TorchDynamo is a Python-level JIT compiler designed to make unmodified PyTorch programs faster. TorchDynamo hooks into the frame evaluation API in CPython (PEP 523) to dynamically modify Python bytecode right before it is executed. It rewrites Python bytecode in order to extract sequences of PyTorch operations into an FX Graph which is then just-in-time compiled with a customizable backend. It creates this FX Graph through bytecode analysis and is designed to mix Python execution with compiled backends to get the best of both worlds: usability and performance.
For more on TorchDynamo you can read our posts on PyTorch dev-discuss or watch a deep-dive video.
This repository also hosts TorchInductor, which is TorchDynamo backend able to translate an FX Graph into Triton for GPUs or C++/OpenMP for CPUs. We have a training performance dashboard comparing the performance of different training backends. You can read more in the TorchInductor post on PyTorch dev-discuss.
TorchDynamo is experimental and under active development. You are welcome to try it out and contribute, but should expect to find bugs and rough edges.
Python 3.8 is recommended. Python 3.7 through 3.10 are supported and tested.
TorchDynamo requires the latest development/nightly build of PyTorch or
building PyTorch from source. TorchDynamo also requires functorch
from the PyTorch repository. The Makefile target make setup_nightly_gpu
contain the commands used by our CI to setup dependencies.
The 1.12 branch contains an older snapshot of TorchDynamo that works on PyTorch 1.12. However, this is missing the latest features and not recommended.
Other development requirements can be installed with:
pip install -r requirements.txtInstall TorchDynamo with:
python setup.py developHere is a basic example of how to use TorchDynamo. One can decorate a function
or a method using torchdynamo.optimize to enable TorchDynamo optimization.
import torch
import torchdynamo
def my_compiler(gm: torch.fx.GraphModule, example_inputs: List[torch.Tensor]):
print("my_compiler() called with FX graph:")
gm.graph.print_tabular()
return gm.forward # return a python callable
@torchdynamo.optimize(my_compiler)
def fn(x, y):
a = torch.cos(x)
b = torch.sin(y)
return a + b
fn(torch.randn(10), torch.randn(10))Running the above example produces this output
my_compiler() called with FX graph:
opcode name target args kwargs
------------- ------ ------------------------------------------------------ ---------- --------
placeholder x x () {}
placeholder y y () {}
call_function cos <built-in method cos of type object at 0x7f1a894649a8> (x,) {}
call_function sin <built-in method sin of type object at 0x7f1a894649a8> (y,) {}
call_function add <built-in function add> (cos, sin) {}
output output output ((add,),) {}
This works for torch.nn.Module as well as shown below
import torch
import torchdynamo
class MockModule(torch.nn.Module):
def __init__(self):
super().__init__()
self.relu = torch.nn.ReLU()
def forward(self, x):
return self.relu(torch.cos(x))
mod = MockModule()
optimized_mod = torchdynamo.optimize(my_compiler)(mod)
optimized_mod(torch.randn(10))In the above examples, TorchDynamo uses a custom compiler (also referred to as
backend in the rest of the doc) my_compiler that just prints the Fx
GraphModule extracted by TorchDynamo's bytecode analysis, and returns the
forward callable. One could write new compilers in a similar fashion.
Let's take a look at one more example with control flow.
from typing import List
import torch
import torchdynamo
def my_compiler(gm: torch.fx.GraphModule, example_inputs: List[torch.Tensor]):
print("my_compiler() called with FX graph:")
gm.graph.print_tabular()
return gm.forward # return a python callable
@torchdynamo.optimize(my_compiler)
def toy_example(a, b):
x = a / (torch.abs(a) + 1)
if b.sum() < 0:
b = b * -1
return x * b
for _ in range(100):
toy_example(torch.randn(10), torch.randn(10))Running this example produces the following output:
my_compiler() called with FX graph:
opcode name target args kwargs
------------- ------- ------------------------------------------------------ ---------------- --------
placeholder a a () {}
placeholder b b () {}
call_function abs_1 <built-in method abs of type object at 0x7f8d259298a0> (a,) {}
call_function add <built-in function add> (abs_1, 1) {}
call_function truediv <built-in function truediv> (a, add) {}
call_method sum_1 sum (b,) {}
call_function lt <built-in function lt> (sum_1, 0) {}
output output output ((truediv, lt),) {}
my_compiler() called with FX graph:
opcode name target args kwargs
------------- ------ ----------------------- ----------- --------
placeholder b b () {}
placeholder x x () {}
call_function mul <built-in function mul> (b, -1) {}
call_function mul_1 <built-in function mul> (x, mul) {}
output output output ((mul_1,),) {}
my_compiler() called with FX graph:
opcode name target args kwargs
------------- ------ ----------------------- --------- --------
placeholder b b () {}
placeholder x x () {}
call_function mul <built-in function mul> (x, b) {}
output output output ((mul,),) {}
Note that the order of the last two graphs is nondeterministic depending on which one is encountered first by the just-in-time compiler.
TorchDynamo has a growing list of backends, which can be found in backends.py
or torchdynamo.list_backends(). Note many backends require installing
additional packages. Some of the most commonly used backends are
Debugging backends:
torchdynamo.optimize("eager")- Uses PyTorch to run the extracted GraphModule. This is quite useful in debugging TorchDynamo issues.torchdynamo.optimize("aot_nop")- Uses AotAutograd with no compiler, i.e, just using PyTorch eager for the AotAutograd's extracted forward and backward graphs. This is useful for debugging, and unlikely to give speedups.
