earlier GCC versions and other compilers
Pull Request resolved: https://github.com/pytorch/pytorch/pull/137092
Approved by: https://github.com/malfet
This adds logs if we can't acquire locks in NCCLUtils and ProcessGroupNCCL for 30s.
This is motivated by some deadlocks were seeing and it's unclear if it's in NCCL or on the PyTorch side of things.
This required replacing most `std::mutex` with `std::timed_mutex` and `std::condition_variable_any` as appropriate.
Test plan:
existing CI for regressions
will add unit tests on `C10D_LOCK_GUARD`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134131
Approved by: https://github.com/c-p-i-o, https://github.com/fduwjj
Summary: Pass process group info into NcclWork
Test Plan: buck2 run mode/dev-nosan kineto/libkineto/fb/integration_tests:pytorch_execution_trace_integration_test
Differential Revision: D61677160
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134269
Approved by: https://github.com/wconstab
This adds logs if we can't acquire locks in NCCLUtils and ProcessGroupNCCL for 30s.
This is motivated by some deadlocks were seeing and it's unclear if it's in NCCL or on the PyTorch side of things.
This required replacing most `std::mutex` with `std::timed_mutex` and `std::condition_variable_any` as appropriate.
Test plan:
existing CI for regressions
will add unit tests on `C10D_LOCK_GUARD`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134131
Approved by: https://github.com/c-p-i-o, https://github.com/fduwjj
Summary:
Re-enable testHelperPrefix test that was erroneously disabled in CI.
Fixes#50701
Test Plan:
Test passes locally:
```
❯ ./TCPStoreTest --gtest_filter=TCPStoreTest.testHelperPrefix
Running main() from
/data/users/cpio/pytorch/third_party/googletest/googletest/src/gtest_main.cc
Note: Google Test filter = TCPStoreTest.testHelperPrefix
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from TCPStoreTest
[ RUN ] TCPStoreTest.testHelperPrefix
[W807 12:01:31.531576727 socket.cpp:462] [c10d] waitForInput: poll for
socket SocketImpl(fd=6, addr=[localhost]:37984,
remote=[localhost]:37171) returned 0, likely a timeout
[W807 12:01:31.531663710 socket.cpp:487] [c10d] waitForInput: socket
SocketImpl(fd=6, addr=[localhost]:37984, remote=[localhost]:37171) timed
out after 100ms
[ OK ] TCPStoreTest.testHelperPrefix (314 ms)
[----------] 1 test from TCPStoreTest (314 ms total)
[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (314 ms total)
[ PASSED ] 1 test.
╭─ ~/local/pytorch/build/bin main *1 +1 ···················· ✔
/home/cpio/local/a/pytorch-env cpio@devgpu011 ─╮
╰─
```
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132916
Approved by: https://github.com/Skylion007
Partially fixes#122980
- change cpp type mapping for complex64 to std::complex<float>
- add `aoti_torch_item_complex64` and `aoti_torch_scalar_to_tensor_complex64`.
- add `expensiveCopyToTensor()` to convert `ArrayRefTensor<T>` type to `AtenTensorHandle` type.
- if we want to fully fix#122980, we still need to let ArrayRef and MiniArrayRef to consider underlying storage number of elements. See more details in https://github.com/pytorch/pytorch/pull/132347 (#132347 broke some internal tests, so we need more work before landing it).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132810
Approved by: https://github.com/desertfire
Fixes#10536
Reattempt of #61467. Thank you so much to @mskoh52 for your excellent work!
As I was trying to create a more efficient LLM data collator, I realized that `pad_sequence` only supports right padding, even though left padding is a very common format for LLMs, like Llama and Mistral.
The proposed alternative implementation was to use multiple flips, which tends to be 1.5x-2x slower. Instead we can add a [`padding_side` parameter as there is for for Hugging Face tokenizers](9d6c0641c4/src/transformers/tokenization_utils_base.py (L1565)), which requires only a very small change in the C++ code.
Here are the benchmarks of the new implementation!
