Continuation after https://github.com/pytorch/pytorch/pull/90163.
Here is a script I used to find all the non-existing arguments in the docstrings (the script can give false positives in presence of *args/**kwargs or decorators):
_Edit:_
I've realized that the indentation is wrong for the last `break` in the script, so the script only gives output for a function if the first docstring argument is wrong. I'll create a separate PR if I find more issues with corrected script.
``` python
import ast
import os
import docstring_parser
for root, dirs, files in os.walk('.'):
for name in files:
if root.startswith("./.git/") or root.startswith("./third_party/"):
continue
if name.endswith(".py"):
full_name = os.path.join(root, name)
with open(full_name, "r") as source:
tree = ast.parse(source.read())
for node in ast.walk(tree):
if isinstance(node, ast.FunctionDef):
all_node_args = node.args.args
if node.args.vararg is not None:
all_node_args.append(node.args.vararg)
if node.args.kwarg is not None:
all_node_args.append(node.args.kwarg)
if node.args.posonlyargs is not None:
all_node_args.extend(node.args.posonlyargs)
if node.args.kwonlyargs is not None:
all_node_args.extend(node.args.kwonlyargs)
args = [a.arg for a in all_node_args]
docstring = docstring_parser.parse(ast.get_docstring(node))
doc_args = [a.arg_name for a in docstring.params]
clean_doc_args = []
for a in doc_args:
clean_a = ""
for c in a.split()[0]:
if c.isalnum() or c == '_':
clean_a += c
if clean_a:
clean_doc_args.append(clean_a)
doc_args = clean_doc_args
for a in doc_args:
if a not in args:
print(full_name, node.lineno, args, doc_args)
break
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/90505
Approved by: https://github.com/malfet, https://github.com/ZainRizvi
Sometimes you want to query the small element of a set of elements and use `sorted(elements)[0]` without a second thought. However, this is not optimal, since the entire list must be sorted first `O(n log n)`. It would be better to use the `min(elements)` method provided for this purpose `O(n)`.
Furthermore `sorted(elements)[::-1]` is not very efficient, because it would be better to use `sorted(elements, reverse=True)` to save the slice operation.
**TLDR: using `sorted(elements)[0]` is slow and can be replaced with `min(elements)`.**
I stumbled across these code snippets while playing around with CodeQL (see https://lgtm.com/query/4148064474379348546/).
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86995
Approved by: https://github.com/jansel
Fix use-dict-literal pylint suggestions by changing `dict()` to `{}`. This PR should do the change for every Python file except test/jit/test_list_dict.py, where I think the intent is to test the constructor.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/83718
Approved by: https://github.com/albanD
This is a new version of #15648 based on the latest master branch.
Unlike the previous PR where I fixed a lot of the doctests in addition to integrating xdoctest, I'm going to reduce the scope here. I'm simply going to integrate xdoctest, and then I'm going to mark all of the failing tests as "SKIP". This will let xdoctest run on the dashboards, provide some value, and still let the dashboards pass. I'll leave fixing the doctests themselves to another PR.
In my initial commit, I do the bare minimum to get something running with failing dashboards. The few tests that I marked as skip are causing segfaults. Running xdoctest results in 293 failed, 201 passed tests. The next commits will be to disable those tests. (unfortunately I don't have a tool that will insert the `#xdoctest: +SKIP` directive over every failing test, so I'm going to do this mostly manually.)
Fixes https://github.com/pytorch/pytorch/issues/71105
@ezyang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82797
Approved by: https://github.com/ezyang
### Description
Across PyTorch's docstrings, both `callable` and `Callable` for variable types. The Callable should be capitalized as we are referring to the `Callable` type, and not the Python `callable()` function.
### Testing
There shouldn't be any testing required.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82487
Approved by: https://github.com/albanD
Summary:
This PR was created to resolve issue brought up in https://fb.workplace.com/groups/319878845696681/permalink/741428653541696/
Changes:
- Adds timeout argument to RpcAgent.join()
- Add optional timeout argument to ThriftRpcAgent barrier()
- During shutdown (ThriftRpcAgent join) calls the barrier, the agent will use the timeout passed to shutdown and pass that timeout into the join().
