.. note:: :class: sphx-glr-download-link-note Click :ref:`here ` to download the full example code .. rst-class:: sphx-glr-example-title .. _sphx_glr_beginner_profiler.py: Profiling your PyTorch Module ------------ **Author:** `Suraj Subramanian `_ PyTorch includes a profiler API that is useful to identify the time and memory costs of various PyTorch operations in your code. Profiler can be easily integrated in your code, and the results can be printed as a table or retured in a JSON trace file. .. note:: Profiler supports multithreaded models. Profiler runs in the same thread as the operation but it will also profile child operators that might run in another thread. Concurrently-running profilers will be scoped to their own thread to prevent mixing of results. .. note:: PyTorch 1.8 introduces the new API that will replace the older profiler API in the future releases. Check the new API at `this page `__. Head on over to `this recipe `__ for a quicker walkthrough of Profiler API usage. -------------- .. code-block:: default import torch import numpy as np from torch import nn import torch.autograd.profiler as profiler Performance debugging using Profiler ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Profiler can be useful to identify performance bottlenecks in your models. In this example, we build a custom module that performs two sub-tasks: - a linear transformation on the input, and - use the transformation result to get indices on a mask tensor. We wrap the code for each sub-task in separate labelled context managers using ``profiler.record_function("label")``. In the profiler output, the aggregate performance metrics of all operations in the sub-task will show up under its corresponding label. Note that using Profiler incurs some overhead, and is best used only for investigating code. Remember to remove it if you are benchmarking runtimes. .. code-block:: default class MyModule(nn.Module): def __init__(self, in_features: int, out_features: int, bias: bool = True): super(MyModule, self).__init__() self.linear = nn.Linear(in_features, out_features, bias) def forward(self, input, mask): with profiler.record_function("LINEAR PASS"): out = self.linear(input) with profiler.record_function("MASK INDICES"): threshold = out.sum(axis=1).mean().item() hi_idx = np.argwhere(mask.cpu().numpy() > threshold) hi_idx = torch.from_numpy(hi_idx).cuda() return out, hi_idx Profile the forward pass ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ We initialize random input and mask tensors, and the model. Before we run the profiler, we warm-up CUDA to ensure accurate performance benchmarking. We wrap the forward pass of our module in the ``profiler.profile`` context manager. The ``with_stack=True`` parameter appends the file and line number of the operation in the trace. .. WARNING:: ``with_stack=True`` incurs an additional overhead, and is better suited for investigating code. Remember to remove it if you are benchmarking performance. .. code-block:: default model = MyModule(500, 10).cuda() input = torch.rand(128, 500).cuda() mask = torch.rand((500, 500, 500), dtype=torch.double).cuda() # warm-up model(input, mask) with profiler.profile(with_stack=True, profile_memory=True) as prof: out, idx = model(input, mask) Print profiler results ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Finally, we print the profiler results. ``profiler.key_averages`` aggregates the results by operator name, and optionally by input shapes and/or stack trace events. Grouping by input shapes is useful to identify which tensor shapes are utilized by the model. Here, we use ``group_by_stack_n=5`` which aggregates runtimes by the operation and its traceback (truncated to the most recent 5 events), and display the events in the order they are registered. The table can also be sorted by passing a ``sort_by`` argument (refer to the `docs `__ for valid sorting keys). .. Note:: When running profiler in a notebook, you might see entries like ``(13): forward`` instead of filenames in the stacktrace. These correspond to ``(line number): calling-function``. .. code-block:: default print(prof.key_averages(group_by_stack_n=5).table(sort_by='self_cpu_time_total', row_limit=5)) """ (Some columns are omitted) ------------- ------------ ------------ ------------ --------------------------------- Name Self CPU % Self CPU Self CPU Mem Source Location ------------- ------------ ------------ ------------ --------------------------------- MASK INDICES 87.88% 5.212s -953.67 Mb /mnt/xarfuse/.../torch/au (10): forward /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ aten::copy_ 12.07% 715.848ms 0 b (12): forward /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ LINEAR PASS 0.01% 350.151us -20 b /mnt/xarfuse/.../torch/au (7): forward /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ aten::addmm 0.00% 293.342us 0 b /mnt/xarfuse/.../torch/nn /mnt/xarfuse/.../torch/nn /mnt/xarfuse/.../torch/nn (8): forward /mnt/xarfuse/.../torch/nn aten::mean 0.00% 235.095us 0 b (11): forward /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ ----------------------------- ------------ ---------- ---------------------------------- Self CPU time total: 5.931s """ Improve memory performance ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Note that the most expensive operations - in terms of memory and time - are at ``forward (10)`` representing the operations within MASK INDICES. Let’s try to tackle the memory consumption first. We can see that the ``.to()`` operation at line 12 consumes 953.67 Mb. This operation copies ``mask`` to the CPU. ``mask`` is initialized with a ``torch.double`` datatype. Can we reduce the memory footprint by casting it to ``torch.float`` instead? .. code-block:: default model = MyModule(500, 10).cuda() input = torch.rand(128, 500).cuda() mask = torch.rand((500, 500, 500), dtype=torch.float).cuda() # warm-up model(input, mask) with profiler.profile(with_stack=True, profile_memory=True) as prof: out, idx = model(input, mask) print(prof.key_averages(group_by_stack_n=5).table(sort_by='self_cpu_time_total', row_limit=5)) """ (Some columns are omitted) ----------------- ------------ ------------ ------------ -------------------------------- Name Self CPU % Self CPU Self CPU Mem Source Location ----------------- ------------ ------------ ------------ -------------------------------- MASK INDICES 93.61% 5.006s -476.84 Mb /mnt/xarfuse/.../torch/au (10): forward /mnt/xarfuse/ /torch/nn (9): /mnt/xarfuse/.../IPython/ aten::copy_ 6.34% 338.759ms 0 b (12): forward /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ aten::as_strided 0.01% 281.808us 0 b (11): forward /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ aten::addmm 0.01% 275.721us 0 b /mnt/xarfuse/.../torch/nn /mnt/xarfuse/.../torch/nn /mnt/xarfuse/.../torch/nn (8): forward /mnt/xarfuse/.../torch/nn aten::_local 0.01% 268.650us 0 b (11): forward _scalar_dense /mnt/xarfuse/.../torch/nn (9): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ ----------------- ------------ ------------ ------------ -------------------------------- Self CPU time total: 5.347s """ The CPU memory footprint for this operation has halved. Improve time performance ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ While the time consumed has also reduced a bit, it’s still too high. Turns out copying a matrix from CUDA to CPU is pretty expensive! The ``aten::copy_`` operator in ``forward (12)`` copies ``mask`` to CPU so that it can use the NumPy ``argwhere`` function. ``aten::copy_`` at ``forward(13)`` copies the array back to CUDA as a tensor. We could eliminate both of these if we use a ``torch`` function ``nonzero()`` here instead. .. code-block:: default class MyModule(nn.Module): def __init__(self, in_features: int, out_features: int, bias: bool = True): super(MyModule, self).__init__() self.linear = nn.Linear(in_features, out_features, bias) def forward(self, input, mask): with profiler.record_function("LINEAR PASS"): out = self.linear(input) with profiler.record_function("MASK INDICES"): threshold = out.sum(axis=1).mean() hi_idx = (mask > threshold).nonzero(as_tuple=True) return out, hi_idx model = MyModule(500, 10).cuda() input = torch.rand(128, 500).cuda() mask = torch.rand((500, 500, 500), dtype=torch.float).cuda() # warm-up model(input, mask) with profiler.profile(with_stack=True, profile_memory=True) as prof: out, idx = model(input, mask) print(prof.key_averages(group_by_stack_n=5).table(sort_by='self_cpu_time_total', row_limit=5)) """ (Some columns are omitted) -------------- ------------ ------------ ------------ --------------------------------- Name Self CPU % Self CPU Self CPU Mem Source Location -------------- ------------ ------------ ------------ --------------------------------- aten::gt 57.17% 129.089ms 0 b (12): forward /mnt/xarfuse/.../torch/nn (25): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ aten::nonzero 37.38% 84.402ms 0 b (12): forward /mnt/xarfuse/.../torch/nn (25): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ INDEX SCORE 3.32% 7.491ms -119.21 Mb /mnt/xarfuse/.../torch/au (10): forward /mnt/xarfuse/.../torch/nn (25): /mnt/xarfuse/.../IPython/ aten::as_strided 0.20% 441.587us 0 b (12): forward /mnt/xarfuse/.../torch/nn (25): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ aten::nonzero _numpy 0.18% 395.602us 0 b (12): forward /mnt/xarfuse/.../torch/nn (25): /mnt/xarfuse/.../IPython/ /mnt/xarfuse/.../IPython/ -------------- ------------ ------------ ------------ --------------------------------- Self CPU time total: 225.801ms """ Further Reading ~~~~~~~~~~~~~~~~~ We have seen how Profiler can be used to investigate time and memory bottlenecks in PyTorch models. Read more about Profiler here: - `Profiler Usage Recipe `__ - `Profiling RPC-Based Workloads `__ - `Profiler API Docs `__ .. rst-class:: sphx-glr-timing **Total running time of the script:** ( 0 minutes 0.000 seconds) .. _sphx_glr_download_beginner_profiler.py: .. only :: html .. container:: sphx-glr-footer :class: sphx-glr-footer-example .. container:: sphx-glr-download :download:`Download Python source code: profiler.py ` .. container:: sphx-glr-download :download:`Download Jupyter notebook: profiler.ipynb ` .. only:: html .. rst-class:: sphx-glr-signature `Gallery generated by Sphinx-Gallery `_