`float32`:
`bool`:
Code:
```python
from __future__ import annotations
import random
import time
from typing import Literal
import numpy as np
import torch
def pad_sequence_with_flips(
sequences: list[torch.Tensor],
batch_first: bool = False,
padding_value: int | float | bool = 0.0,
padding_side: Literal["left", "right"] | str = "left",
) -> torch.Tensor:
if padding_side == 'right':
padded_sequence = torch._C._nn.pad_sequence([t.flatten() for t in sequences], batch_first=batch_first, padding_value=padding_value)
elif padding_side=='left':
padded_sequence = torch._C._nn.pad_sequence([t.flatten().flip(0) for t in sequences], batch_first=batch_first, padding_value=padding_value) # pyright: ignore[reportArgumentType]
padded_sequence = padded_sequence.flip(int(batch_first))
else:
raise ValueError(f"padding_side should be either 'right' or 'left', but got {padding_side}")
return padded_sequence
sequence_lengths: list[int] = []
flip_left_pad_times: list[float] = []
flip_left_pad_times_std: list[float] = []
left_pad_times: list[float] = []
left_pad_times_std: list[float] = []
RUNS_PER_LOOP: int = 100
for i in range(1, 7):
sequence_length = i * int(1e6) // 6
sequence_lengths.append(sequence_length)
sequences = [torch.randint(0, 2, (random.randint(1, sequence_length),), dtype=torch.bool) for _ in range(64)]
inner_left_pad_times: list[float] = []
inner_right_pad_times: list[float] = []
inner_flip_left_pad_times: list[float] = []
inner_flip_right_pad_times: list[float] = []
for _ in range(RUNS_PER_LOOP):
start = time.perf_counter()
torch._C._nn.pad_sequence(sequences, batch_first=True, padding_value=False, padding_side="left")
end = time.perf_counter()
inner_left_pad_times.append(end - start)
start = time.perf_counter()
pad_sequence_with_flips(sequences, batch_first=True, padding_value=False, padding_side="left")
end = time.perf_counter()
inner_flip_left_pad_times.append(end - start)
left_pad_times.append(sum(inner_left_pad_times) / len(inner_left_pad_times))
left_pad_times_std.append(np.std(inner_left_pad_times))
flip_left_pad_times.append(sum(inner_flip_left_pad_times) / len(inner_flip_left_pad_times))
flip_left_pad_times_std.append(np.std(inner_flip_left_pad_times))
print(f"Sequence Length: {sequence_length}, Left Pad Time: {left_pad_times[-1]}, Left with Flips Pad Time: {flip_left_pad_times[-1]}")
import matplotlib.pyplot as plt
plt.plot(sequence_lengths, left_pad_times, label="new pad_sequence left")
plt.scatter(sequence_lengths, left_pad_times)
plt.errorbar(sequence_lengths, left_pad_times, yerr=left_pad_times_std, linestyle='None', marker='^')
plt.plot(sequence_lengths, flip_left_pad_times, label="old pad_sequence left (2 flips)")
plt.scatter(sequence_lengths, flip_left_pad_times)
plt.errorbar(sequence_lengths, flip_left_pad_times, yerr=flip_left_pad_times_std, linestyle='None', marker='^')
plt.xlabel("Sequence Length")
plt.ylabel("Time (s)")
plt.legend(loc="upper right")
# Sequence Length: 166666, Left Pad Time: 0.06147645162009212, Left with Flips Pad Time: 0.09842291727001794
# Sequence Length: 333333, Left Pad Time: 0.08933195920990329, Left with Flips Pad Time: 0.15597836187991562
# Sequence Length: 500000, Left Pad Time: 0.08863158334006585, Left with Flips Pad Time: 0.15224887342999863
# Sequence Length: 666666, Left Pad Time: 0.10524682551997103, Left with Flips Pad Time: 0.18177212480995877
# Sequence Length: 833333, Left Pad Time: 0.11801802741003485, Left with Flips Pad Time: 0.20821274195001024
# Sequence Length: 1000000, Left Pad Time: 0.131894061660023, Left with Flips Pad Time: 0.23223503091008751
```
Co-authored-by: mskoh52 <mskoh52@users.noreply.github.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131884
Approved by: https://github.com/ezyang
Sometimes, it could be difficult to write a fake class e.g. when the original implementation is using some third-party libraries or users are certain that the class is safe to trace with the real object.
This PR allows user to specify their intention by implementing a "safe_to_trace_with_real_obj" method on their script class.
Test Plan:
`pytest test/export/test_torchbind.py -k safe`
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129586
Approved by: https://github.com/zou3519
As titled. Previously, __obj_flatten__ can run in a fake tensor mode, e.g. in process_input of aot_autograd, which is surrounded by a fake tensor mode. This causes the tensor ops inside __obj_flatten__ to run under fake tensor mode. However, tensors inside of script obejct are real tensors, this causes the fake tensor mode to error out saying that we need to first fakify fall the tensors (because allow_non_fake_inputs is set to True).
In this PR, we disable all the dispatch modes when running to_fake_obj.
Note that, the output of `__obj_flatten__` will be fakified and filled inside of the corresponding FakeScriptObject. So during traicng, we'll be using FakeScriptObject that has fake tensor contents.
Test Plan:
Add a new test: pytest test/export/test_torchbind.py -k test_compile_tensor_op_in_tensor_flatten
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129605
Approved by: https://github.com/angelayi
We've been facing issues where TCPStore can successfully connect but then fail in the validate() function due to resets from listen backlog queue overflow when combined with reset enabled as well as long init times.
This PR does a few things:
* Retry that connect and validate up to the specified timeout.
* Use exponential backoff for the retry logic with jitter instead of a fixed 1s sleep.
* Eliminate the `sleep(std::chrono::milliseconds(numWorkers))` on init which can add significant delays to startup. This is no longer necessary per @XilunWu https://github.com/pytorch/pytorch/pull/116141
Test plan:
```
python test/distributed/test_store.py -v
./build/bin/BackoffTest
```
Will do internal testing with some large scale jobs to ensure TCPStore works correctly.