- Update API.py to also include fix bug (missing timeout for signal)
- Change default shutdown timeout to 0 (no timeout). Existing functionality in _all_gather will remain the same and wait indefinitely for signal if no timeout is set for the function. New functionality has user specify timeout for both the signal and rpc calls.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76194
Test Plan:
Modified barrier test
buck test torch/fb/distributed/thriftRpcBackend/test:ThriftRpcAgentTest -- BarrierTest
Reviewed By: mrshenli
Differential Revision: D35825382
fbshipit-source-id: e91e9ab5d9fca08787cb6b6b8125a4b03d1c7cde
(cherry picked from commit fcf899a387001574bf4e39a213ea741611d76097)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/75616
Kineto introduced a new profiler to read performance counters from NVIDIA GPUs (CUPTI Range Profiler API)
Here we are adding support to configure this Kineto range profiler mode
Example
```
with torch.profiler.profile(
activities=[ProfilerActivity.CUDA],
record_shapes=True,
on_trace_ready=trace_handler,
experimental_config=torch.profiler._ExperimentalConfig(
profiler_metrics=[
"kineto__tensor_core_insts",
"dram__bytes_read.sum",
"dram__bytes_write.sum"],
profiler_measure_per_kernel=False),
) as prof:
res = train_batch(modeldef)
prof.step()
```
## Details
* Introduce a new structure `KinetoProfilerConfig` so users can configure Kineto specific options, keeps profiler API consistent.
* Populate configuration options for Kineto.
Test Plan: CI and tested on resnet50
Reviewed By: robieta
Differential Revision: D34489487
fbshipit-source-id: 8ef82d2593f4f4d5824ca634f7d25507bc572caa
(cherry picked from commit 4a2af70629db55a605d4b8d0a54d41df2b247183)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/73372
This PR which allows for optional `world_size` argument in init_rpc. This makes changes in rendezvous to allow for `NoneType` for world_size and creates a new code path when initializing TensorPipe agent for init_rpc. The TensorPipe agent is protected by a critical section enforced using the store, so that only one node can create a TPAgent at a time.
This PR does not yet enable RPC commands between ranks.
Previously:
```python
os.environ['MASTER_ADDR'] = 'localhost'
os.environ['MASTER_PORT'] = '29500'
init_rpc("worker0", world_size=1, rank=0)
```
Now (only rank is needed):
```python
os.environ['MASTER_ADDR'] = 'localhost'
os.environ['MASTER_PORT'] = '29500'
init_rpc("worker0", rank=0)
```
Test Plan: Imported from OSS
Reviewed By: mrshenli
Differential Revision: D34621651
Pulled By: H-Huang
fbshipit-source-id: 09dbb511d5a00c219a6ce0a35501ff2e388998b0
(cherry picked from commit 834aedc3256167399c323169ef2f0c9b3cf98dff)
Summary:
From https://github.com/pytorch/pytorch/issues/67626: RRefProxy (rref.rpc_async, rref.rpc_sync, rref.remote) currently uses a blocking RPC call to the owner
This is done by chaining async calls. In the sync case we wait on the
resulting Future.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/70206
Test Plan:
I ran rpc_tests using tensorpipe_rpc_agent_test_fixture.py and had to
adjust test_rref_proxy_timeout to the new behavior.
I ran into test_tensorpipe_set_default_timeout failing due to the
timeout being too small. Doesn't look related to this change.
mrshenli
Fixes https://github.com/pytorch/pytorch/issues/67626
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse SciPioneer H-Huang
Reviewed By: pritamdamania87
Differential Revision: D33243348
Pulled By: kumpera
fbshipit-source-id: e1e8c34bb3d170407c0a793e2e585357f905d3c6
(cherry picked from commit 1ad5a7ceea)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/68128
Reland of D31762735 (0cbfd466d2).
This diff was originally reverted due to failure in test_send_export_type_through_rpc_with_custom_pickler.
I updated rpc_pickler_test.py to prevent a race condition where processes were not registering their pickler before handling their rpc_sync calls.
Test Plan:
rpc_pickler_test file:
buck test mode/dev-nosan -c 'cxx.coverage_only=caffe2' //caffe2/torch/fb/training_toolkit/backend/metrics/tests:rpc_pickler_test //caffe2/torch/fb/training_toolkit/backend/metrics/collectors/fbdata_aggregator/tests:batch_collector_test -- --run-disabled --collect-coverage '--code-coverage-session=test_session' --force-tpx
rpc_pickler stress test:
buck test mode/dev-nosan -c 'cxx.coverage_only=caffe2' //caffe2/torch/fb/training_toolkit/backend/metrics/tests:rpc_pickler_test -- --exact 'caffe2/torch/fb/training_toolkit/backend/metrics/tests:rpc_pickler_test - test_send_export_type_through_rpc_with_custom_pickler (caffe2.torch.fb.training_toolkit.backend.metrics.tests.rpc_pickler_test.CythonTypeRpcSpawnTest)' --run-disabled --collect-coverage '--code-coverage-session=test_session' --force-tpx --jobs 18 --stress-runs 10 --record-results
Reviewed By: mrshenli
Differential Revision: D32316077
fbshipit-source-id: e58de2335fbaa3ab46d46fe222c659197633a5e4
Summary:
Fixes [issue#64](https://github.com/MLH-Fellowship/pyre-check/issues/64)
This PR fixes the type checking errors in torch/distributed/rpc/options.py.
