OSError: Missing: valgrind, callgrind_control, callgrind_annotate

The error occurs on below step:
8. Collecting instruction counts with Callgrind:

Traceback (most recent call last):
File "benchmark.py", line 805, in
stats_v0 = t0.collect_callgrind()
File "/usr/local/lib/python3.8/dist-packages/torch/utils/benchmark/utils/timer.py", line 486, in collect_callgrind
result = valgrind_timer_interface.wrapper_singleton().collect_callgrind(
File "/usr/local/lib/python3.8/dist-packages/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.py", line 526, in collect_callgrind
self._validate()
File "/usr/local/lib/python3.8/dist-packages/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.py", line 512, in _validate
raise OSError("Missing: " + ", ".join(missing_cmds))
OSError: Missing: valgrind, callgrind_control, callgrind_annotate

[svekars]

Thank you for notifying us about this issue! Can you please provide the following information:

• Link to the tutorial on the website
• Version of PyTorch used
• Environment in which the error occurs (i.e. Colab, macOS, Linux, etc.)

Thank you for notifying us about this issue! Can you please provide the following information:

• Link to the tutorial on the website
• Version of PyTorch used
• Environment in which the error occurs (i.e. Colab, macOS, Linux, etc.)

1. Link to the tutorial on the website
   https://pytorch.org/tutorials/recipes/recipes/benchmark.html

2. Version of PyTorch used

import torch
print(torch.version)
1.14.0a0+410ce96

3. Environment in which the error occurs (i.e. Colab, macOS, Linux, etc.)
   Linux-based Nvidia DGX OS with nvidia pytorch docker image (nvcr.io/nvidia/pytorch:22.12-py3 )

and we had below error also.

Traceback (most recent call last):
File "test8.py", line 30, in
os.path.join(os.getenv('CONDA_PREFIX'), "include")
File "/usr/lib/python3.8/posixpath.py", line 76, in join
a = os.fspath(a)
TypeError: expected str, bytes or os.PathLike object, not NoneType

[svekars]

cc: @robieta @malfet @albanD Can you take a look?

[albanD]

Hi,

I think the error just says that to be able to run this part of the tutorial you will need to install valgrind. You should be able to install it with your OS package manager.

[svekars]

I think the error is actually different:

TypeError                                 Traceback (most recent call last)
[<ipython-input-16-5ae88987f723>](https://localhost:8080/#) in <module>
     28     extra_include_paths=[
     29         # `load_inline` needs to know where to find Pybind11 headers.
---> 30         os.path.join(os.getenv('CONDA_PREFIX'), 'include')
     31     ],
     32     functions=['batched_dot_mul_sum_v0', 'batched_dot_mul_sum_v1']

[/usr/lib/python3.8/posixpath.py](https://localhost:8080/#) in join(a, *p)
     74     will be discarded.  An empty last part will result in a path that
     75     ends with a separator."""
---> 76     a = os.fspath(a)
     77     sep = _get_sep(a)
     78     path = a

TypeError: expected str, bytes or os.PathLike object, not NoneType

[albanD]

That is the a different error from the one at the top yes.

In this case,

tutorials/recipes_source/recipes/benchmark.py

Line 727 in b24c7c3

os.path.join(os.getenv('CONDA_PREFIX'), 'include')

does assume that the user is using conda. Which is most likely wrong (but a different problem from the one at the top).