At 4k scale: 4x improvement
```
tristanr@devvm4382 ~/pt_tests [SIGABRT]> time TORCH_SHOW_CPP_STACKTRACES=1 python tcpstore_large_test.py (pytorch-3.10)
started 0
init 0
set 0
joined all
________________________________________________________
Executed in 1.98 secs fish external
usr time 0.93 secs 91.00 micros 0.93 secs
sys time 1.98 secs 954.00 micros 1.97 secs
tristanr@devvm4382 ~/pt_tests> conda activate torchdrive-3.10 (pytorch-3.10)
tristanr@devvm4382 ~/pt_tests> time TORCH_SHOW_CPP_STACKTRACES=1 python tcpstore_large_test.py (torchdrive-3.10)
started 0
init 0
set 0
joined all
________________________________________________________
Executed in 8.20 secs fish external
usr time 2.15 secs 0.00 micros 2.15 secs
sys time 2.76 secs 843.00 micros 2.76 secs
```
```py
import time
import os
import threading
from multiprocessing import Pool
WORLD_SIZE = 10000
import torch.distributed as dist
def run(rank):
should_log = rank % (WORLD_SIZE // 10) == 0
if should_log:
print(f"started {rank}")
store = dist.TCPStore(
host_name="devvm4382.nao0.facebook.com",
port=29500,
world_size=WORLD_SIZE,
is_master=rank == 0,
use_libuv=True,
)
if should_log:
print(f"init {rank}")
store.set(f"key{rank}", "1234")
if should_log:
print(f"set {rank}")
del store
def noop(rank):
pass
print("starting pool")
with Pool(WORLD_SIZE) as pool:
pool.map(noop, range(WORLD_SIZE), 1)
print("pool hot")
start = time.time()
pool.map(run, range(WORLD_SIZE), 1)
print("run finished", time.time()-start)
```
```
tristanr@devvm4382 ~/pt_tests> python tcpstore_large_test.py (pytorch-3.10)
starting pool
pool hot
started 0
[W624 16:58:09.086081750 TCPStore.cpp:343] [c10d] Starting store with 10000 workers but somaxconn is 4096.This might cause instability during bootstrap, consider increasing it.
started 1000
init 1000
set 1000
started 2000
init 2000
set 2000
started 3000
init 3000
set 3000
started 4000
init 4000
set 4000
started 5000
init 5000
set 5000
started 6000
init 6000
set 6000
started 7000
init 7000
set 7000
started 8000
init 8000
set 8000
started 9000
init 9000
set 9000
init 0
set 0
run finished 0.705092191696167
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129261
Approved by: https://github.com/rsdcastro, https://github.com/wconstab, https://github.com/kurman, https://github.com/XilunWu, https://github.com/c-p-i-o
This PR does two things:
1. it duplicates the fake script object because aot_export trace the program twice. The result of tracing in the first time would cause the tracing result of second time be wrong.
2. Also add a new test for methods that return constant outputs. Before the PR, there's is no meta["val"] for these nodes because fx won't track these constants. We still need to preserve these constant return operators in the graph because torchbind objects are stateful and deleting it would remove the implicit state mutation inside of the object.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128844
Approved by: https://github.com/angelayi
Adds `C10_UBSAN_ENABLED` macro and use it to disable `SymIntTest::Overflows` (fails under `signed-integer-overflow` UBSAN check).
Also cleans up UBSAN guard in `jit/test_misc.cpp` to use `C10_UBSAN_ENABLED` and the existing `C10_ASAN_ENABLED` instead of locally defining `HAS_ASANUBSAN`.
> NOTE: This should fix `SymIntTest::Overflows` failing under ubsan in fbcode too...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127967
Approved by: https://github.com/atalman, https://github.com/d4l3k, https://github.com/malfet
Summary:
Split out `seq_id` into `collective_seq_id` and `p2p_seq_id`. The main idea here is that collectives that go to all machines should have identical `collective_seq_id` and therefore it makes it easier to spot if one of machines isn't handling a collective operation.
Next, we can attempt to match up p2p operations to ensure that the sender(s)/receivers(s) are in sync.
Resolves issue: https://github.com/pytorch/pytorch/issues/125173
Test Plan:
Unit tests.
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125727
Approved by: https://github.com/zdevito
Copy of #126089, with some additional fixes & tests
Partial fix for #125635: previously, the deepcopy implementation would group together any tensors with any aliasing relationship and assign them to the same tensor. This was sort of good if you have two tensors `b = a.detach()`, because then if you deepcopy `list = [a, b]` to `list2 = list.deepcopy()`, then writes to `list2[0]` will also modify `list2[1]`. But for the most part, it's bad; (1) if you have `b = a.as_strided((4, 4), (16, 1), 16)`, then it'll make `b == a` in the deepcopied implementation, which is completely wrong; and (2) even if you have `b = a.detach()`, these are still initially two different tensors which become the same tensor after the old deepcopy implementation.
The new implementation only groups together tensors that have the same identity. This is a partial fix, but it's more reasonable. What changes:
* (becomes more correct): different views of the same base tensor will no longer all become equal after deepcopying
* (still kind of wrong): views won't actually alias each other after deepcopying.
* (arguably a minor regression): equivalent views of the same tensor will no longer be copied to the same tensor - so they won't alias.
BC breaking: C++ deepcopy interface changes from accepting `IValue::HashAliasedIValueMap memo` to accepting `IValue::HashIdentityIValueMap memo`. If there are objections, we can keep the old API. However, it seems likely that users generally won't try to deepcopy from C++.
Differential Revision: [D57406306](https://our.internmc.facebook.com/intern/diff/D57406306)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126126
Approved by: https://github.com/ezyang
This PR makes libtorch behave the same as PyTorch when loading optimizer state from archive. With PyTorch, options of parameter groups are loaded from the archive, which is missing currently in libtorch.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125215
Approved by: https://github.com/janeyx99
Summary:
We found that some dumps are missing when monitoring thread timeout.