The variable types in 84:8 and 85:8 were declared to have type `List` but were sometimes assigned a value of `None`. This caused an incompatitble variable type error. Therefore, I changed the type from `List` to `Optional[List]` . Hence, this fixes the incompatitble variable type error.
Signed-off-by: Onyemowo Agbo
onionymous
0xedward
Pull Request resolved: https://github.com/pytorch/pytorch/pull/68056
Reviewed By: zou3519
Differential Revision: D32282289
Pulled By: mrshenli
fbshipit-source-id: ee410165e623834b4f5f3da8d44bd5a29306daae
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67924
This diff reverts the changes made in D31762735 (0cbfd466d2)
Test Plan: Wait for CI
Reviewed By: derekmod-fb
Differential Revision: D32214744
fbshipit-source-id: e0a65b6a31a88216ae1243549fcbc901ef812374
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67363
ProcessGroup RPC backend is deprecated. In 1.10 it would throw an error to the user to be more user friendly. This PR now removes it completely.
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse SciPioneer H-Huang
Test Plan: Imported from OSS
Reviewed By: bdhirsh
Differential Revision: D32138321
Pulled By: H-Huang
fbshipit-source-id: b4f700d8f1b1d46ada7b5062d3f754646571ea90
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65391
TSAN tests are much slower than the usual dev/opt mode, about 5-10x
slower.
As a result, for TSAN build mode we use a much higher timeout for distributed
tests.
ghstack-source-id: 138584613
Test Plan: waitforbuildbot
Reviewed By: cbalioglu
Differential Revision: D31076575
fbshipit-source-id: 44a485f07101deac536470ceeff2a52cac4f9e0b
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61792
KinetoEvent
This PR adds module hierarchy information to events.
What is module hierarchy information attached to events?
During profiling a TorchScript module, when events are added, we ask JIT
what is the module hierarchy associated with the node being
executed. At the time of execution of that node, there might be multiple
frames in the stack of interpreter. For each frame, we find
corresponding node and the corresponding module hierarchy is queried.
Module hierarchy corresponding to the node is associated with node's
InlinedCallStack. InlinedCallStack of node tracks the path via which the
node is inlined. Thus during the inlining process we annotate
module information corresponding to the CallMethod nodes being inlined.
With this PR, chrome trace will contain additional metadata:
"Module Hierarchy". This can look like this:
TOP(ResNet)::forward.SELF(ResNet)::_forward_impl.layer1(Sequential)::forward.0(BasicBlock)::forward.conv1(Conv2d)::forward.SELF(Conv2d)::_conv_forward
It contains module instance, type name and the method name in the
callstack.
Test Plan:
test_profiler
Imported from OSS
Reviewed By: raziel, ilia-cher
Differential Revision: D29745442
fbshipit-source-id: dc8dfaf7c5b8ab256ff0b2ef1e5ec265ca366528
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61931
This PR consolidates the profiling code around a new C++ implementation
(profiler_kineto.h/cpp) and uses it unconditionally from
torch.autograd.profiler/torch.profiler:
1. Always use profiler_kineto.h/cpp as the C++ implementation
2. Simplify profiler.py to remove unneeded parts depending on legacy
impl
3. Move some of the legacy logic into profiler_legacy.py (to be fully
deleted later)
Test Plan:
USE_KINETO=1 USE_CUDA=1 USE_MKLDNN=1 BLAS=MKL BUILD_BINARY=1 python setup.py develop install --cmake
python test/test_profiler.py -v
USE_KINETO=0 USE_CUDA=1 USE_MKLDNN=1 BLAS=MKL BUILD_BINARY=1 python setup.py develop install --cmake
python test/test_profiler.py -v
Imported from OSS
Reviewed By: gdankel
Differential Revision: D29801599
fbshipit-source-id: 9794d29f2af38dddbcd90dbce4481fc8575fa29e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62409
This a reland of #61907 because removing process_group_agent.h / cpp broke facebook specific tests. I will remove the files and update the internal test code in a separate PR.