Training & inference backends:
torchdynamo.optimize("inductor")- Uses TorchInductor backend with AotAutograd and cudagraphs. Read moretorchdynamo.optimize("nvfuser")- nvFuser with TorchScript. Read moretorchdynamo.optimize("aot_nvfuser")- nvFuser with AotAutograd. Read moretorchdynamo.optimize("aot_cudagraphs")- cudagraphs with AotAutograd. Read more
Inference-only backends:
torchdynamo.optimize("ofi")- Uses Torchscript optimize_for_inference. Read moretorchdynamo.optimize("fx2trt")- Uses Nvidia TensorRT for inference optimizations. Read moretorchdynamo.optimize("onnxrt")- Uses ONNXRT for inference on CPU/GPU. Read moretorchdynamo.optimize("ipex")- Uses IPEX for inference on CPU. Read more
Torchdynamo supports training, using AotAutograd to capture backwards:
- only the .forward() graph is captured by torchdynamo's python evalframe frontend
- for each segment of .forward() that torchdynamo captures, it uses AotAutograd to generate a backward graph segment
- each pair of forward, backward graph are (optionally) min-cut partitioned to save the minimal state between forward/backward
- the forward, backward pairs are wrapped in autograd.function modules
- usercode calling .backward() still triggers eager's autograd engine, which runs each 'compiled backward' graph as if it were one op, also running any non-compiled eager ops' .backward() functions
Current limitations:
- optimizer ops are currently not captured at all, and thus not compiled (under investigation to add support)
- DDP and FSDP, which rely on autograd 'hooks' firing between backward ops to schedule communications ops, may be pessimized by having all communication ops scheduled after whole compiled regions of backwards ops (WIP to fix this)
Example
model = ...
optimizer = ...
@torchdynamo.optimize("inductor")
def training_iteration(...):
return model(...)
for _ in range(100):
loss = training_iteration(...)
loss.backward()
optimizer.step()See Troubleshooting.
One could replace my_compiler() in the examples above with something that generates faster
code, for example one using optimize_for_inference:
def my_compiler(gm: torch.fx.GraphModule, example_inputs: List[torch.Tensor]):
scripted = torch.jit.trace(gm, example_inputs)
return torch.jit.optimize_for_inference(scripted)TorchDynamo also includes many backends, which can be found in
backends.py or torchdynamo.list_backends(). Note many backends
require installing additional packages. You can combine these backends
together with code like:
from torchdynamo.optimizations import BACKENDS
def my_compiler(gm: torch.fx.GraphModule, example_inputs: List[torch.Tensor]):
trt_compiled = BACKENDS["tensorrt"](gm, example_inputs)
if trt_compiled is not None:
return trt_compiled
# first backend failed, try something else...
cudagraphs_compiled = BACKENDS["cudagraphs"](gm, example_inputs)
if cudagraphs_compiled is not None:
return cudagraphs_compiled
return gm.forwardTorchDynamo operates just-in-time and specializes graphs based on dynamic properties. For example, the first graph above has the following guards:
GUARDS:
- local 'a' TENSOR_MATCH
- local 'b' TENSOR_MATCH
- global 'torch' FUNCTION_MATCH
If any of those guards fail, the graph will be recaptured and recompiled.
The interesting guard type there is TENSOR_MATCH, which checks the
following torch.Tensor properties:
- Python class of the tensor (tensor subclassing, etc)
- dtype
- device
- requires_grad
- dispatch_key (with thread-local includes/excludes applied)
- ndim
- sizes* (optional)
- strides* (optional)
*For sizes/strides you can disable this specialization by setting:
torchdynamo.config.dynamic_shapes = TrueThe full specialization mode allows the backend compiler to assume an entirely static graph. Unfortunately, most backends require this. Operators which return dynamic shapes will trigger a graph break when not in dynamic shape mode.
In some cases, you may not want unexpected compiles after a program has warmed up. For example, if you are serving production traffic in a latency critical application. For this, TorchDynamo provides an alternate mode where prior compiled graphs are used, but no new ones are generated:
frozen_toy_example = torchdynamo.run(toy_example)
frozen_toy_example(torch.randn(10), torch.randn(10))In some cases, you may want to ensure there are no graph breaks in your
program to debug performance issues. You can turn graph breaks into
errors by setting
nopython=True:
@torchdynamo.optimize(my_compiler, nopython=True)
def toy_example(a, b):Which will trigger the following error in the example program above:
Traceback (most recent call last):
...
torchdynamo.exc.Unsupported: generic_jump TensorVariable()
Processing original code:
File "example.py", line 7, in toy_example
if b.sum() < 0:As background reading, I'd suggest looking at the PyTorch, functorch, and TorchBench setup docs. Since these projects work together in different ways.
The following instructions use Miniconda.
conda create --name=torchdynamo python=3.8
conda activate torchdynamo
# install pytorch nightlies
# for CUDA version, replace `cpuonly` with `cudatoolkit=11.6`
conda install pytorch torchvision torchaudio torchtext cpuonly -c pytorch-nightly
pip install -v "git+https://github.com/pytorch/pytorch.git@`python -c "import torch.version; print(torch.version.git_version)"`#subdirectory=functorch"
git clone git@github.com:pytorch/torchdynamo.git
cd torchdynamo
pip install -r requirements.txt
# check if everything works
make test
If see errors about missing symbols from guards.so, that may mean your
C++ compiler is incompatible CUDA and/or with the one used to compile
PyTorch. You may need to change your compiler version or build PyTorch
from source.
Run tests with
pytest testTo debug a specific test (with more debug prints) do:
pytest -vsk <test name>Test on torchbenchmark models with:
python benchmarks/torchbench.py
Install format/linter deps with pip install -r requirements.txt, then:
make format # reformat all files (in-place)
make lint # run the lintersTorchDynamo has a BSD-style license, as found in the LICENSE file.