Skip to content

Add documentation to logging #26175

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
Krovatkin wants to merge 4 commits into from
Closed

Add documentation to logging #26175

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
06f5865
logging documentation
Krovatkin Sep 13, 2019
f449e7d
fix a copy/paste typo
Krovatkin Sep 13, 2019
897ff72
Update jit_log.h
Krovatkin Sep 13, 2019
9cdd6d2
fix a typo in currently
Krovatkin Sep 13, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 26 additions & 0 deletions torch/csrc/jit/docs/OVERVIEW.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1026,6 +1026,32 @@ with prim::DifferentiableGraph_0 = graph(%13 : Float(*, *),
return (%hy, %cy)
```

## JIT Logging ##

Logging is a very useful debugging technique, especially in the context of compilers. Compilers perform a series of passes and analyses and logging can help to trace issues such as wrong results or segmentation faults
all the way back to the original erroneous transformation.

`TorchScript` offers a simple logging facility that can enabled by setting an environment variable `PYTORCH_JIT_LOG_LEVEL`.

Logging is enabled on a per file basis. To enable logging in `dead_code_elimination.cpp`, `PYTORCH_JIT_LOG_LEVEL` should be
set to `dead_code_elimination.cpp` or, simply, to `dead_code_elimination` (i.e. `PYTORCH_JIT_LOG_LEVEL=dead_code_elimination`).

Multiple files can be logged by separating each file name with a colon `:` as in the following example, `PYTORCH_JIT_LOG_LEVEL=dead_code_elimination:guard_elimination`

There are 3 logging levels available for your use ordered by the detail level from lowest to highest.

* `GRAPH_DUMP` should be used for printing entire graphs after optimization passes
* `GRAPH_UPDATE` should be used for reporting graph transformations (i.e. node deletion, constant folding, etc)
* `GRAPH_DEBUG` should be used for providing information useful for debugging
the internals of a particular optimization pass or analysis

The current logging level is `GRAPH_UPDATE` meaning that both `GRAPH_DUMP` and `GRAPH_UPDATE` will be enabled when
one specifies a file(s) in `PYTORCH_JIT_LOG_LEVEL`.

`GRAPH_DEBUG` can be enabled by prefixing a file name with an `>` as in `>alias_analysis`.
`>>` and `>>>` are also valid and **currently** are equivalent to `GRAPH_DEBUG` as there is no logging level that is
higher than `GRAPH_DEBUG`.

## DifferentiableGraphOp ##

[graph_executor.cpp](../graph_executor.cpp)
Expand Down
37 changes: 30 additions & 7 deletions torch/csrc/jit/jit_log.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,15 +3,38 @@
#include <string>
#include <torch/csrc/WindowsTorchApiMacro.h>

// To enable logging please set(export) PYTORCH_JIT_LOG_LEVEL to
// the ordinal value of one of the following logging levels: 1 for GRAPH_DUMP,
// 2 for GRAPH_UPDATE, 3 for GRAPH_DEBUG.
// * Use GRAPH_DUMP for dumping graphs after optimization passes
// * Use GRAPH_UPDATE for reporting graph transformations (i.e. node deletion,
// constant folding, CSE)
// * Use GRAPH_DEBUG to provide information useful for debugging
// `TorchScript` offers a simple logging facility that can enabled by setting an
// environment variable `PYTORCH_JIT_LOG_LEVEL`.

// Logging is enabled on a per file basis. To enable logging in
// `dead_code_elimination.cpp`, `PYTORCH_JIT_LOG_LEVEL` should be
// set to `dead_code_elimination.cpp` or, simply, to `dead_code_elimination`
// (i.e. `PYTORCH_JIT_LOG_LEVEL=dead_code_elimination`).

// Multiple files can be logged by separating each file name with a colon `:` as
// in the following example,
// `PYTORCH_JIT_LOG_LEVEL=dead_code_elimination:guard_elimination`

// There are 3 logging levels available for your use ordered by the detail level
// from lowest to highest.

// * `GRAPH_DUMP` should be used for printing entire graphs after optimization
// passes
// * `GRAPH_UPDATE` should be used for reporting graph transformations (i.e.
// node deletion, constant folding, etc)
// * `GRAPH_DEBUG` should be used for providing information useful for debugging
// the internals of a particular optimization pass or analysis

// The current logging level is `GRAPH_UPDATE` meaning that both `GRAPH_DUMP`
// and `GRAPH_UPDATE` will be enabled when
// one specifies a file(s) in `PYTORCH_JIT_LOG_LEVEL`.

// `GRAPH_DEBUG` can be enabled by prefixing a file name with an `>` as in
// `>alias_analysis`.
// `>>` and `>>>` are also valid and **currently** are equivalent to
// `GRAPH_DEBUG` as there is no logging level that is
// higher than `GRAPH_DEBUG`.

namespace torch {
namespace jit {

Expand Down