This is likely due to multiple PGs could still dump the same records
at the same time. So we should allow only PG0 to actualy dump
Test Plan:
unit test
python test/run_test.py --cpp --verbose -i cpp/ProcessGroupNCCLErrorsTest
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125356
Approved by: https://github.com/c-p-i-o
This PR continues to clean clang-tidy warnings in torch/csrc/distributed/c10d, following #124701. In addition, libfmt dependency is added in CMake code to enable using it in the headers. The libfmt has to be added as private dependency to torch_cuda and torch_hip because they include torch/csrc/distributed/c10d/Utils.hpp which uses libfmt.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124987
Approved by: https://github.com/malfet
Summary: Now that we have reached nanosecond granularity, we can now remove the temporary guards that were previously required for nanosecond precision.
Test Plan: Regression should cover this change
Reviewed By: aaronenyeshi
Differential Revision: D56444570
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124734
Approved by: https://github.com/aaronenyeshi
Summary: In AOTInductor generated CPU model code, there can be direct references to some aten/c10 utility functions and data structures, e.g. at::vec and c10::Half. These are performance critical and thus it doesn't make sense to create C shim for them. Instead, we make sure they are implemented in a header-only way, and use this set of tests to guard future changes.
There are more header files to be updated, but we will do it in other followup PRs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123848
Approved by: https://github.com/jansel
ghstack dependencies: #123847
Summary:
Kineto traces use microsecond level granularity because of chrome tracing defaults to that precision. Fix by adding preprocessor flag to TARGETS and BUCK files. Also remove any unnecessary ns to us conversions made in the profiler itself.
This diff contains profiler changes only. Libkineto changes found in D54964435.
Test Plan:
Check JSON and chrome tracing to make sure values are as expected. Tracing with flags enabled should have ns precision. Tracings without flags should be same as master.
Zoomer: https://www.internalfb.com/intern/zoomer/?profiling_run_fbid=796886748550189
Ran key_averages() to make sure FunctionEvent code working as expected:
-- ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg Self CUDA Self CUDA % CUDA total CUDA time avg # of Calls
ProfilerStep* 0.74% 3.976ms 64.40% 346.613ms 69.323ms 0.000us 0.00% 61.710ms 12.342ms 5
Optimizer.zero_grad#SGD.zero_grad 0.76% 4.109ms 0.76% 4.109ms 821.743us 0.000us 0.00% 0.000us 0.000us 5
## forward ## 6.89% 37.057ms 27.19% 146.320ms 29.264ms 0.000us 0.00% 58.708ms 11.742ms 5
aten::conv2d 0.22% 1.176ms 7.74% 41.658ms 157.199us 0.000us 0.00% 27.550ms 103.962us 265
aten::convolution 0.79% 4.273ms 7.52% 40.482ms 152.762us 0.000us 0.00% 27.550ms 103.962us 265
aten::_convolution 0.69% 3.688ms 6.73% 36.209ms 136.637us 0.000us 0.00% 27.550ms 103.962us 265
aten::cudnn_convolution 6.04% 32.520ms 6.04% 32.520ms 122.719us 27.550ms 8.44% 27.550ms 103.962us 265
aten::add_ 2.42% 13.045ms 2.42% 13.045ms 30.694us 12.700ms 3.89% 12.700ms 29.882us 425
aten::batch_norm 0.19% 1.027ms 8.12% 43.717ms 164.971us 0.000us 0.00% 16.744ms 63.185us 265
aten::_batch_norm_impl_index 0.31% 1.646ms 7.93% 42.691ms 161.096us 0.000us 0.00% 16.744ms 63.185us 265
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------
Differential Revision: D55925068
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123650
Approved by: https://github.com/aaronenyeshi
Summary:
Kineto traces use microsecond level granularity because of chrome tracing defaults to that precision. Fix by adding preprocessor flag to TARGETS and BUCK files. Also remove any unnecessary ns to us conversions made in the profiler itself.
This diff contains profiler changes only. Libkineto changes found in D54964435.
Test Plan:
Check JSON and chrome tracing to make sure values are as expected. Tracing with flags enabled should have ns precision. Tracings without flags should be same as master.
Tracing with flags enabled: https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/dynocli/devvm2185.cco0.facebook.com/rank-0.Mar_18_14_37_22.4155151.pt.trace.json.gz&bucket=gpu_traces
Tracing without flags enabled: https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/dynocli/devvm2185.cco0.facebook.com/rank-0.Mar_18_14_39_15.4166047.pt.trace.json.gz&bucket=gpu_traces
Tracing on main: https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/dynocli/devvm2185.cco0.facebook.com/rank-0.Mar_18_14_42_43.4177559.pt.trace.json.gz&bucket=gpu_traces
Ran key_averages() to make sure FunctionEvent code working as expected:
-- ------------ ------------
Name Self CPU % Self CPU CPU total % CPU total CPU time avg Self CUDA Self CUDA % CUDA total CUDA time avg # of Calls
ProfilerStep* 0.74% 3.976ms 64.40% 346.613ms 69.323ms 0.000us 0.00% 61.710ms 12.342ms 5
Optimizer.zero_grad#SGD.zero_grad 0.76% 4.109ms 0.76% 4.109ms 821.743us 0.000us 0.00% 0.000us 0.000us 5
## forward ## 6.89% 37.057ms 27.19% 146.320ms 29.264ms 0.000us 0.00% 58.708ms 11.742ms 5
aten::conv2d 0.22% 1.176ms 7.74% 41.658ms 157.199us 0.000us 0.00% 27.550ms 103.962us 265
aten::convolution 0.79% 4.273ms 7.52% 40.482ms 152.762us 0.000us 0.00% 27.550ms 103.962us 265
aten::_convolution 0.69% 3.688ms 6.73% 36.209ms 136.637us 0.000us 0.00% 27.550ms 103.962us 265
aten::cudnn_convolution 6.04% 32.520ms 6.04% 32.520ms 122.719us 27.550ms 8.44% 27.550ms 103.962us 265
aten::add_ 2.42% 13.045ms 2.42% 13.045ms 30.694us 12.700ms 3.89% 12.700ms 29.882us 425
aten::batch_norm 0.19% 1.027ms 8.12% 43.717ms 164.971us 0.000us 0.00% 16.744ms 63.185us 265
aten::_batch_norm_impl_index 0.31% 1.646ms 7.93% 42.691ms 161.096us 0.000us 0.00% 16.744ms 63.185us 265
------------------------------------------------------- ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------ ------------
Differential Revision: D55087993
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122425
Approved by: https://github.com/aaronenyeshi
This PR updates the error message in autograd when an input tensor does not set to `require_grad`. The original message does not contain the index info, making users hard to debug.