Test Plan: Imported from OSS
Reviewed By: mrshenli
Differential Revision: D29990001
Pulled By: H-Huang
fbshipit-source-id: 2ee333322247d8b72691152308c3297e8c0c006d
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61907
Removing the code for faulty process group agent since it was replaced by faulty tensorpipe agent
Test Plan: Imported from OSS
Reviewed By: mrshenli
Differential Revision: D29794666
Pulled By: H-Huang
fbshipit-source-id: 0b35191cc07220b6774ecacc8d004f25fd2e87f0
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61421
This PR adds the faulty tensorpipe agent implementation and replaces all faulty process group agent tests with it. The faulty tensorpipe agent code is very similar to that of faulty process group agent. It allows the user to fail or delay certain types of rpc messages, which is used in the faulty agent tests. These changes are needed to deprecate the process group rpc backend.
Summary of changes:
- Add faulty tensorpipe agent class
- Update tensorpipe pipeWrite function to allow to be overwritten and add delay
- Update test backend registry and faulty agent tests to use the FAULTY_TENSORPIPE_AGENT backend.
This effects all faulty agent tests, here a few of them as sample commands:
`pytest test/distributed/rpc/test_faulty_agent.py -vs -k test_verify_backend_options`
`pytest test/distributed/rpc/test_faulty_agent.py -vs -k test_no_faulty_messages`
`pytest test/distributed/rpc/test_faulty_agent.py -vs -k test_builtin_remote_message_dropped_timeout`
Test Plan: Imported from OSS
Reviewed By: mrshenli
Differential Revision: D29773739
Pulled By: H-Huang
fbshipit-source-id: 6b2bc366735d70b79943d4207f454bc9555bbf5f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61094
`_wait_all_workers` was swallowing exceptions and as a result if there
were any errors it would still continue with rpc_agent.join() which would hang
since something already failed before.
To fix this, I've ensured that wait_all_workers throws and in that case we just
proceed with an ungraceful shutdown without joining.
ghstack-source-id: 133160706
Test Plan:
1) Added unit test.
2) waitforbuildbot
Reviewed By: rohan-varma
Differential Revision: D29509286
fbshipit-source-id: 7c3f1c68d712ae2f63e10e0216580db8e9bcc29d
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/59801
Fixes https://github.com/pytorch/pytorch/issues/59795.
The RPC calls in shutdown no longer able to finish within 5s if
there is no other RPCs before `rpc.shutdown()` in that process,
because agent initialization can take longer than 5s. We don't
have this problem previously, because TensorPipe's backend
registry used to use RPC to communicate CUDA devices in `init_rpc`.
However, after #58753, `init_rpc` uses ProcessGroup to communicate
devices, and hence the channels/transport could be uninitialized
after `init_rpc`.
Differential Revision:
D29039238
D29039238
Test Plan: Imported from OSS
Reviewed By: rohan-varma
Pulled By: mrshenli
fbshipit-source-id: 46f89b01a058a51d271ddef9084a67b220a067b7
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58753
TSAN was (rightfully!) detecting and complaining about a race due to the fact that upon init the TP agent exchanges the device maps between nodes using RPC requests (and by doing so it accesses the device maps) and then sets the reverse device maps (thus possibly modifying the set of devices). This resulted in a data race, i.e., simultaneously reading and writing the set of devices without synchronizing.
One solution is to add a mutex around the devices, which works, but is "annoying". An alternative solution is to make the set of devices immutable (i.e., `const`). For that to work, we need to exchange the device maps without using RPC calls. We can do so using the process group that we need to create anyways.
Since now there's a lot more logic in Python, I've moved (and restructured) all safety checks over there, and removed them from C++.
ghstack-source-id: 130583775
Test Plan: Unit tests
Reviewed By: mrshenli
Differential Revision: D28603754
fbshipit-source-id: 88533e65d72d1eb806dc41bec8d55def5082e290
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53423closes#40166
This change exposes a new API, rpc.barrier() which blocks the main processes of all workers running RPC until the whole group completes this function. Optionally rpc.barrier can take in a set of worker_names and only synchronize across those worker names.
