gdb special command to print tensors #54339

antocuni · 2021-03-19T17:36:55Z

This is something which I wrote because it was useful during my debugging sessions, but I think it might be generally useful to other people as well so I took the liberty of proposing an official pytorch-gdb extension.

pytorch-gdb is a gdb script written in python. Currently, it contains only one command: torch-tensor-repr, which prints a human-readable repr of an at::Tensor object. Example:

Breakpoint 1, at::native::neg (self=...) at [...]/pytorch/aten/src/ATen/native/UnaryOps.cpp:520
520     Tensor neg(const Tensor& self) { return unary_op_impl(self, at::neg_out); }
(gdb) # the default repr of 'self' is not very useful
(gdb) p self
$1 = (const at::Tensor &) @0x7ffff72ed780: {impl_ = {target_ = 0x5555559df6e0}}
(gdb) torch-tensor-repr self
Python-level repr of self:
tensor([1., 2., 3., 4.], dtype=torch.float64)

The idea is that by having an official place where to put these things, pytorch-gdb will slowly grow other useful features and make the pytorch debugging experience nicer and faster.

…epr command

facebook-github-bot · 2021-03-19T18:05:34Z

💊 CI failures summary and remediations

As of commit 4c217dd (more details on the Dr. CI page):

💚 💚 Looks good so far! There are no failures yet. 💚 💚

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

ezyang · 2021-03-23T03:22:11Z

torch/csrc/utils.cpp

+  PyGILState_Release(gil);
+  return result;
+
+error:


Not using RAII here is kind of weird. Is there a particular reason, e.g., gdb doesn't like it when you have stack unwinding?

the most honest answer is that I am used to write Python API code in C and I didn't think that this was C++ 🤦‍♂️

But also, I am not aware of any standard smart-pointer-like class to manage PyObject* in C++ with RAII. Do we have any helper for that?

Or you were suggesting do write a custom class just to handle this specific function?

facebook-github-bot · 2021-03-23T03:22:57Z

@ezyang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

facebook-github-bot · 2021-03-23T19:32:59Z

@ezyang merged this pull request in 21a9a93.

Summary: #54339 broke Flake8. This PR fixes it. Pull Request resolved: #54540 Test Plan: ``` flake8 ``` Reviewed By: walterddr Differential Revision: D27274171 Pulled By: samestep fbshipit-source-id: 4b440d72b4b5615f45e6fcb25f7a4c0423add272

wwwwcu · 2021-07-02T08:41:20Z

I saw the tool (pytorhc-gdb), so I want to have a try.
Following the guide, I can't get the value of at::tensor since my code walk into libtorch_cpu.so instead of UnaryOps.cpp:520

gdb python
GNU gdb (Ubuntu 8.1-0ubuntu3.2) 8.1.0.20180409-git
[...]
(gdb) break at::native::negative
Function "at::native::negative" not defined.
Make breakpoint pending on future shared library load? (y or [n]) y
Breakpoint 1 (at::native::negative) pending.
(gdb) run
Starting program: /root/anaconda3/envs/pytorch/bin/python 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Python 3.6.10 |Anaconda, Inc.| (default, May  8 2020, 02:54:21) 
[GCC 7.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> t = torch.tensor([1, 2, 3, 4], dtype=torch.float64)
>>> t.negative()

Breakpoint 1, 0x00007fffe32c5f20 in at::native::negative(at::Tensor const&)@plt () from /root/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/lib/libtorch_cpu.so

I built pytorch from source with "DEBUG=1", I'm not sure if I use the tool correctly.

antocuni · 2021-07-02T14:22:55Z

I saw the tool (pytorhc-gdb), so I want to have a try.
Following the guide, I can't get the value of at::tensor since my code walk into libtorch_cpu.so instead of UnaryOps.cpp:520

you are correct, the example was out of date. I fixed it in #61175

…1175) Summary: As pointed out by #54339 (comment), the `pytorch-gdb` example is currently broken because the code has been refactored. This PR updates the example so that it works again. Pull Request resolved: #61175 Reviewed By: heitorschueroff Differential Revision: D29660336 Pulled By: ezyang fbshipit-source-id: 8bcd32fc583c0b28a705ef37203ce7ad4d636732

antocuni added 3 commits Mar 19, 2021

introduce the pytorch-gdb extension, and implement the torch-tensor-r…

a266c5d

…epr command

[ci skip] add some documentation about pytorch-gdb

31e387d

[ci skip] small tweaks

4c217dd

facebook-github-bot added the cla signed label Mar 19, 2021

pytorchbot added the open source label Mar 19, 2021

heitorschueroff requested review from ezyang and gchanan Mar 22, 2021

heitorschueroff added the triaged label Mar 22, 2021

ezyang reviewed Mar 23, 2021

View changes

ezyang approved these changes Mar 23, 2021

View changes

facebook-github-bot closed this in pytorch/pytorch@21a9a93 Mar 23, 2021

facebook-github-bot added the Merged label Mar 23, 2021

samestep mentioned this pull request Mar 23, 2021

Fix Flake8 #54540

Closed

antocuni mentioned this pull request Jul 2, 2021

update the pytorch-gdb example so that it works on current master #61175

Closed

pytorch / pytorch Public

gdb special command to print tensors #54339

gdb special command to print tensors #54339

antocuni commented Mar 19, 2021

facebook-github-bot commented Mar 19, 2021 •

edited

ezyang Mar 23, 2021

antocuni Mar 30, 2021

facebook-github-bot commented Mar 23, 2021

facebook-github-bot commented Mar 23, 2021

wwwwcu commented Jul 2, 2021 •

edited

antocuni commented Jul 2, 2021

pytorch / pytorch Public

gdb special command to print tensors #54339

gdb special command to print tensors #54339

Conversation

antocuni commented Mar 19, 2021

facebook-github-bot commented Mar 19, 2021 • edited

💊 CI failures summary and remediations

ezyang Mar 23, 2021

Choose a reason for hiding this comment

antocuni Mar 30, 2021

Choose a reason for hiding this comment

facebook-github-bot commented Mar 23, 2021

facebook-github-bot commented Mar 23, 2021

wwwwcu commented Jul 2, 2021 • edited

antocuni commented Jul 2, 2021

facebook-github-bot commented Mar 19, 2021 •

edited

wwwwcu commented Jul 2, 2021 •

edited