The error message style consists with that on line 105-109.
Co-authored-by: Jeffrey Wan <soulitzer@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123154
Approved by: https://github.com/soulitzer
This PR only adds abstract class registration logic without touching existing tests so they still trace with real script object. The added tests are only for registration APIs and test error messages.
Our design is that the abstract implementation should be in Python. This is much better in terms of usability. But this also has implications for custom op that takes script object as input, which is detailed later in this stack.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122622
Approved by: https://github.com/zou3519
ghstack dependencies: #122619, #122620, #122621
Summary:
During tracing, some constants (tensor_constant{idx}) are being generated internally.
Those constants are neither parameters or buffers, and users have zero control on them.
To accomodate this, we should allow users not passing in those constants generated internally but still be able the constants in the model.
Test Plan:
Included in commit.
```
build/bin/test_aot_inductor
```
Reviewed By: zoranzhao
Differential Revision: D55354548
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122690
Approved by: https://github.com/khabinov
Summary:
During tracing, some constants (tensor_constant{idx}) are being generated internally.
Those constants are neither parameters or buffers, and users have zero control on them.
To accomodate this, we should allow users not passing in those constants generated internally but still be able the constants in the model.
Test Plan:
Included in commit.
```
build/bin/test_aot_inductor
```
Differential Revision: D55286634
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122562
Approved by: https://github.com/chenyang78, https://github.com/khabinov
Fix and test issues with both coalesced and individual send/recv ops
Considered an alternate approach and then ditched it
- alternate approach: #119757
- reason ditched: prefer recording individual collective events inside
coalescing region instead of just the event at the end of the region,
which also would not have tensor sizes or opnames without additional
state variables added
Another approach also ditched
- record events on workEnqueue instead of initWork
- reason ditched: too messy to get input/output shapes tagged on
recording when recording in workEnqueue. Adding the info onto the
Work obj would be possible, but adds to overhead of copying Works
which we do on every collective. We can get info off the input/output
tensors directly in initWork, but we don't want to keep refs to those
tensors alive while the work is Enqueued, so we'd have to specifically
copy size lists or something.
This PR instead avoids creating a work inside pointToPoint when
coalescing is active. Instead, only at endCoalescing() is a work finally
intialized and enqueued. But it adds a record() call inside
pointToPoint() instead of creating a work, during coalescing. This
record() call picks up tensor shapes and op names.
It ALSO changes initWork to accept a 'record' argument. This defaults to
false, and should only be set to true if the caller ensures the work
will be enqueued by workEnqueue, ensuring its cuda events are live when
used by flight recorder's update_state().
The testing uncovers some odd pre-existing behavior and leaves them
alone for now. We could change some of these
- seq starts off at 1, not 0 for first op (but this is inconistent)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/120270
Approved by: https://github.com/shuqiangzhang
ghstack dependencies: #120724
Summary:
The current dump timeout logic is a bit cumbersome as it needs 2 times: 1.
timeout, 2. wake up time. And in theory the caller just needs to wait
for a max of timeout value for the dump and declare the dump to be
either successful or not. Also we unify the async call using std::async
instead of a customized async lauch function for each operation.
Test Plan:
Unit tests
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/120331
Approved by: https://github.com/wconstab
Do not run test ConstantPropagation.CustomClassesCanBePropagated on a platform where QNNPACK is not supported.
For example, this test fails on M1 Mac because QNNPACK is not supported on M1 Mac:
[----------] 1 test from ConstantPropagation
[ RUN ] ConstantPropagation.CustomClassesCanBePropagated
unknown file: Failure
as described in more details in the issue #88613.
After the PR, test passes successfully as below:
[----------] 1 test from ConstantPropagation
[ RUN ] ConstantPropagation.CustomClassesCanBePropagated
[ OK ] ConstantPropagation.CustomClassesCanBePropagated (0 ms)
[----------] 1 test from ConstantPropagation (0 ms total)
Fixes#88613
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119139
Approved by: https://github.com/jcaip
Recently we made it possible to serialize ExportedPrograms with fake parameters/buffers/etc.
The serialization regime was kind of whacky; basically we serialized a stub and reassembled the FakeTensor using metadata that we had stashed elsewhere in the Graph state.
This was bad for a few reasons:
- Storing the metadata separately from the actual serialized object caused situations where you could have one but not the other. An example case is if you had a FakeTensor contained inside a TorchBind object—there was no obviously place to store the metadata for this. This actually happens—TensorQueue in fbgemm does this.