Example:
```python
import os
import torch.multiprocessing as mp
import torch.distributed.rpc as rpc
os.environ["MASTER_ADDR"] = "localhost"
os.environ["MASTER_PORT"] = "5678"
world_size = 4
odd_num_workers = [f"worker{i}" for i in range(world_size) if i % 2]
even_num_workers = [f"worker{i}" for i in range(world_size) if not i % 2]
def worker(i):
print(i)
rpc.init_rpc(f"worker{i}", rank=i, world_size=world_size)
if i % 2:
print(f"start barrier {i}")
rpc.barrier(set(odd_num_workers))
else:
print(f"start barrier {i}")
rpc.barrier(set(even_num_workers))
rpc.shutdown()
print(f"shutdown{i}")
if __name__ == '__main__':
with mp.Pool(processes=world_size) as pool:
pool.map(worker, range(world_size))
```
Test Plan: Imported from OSS
Reviewed By: rohan-varma
Differential Revision: D27737145
Pulled By: H-Huang
fbshipit-source-id: 369196bc62446f506d1fb6a3fa5bebcb0b09da9f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/59242
#Oringal PR Issue: https://github.com/pytorch/pytorch/issues/58274
This can be a workaround: Instead of passing a script `RemoteModule` over RPC, pass its `module_rref` field over RPC, and then construct a new `RemoteModule` on the receiver end.
ghstack-source-id: 130268018
Test Plan:
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_send_remote_module_over_the_wire_script_not_supported
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_remote_module_py_pickle_not_supported_script
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_create_remote_module_by_module_rref
Reviewed By: vipannalla
Differential Revision: D28794905
fbshipit-source-id: 1a677ff0d4b47c078ad47b50d7102a198a1fc39b
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/55728
Full design: https://github.com/pytorch/pytorch/issues/55207
This PR introduces ChunkShardingSpec (SingleShardingSpec in the design). Used
the name ChunkShardingSpec since it is very similar to `torch.chunk` in terms
of how a Tensor is split up and feels more clear compared to SingleShardingSpec.
ghstack-source-id: 129603318
Test Plan: waitforbuildbot
Reviewed By: SciPioneer
Differential Revision: D27694108
fbshipit-source-id: c8764abe6a4d5fc56d023fda29b74b5af2a73b49
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58020
Previously there is no RPC pickler for `RecursiveScriptModule`. Although it is a subclass of `ScriptModule`, the reducer of `ScriptModule` is not triggered for `RecursiveScriptModule` when a script remote module is sent over RPC.
This PR checkpoints the investigation of #58274, which makes sure that a RPC pickler is invoked here. This still cannot fix `test_send_remote_module_over_the_wire_script`. Will revisit this bug once there is a feature request from users.
ghstack-source-id: 128949642
Test Plan:
TODO: re-enable these tests
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_send_remote_module_over_the_wire_script
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_remote_module_py_pickle_not_supported_script
Reviewed By: rohan-varma
Differential Revision: D28346758
fbshipit-source-id: 3cff84ca665da03da6ed6acb094a1f594fcd945e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58019
In order to support sending `RemoteModule` over PRC, previously the pickling/unpickling of `RemoteModule` was implemented based on `__setstate__` and `__getstate__`. However, this means that the user can call regular Python pickler/unpickler to invoke the same logic,which should not be allowed.
This PR ensures that the pickling can only happen over RPC and not via regular python pickle.
Additionally, when a new attribute is added to `RemoteModule`, if it's not added to either `_REMOTE_MODULE_PICKLED_ATTRIBUTES` or `_REMOTE_MODULE_ATTRIBUTES_IGNORE_FOR_PICKLING`, this attribute will be ignored and an error message will be printed to std.err. However, it will not raise an exception like before, because such exception raised at the RPC layer will somehow cause timeout.
#Closes: https://github.com/pytorch/pytorch/issues/57516
ghstack-source-id: 128868501
Test Plan:
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_send_remote_module_over_the_wire
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_remote_module_py_pickle_not_supported
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- test_send_remote_module_with_a_new_attribute_ignored_over_the_wire
buck test mode/dev-nosan caffe2/test/distributed/rpc:process_group_agent -- RemoteModule
buck test mode/dev-nosan //caffe2/torch/fb/csrc/concurrency/test:atomic_int_interprocess_test -- --exact 'caffe2/torch/fb/csrc/concurrency/test:atomic_int_interprocess_test - test_multiple_processes (caffe2.torch.fb.csrc.concurrency.test.atomic_int_interprocess_test.ForkMultipleProcessTest)'
buck test mode/dev //caffe2/torch/distributed/fb/test:app_test -- --exact 'caffe2/torch/distributed/fb/test:app_test - test_custom_init_rpc (caffe2.torch.distributed.fb.test.app_test.TestRpc)'
Reviewed By: mrshenli
Differential Revision: D28318270
fbshipit-source-id: 7e7df2a6690f0860c4531a244d38789db424496f