- It created an annoying cycle: we had to deserialize the Graph's tensor metadata in order to deserialize (potentially faked) constants, but we need constants in order to deserialize the Graph.
This fixes all that. The basic idea is to patch the reducer function for FakeTensor at serialization time, and serialize a copy of the FakeTensor metadata. We already are policing BC for the TensorMeta schema struct so it's not a net increase in the BC surface.
As a bonus, I fixed a weird bug with torchbind tracing where we were accidentally reinterpreting a torch.ScriptObject as a torch.ScriptModule (which was the root cause of some weird behavior @bahuang was seeing last week).
Differential Revision: [D53601251](https://our.internmc.facebook.com/intern/diff/D53601251/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119531
Approved by: https://github.com/zhxchen17
Summary:
Add Runtime Constant-folding for AOTInductor.
This also include the invocation of constant folding at load time.
The constant folding lowering is a 2-step process.
First, we split the graph into 2 modules, one of it is the constant module, which doesn't depend on any input and the whole module could be inferred (constant-folded) one-time and be reused. The constant module, is lowered, and being codegen-ed as usual and cached (let's call this constant code). The constant code reuses the whole lowering/profiling/etc. process, only difference is that we do not generate any headers or initialization for the constant code.
Second, after handling the constant module, we take care of the main module (which is the part that would depend on the user input.) For the main module, we take in one additional component, the constant code, compare with a normal lowering. Addition step we do here is that, we inject the constant code into the codegen-ed main module, and create the caller for the main module to consume the result of the constant module.
Test Plan: Unit tests included in commit.
Differential Revision: D53274382
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118765
Approved by: https://github.com/chenyang78
Summary: `constraints` argument for `torch.export` has been deprecated in favor of the `dynamic_shapes` argument. This PR updates the use of the deprecated API in `caffe2/test/cpp` and `torchrec/distributed/test/test_pt2`.
Test Plan: CI
Differential Revision: D52977354
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118026
Approved by: https://github.com/chenyang78
This PR adds the bare minimum functionality to get torchbind working in an e2e testable way on PT2.
It implements:
* ProxyTensor support
* Simple torch.export support (proxytensor-only path, e.g. non-strict).
* add some tests exercising the path.
Because all this is not fully baked, I hide the functionality behind a feature flag (`enable_torchbind_tracing()`) so it does not affect regular users for now.
Still on the agenda:
* Dynamo support
* Actual FakeMode support
* Mutability support
Hoping to get this first bit in as a standalone, as it will unblock some more extensive experimentation/testing going on internally.
Differential Revision: [D51825372](https://our.internmc.facebook.com/intern/diff/D51825372/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/117697
Approved by: https://github.com/SherlockNoMad
Today watchdog's sleep interval is 1s. That's a bit long compared to modern GPU link's (or network link's) speed.
Take DDP and Ampere for example:
DDP's bucket size = 25 MB
Ampere's NVLink speed = 250 GB/s
25 MB / 250 GB/s = 100 ms.
So we are updating the interval to 100 ms.
Update:
25 MB / 250 GB/s = 0.1 ms
But let's see how it goes so far between making the checking more aggressive.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/117297
Approved by: https://github.com/fduwjj
Previously, we have the writer register to each NCCL PG(backend), so for every pg, we have a NCCL PG instance, so if we use some customized writer when multiple sub-PGs are used, we need to ensure user to register the writer for every backend which indicates a bad UX. Furthermore, the debug info is global, so it does not make sense to have the writer for each instance. We even have a static mutex in the `dumpDebuggingInfo` to ensure we serialize the write, that makes it more obvious that we can make the writer a singleton so that we only have one writer instance for all PG instances.
Although the rationale is clear, the implementation may vary a lot. So this PR is RFC for now to see if this implementation makes sense or not.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/116489
Approved by: https://github.com/kwen2501
Summary:
Refactor update inactive constant buffer to allow updating with active
buffer.
Test Plan:
Existing test to test inactive buffer updates.
UpdateConstantsCuda in cpp test for active buffer updates.
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/116001
Approved by: https://github.com/chenyang78
Replaces the "always sleep 30 sec before abort" with "wait up to 30 sec
for the future to complete then abort". The difference in this case is
the abort happens as soon as the dump finishes up to a maximum, instead
of always waiting the maximum.
Allows multiple calls to dump, which will be serialized.
Renames tryWriteDebugInfo to launchAsyncDebugDump in spirit of the
change to support more than one launch and to always launch rather than
only launching on the first call.
Adds a test for dumping on timeout.
This reverts commit ac7d14baad.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/115332
Approved by: https://github.com/fduwjj
Replaces the "always sleep 30 sec before abort" with "wait up to 30 sec
for the future to complete then abort". The difference in this case is
the abort happens as soon as the dump finishes up to a maximum, instead
of always waiting the maximum.
Allows multiple calls to dump, which will be serialized.
Renames `tryWriteDebugInfo` to `launchAsyncDebugDump` in spirit of the
change to support more than one launch and to always launch rather than
only launching on the first call.
Adds a test for dumping on timeout.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/115176
Approved by: https://github.com/zdevito
Summary:
This adds function to model container doing weight swapping with double buffering.
There are 2 parts for double buffering
a) Write constants into inactive buffer
b) Swap active buffer
For (a), we write the constants into the buffer that's currently not in use, and store the information in both constants map and the corresponding constant array to read.
For (b), we obtain the lock, and activate the constant map/constant array that is inactive, and flag the one that's currently in use to inactive.
Test Plan:
test/cpp/aot_inductor/test.cpp
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D51543732](https://our.internmc.facebook.com/intern/diff/D51543732)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/114446
Approved by: https://github.com/chenyang78, https://github.com/eellison
Previously:
```
[W Utils.hpp:133] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function getCvarInt)
[W Utils.hpp:133] Warning: Environment variable NCCL_ASYNC_ERROR_HANDLING is deprecated; use TORCH_NCCL_ASYNC_ERROR_HANDLING instead (function getCvarInt)
```
With this PR, those warnings disappear. They were introduced in #114077
This change was generated with this sed script, applied with `sed -i -f /tmp/x **/*.{py,hpp,cpp,cc}` and hand inspected.
```
s/\bNCCL_BLOCKING_WAIT\b/TORCH_NCCL_BLOCKING_WAIT/g
s/\bNCCL_ENABLE_TIMING\b/TORCH_NCCL_ENABLE_TIMING/g
s/\bNCCL_DESYNC_DEBUG\b/TORCH_NCCL_DESYNC_DEBUG/g
s/\bNCCL_ASYNC_ERROR_HANDLING\b/TORCH_NCCL_ASYNC_ERROR_HANDLING/g
s/\bENABLE_NCCL_HEALTH_CHECK\b/TORCH_ENABLE_NCCL_HEALTH_CHECK/g
s/\bNCCL_USE_TENSOR_REGISTER_ALLOCATOR_HOOK\b/TORCH_NCCL_USE_TENSOR_REGISTER_ALLOCATOR_HOOK/g
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/114880
Approved by: https://github.com/kwen2501
- [c10d] (retry) Opportunistically use `ncclCommSplit` when creating new NCCL groups (#112889)
- Guard use of `split_from` with a `hasattr` check for cases when NCCL (or RCCL) lacks `ncclCommSplit`
Fixes cause of revert of original PR
Pull Request resolved: https://github.com/pytorch/pytorch/pull/114385
Approved by: https://github.com/huydhn
Currently `ncclCommInitRankConfig` is always used when creating new
communicator groups. This is wasteful as it creates non-shared pairs
of endpoint queues as well as costs time to re-establish
communication.
This change is transparent and opportunistic; when `dist.new_group` is
called, it will use the existing, healthy world process group to
select the right ranks to include in the process group.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112889
Approved by: https://github.com/kwen2501
NCCL_ prefix should only be used for NCCL library's environment variables. We currently use a few environment variables in PyTorch with the NCCL_ prefix that are the NCCL library does not understand.
This patch renames such environment variables to use the TORCH_NCCL_ prefix instead. We still maintain the old NCCL_ variables, but throw a warning when they are used.
The following env changes have been made:
`NCCL_BLOCKING_WAIT` -> `TORCH_NCCL_BLOCKING_WAIT`
`NCCL_ENABLE_TIMING` -> `TORCH_NCCL_ENABLE_TIMING`
`NCCL_DESYNC_DEBUG` -> `TORCH_NCCL_DESYNC_DEBUG`
`NCCL_ASYNC_ERROR_HANDLING` -> `TORCH_NCCL_ASYNC_ERROR_HANDLING`
`ENABLE_NCCL_HEALTH_CHECK` -> `TORCH_ENABLE_NCCL_HEALTH_CHECK`
`NCCL_USE_TENSOR_REGISTER_ALLOCATOR_HOOK` -> `TORCH_NCCL_USE_TENSOR_REGISTER_ALLOCATOR_HOOK`
Fixes #ISSUE_NUMBER
Pull Request resolved: https://github.com/pytorch/pytorch/pull/114077
Approved by: https://github.com/fduwjj
Summary:
The getCvar* functions allow us to provide multiple environment variables for the same value. This allows us to deprecate some variables in favor of others, while still allowing users to temporarily use the old variables for some time.
Test Plan: OSS CI
Reviewed By: fduwjj, XilunWu
Differential Revision: D51225487
Fixes #ISSUE_NUMBER
Pull Request resolved: https://github.com/pytorch/pytorch/pull/113797
Approved by: https://github.com/fduwjj
There was missing support for bfloat scalars. When I use gloo backend
`torch.distributed.init_process_group(backend='gloo')`
and run
`torch.nn.parallel.DistributedDataParallel(model)`
and _model_ has Bfloat16 features I receive following error:
`RuntimeError: Invalid scalar type`
This change fix this issue.
c10::BFloat16 defines conversions from/to float, so calculations are made on float for bfloat.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/113557
Approved by: https://github.com/XilunWu, https://github.com/jgong5
This PR is to enable the store of NCCL flight recorder to storage and make it configurable by letting users register their own way of storing the debug info. We will then provide users a script to offline parse and process the dumped blobs.
One thing, this PR is not trying to resolve is to decide where to dump the debug info. I will send a follow-up PR to address that.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/113503
Approved by: https://github.com/zdevito
This PR has the following goals:
1. Detect unhealthy nccl watchdog thread by implementing a heartbeat. NCCL watchdog sometimes can hang for several reasons such as nccl/cuda API bugs or unexpected blocking behaviors. This is the last resort to ensure that we don't silently keep the training job run for hours.
2. Sometimes, the process gets stuck in the destroy of NCCL PG, and this PR will ensure that we will eventually abort it after some time (by default 2 mins)
3. Once heartbeat cannot be heard, we dump debug information (for now, we just use the flight recorder implemented in https://github.com/pytorch/pytorch/pull/110960/files) to disk. (How and where to dump the debug info will be addressed in the following PR).
4. Finally, we initiate std::abort via `LOG(FATAL)` to kill the process.
To clarify further what this PR is trying to solve, we first list are four cases when a NCCL PG can end up with:
- case 1: ncclwatchdog gets stuck (maybe some blocking API) and heartbeat monitor kills it during regular heartbeat monitor loop.
- case 2: ncclwatchdog timeout and desync report or destroy kicked in(let's call it shutdown) but this shutdown takes so long and heartbeat believes it has to kills the process anyway.
- case 3: ncclwatchdog aborts the process (heartbeat monitor not involved)
- case 4: program exits cleanly (heartbeat monitor not involved)
As we can see here, this PR is trying to address case one and two and we also want to ensure adding one more monitor thread does not interfere what we are currently doing in case three and four. That's why we added two flags `terminateHeartbeatMonitorThread_` and `collectiveDebugInfoMode_`.
For case three and four, either `monitorWakeUpCV_` will be waked up in the destructor or `terminateHeartbeatMonitorThread_` will be set to true. So that monitor thread will just exit ASAP.
For case one, both `terminateHeartbeatMonitorThread_` and `collectiveDebugInfoMode_` will still false when monitor thread see there are no heartbeat, so it will directly kill the process. For case two, either `terminateHeartbeatMonitorThread_` and `collectiveDebugInfoMode_` will be true, the monitor thread will wait extra time before killing the process.
Differential Revision: [D51146305](https://our.internmc.facebook.com/intern/diff/D51146305)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112518
Approved by: https://github.com/kwen2501, https://github.com/wconstab
As this is the oldest gcc that is fully compatible with C++17 standard.
- Replace number of conditional version with simpler `if(CMAKE_COMPILER_IS_GNUCXX)` or `append_cxx_flag_if_supported`.
- As `-Wsuggest-override` condition was hidden before incorrect guard, add missing `override` keywords to `torch::autograd::PyFunctionTensorPostAccGradHooks::apply_with_saved` , `caffe2::python::TensorFeeder::Feed` and `cafee2::NetObserverReporterPrint::report```
Fixes https://github.com/pytorch/pytorch/issues/101839
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112858
Approved by: https://github.com/Skylion007, https://github.com/albanD
As this is the oldest gcc that is fully compatible with C++17 standard.
- Replace number of conditional version with simpler `if(CMAKE_COMPILER_IS_GNUCXX)` or `append_cxx_flag_if_supported`.
- As `-Wsuggest-override` condition was hidden before incorrect guard, add missing `override` keywords to `torch::autograd::PyFunctionTensorPostAccGradHooks::apply_with_saved` , `caffe2::python::TensorFeeder::Feed` and `cafee2::NetObserverReporterPrint::report```
Fixes https://github.com/pytorch/pytorch/issues/101839
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112858
Approved by: https://github.com/Skylion007, https://github.com/albanD
If code is compiled without `glog`, there are no way to control log levels other than explicitly calling `c10::initLogging()`
Test plan: Run `TORCH_CPP_LOG_LEVEL=0 ./bin/ProcessGroupNCCLTest` and observe extra log messages
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112809
Approved by: https://github.com/fduwjj
Summary:
Move the profiler's Approximate Clock from libtorch to libc10. The main reason is to allow c10 features to get time.
The clock is using TSC when available for performance. CUDA Caching Allocator's implementation of memory snapshot will add the timestamps to memory events with this same clock in subsequent diff.
Test Plan: CI
Differential Revision: D50601935
Pulled By: aaronenyeshi
Pull Request resolved: https://github.com/pytorch/pytorch/pull/111972
Approved by: https://github.com/davidberard98
Updates `_export.aot_compile` to pass a torch IR graph to inductor, allowing inductor to now run the pre_grad_passes, and reuse more of inductor's code.
Also updates the API to only return the `so_path`, and not returning the exported program. The pytree call spec is now serialized and placed inside of the generated model code. When calling the model, because there is no c++ pytree implementation linked yet, we can access the call specs through `get_call_spec()`, and call pytree flatten/unflattenin python.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110020
Approved by: https://github.com/desertfire
- rename `__HIP_PLATFORM_HCC__` to `__HIP_PLATFORM_AMD__`
- rename `HIP_HCC_FLAGS` to `HIP_CLANG_FLAGS`
- rename `PYTORCH_HIP_HCC_LIBRARIES` to `PYTORCH_HIP_LIBRARIES`
- workaround in tools/amd_build/build_amd.py until submodules are updated
These symbols have had a long deprecation cycle and will finally be removed in ROCm 6.0.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/111975
Approved by: https://github.com/ezyang, https://github.com/hongxiayang
Keep a buffer of the last 16384 nccl work actions, including the stack
trace that launched the event.
When torch._C._distributed_c10d._dump_nccl_trace(), it an dump these to
a pickled archive.
For each action we get:
process_group_id, seq_id, collective_name, size_of_first_tensor, stack trace
state - issued, started, completed (based on cuda events and queried if
necessary when the dump is requested)
I tested that it is possible to query event state when the streams are
otherwise stuck.
Differential Revision: [D50138956](https://our.internmc.facebook.com/intern/diff/D50138956)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110960
Approved by: https://github.com/wconstab
