# Introduction #
-This package provides a function that generates a dot file from the
-auto-grad graph.
+This package provides a function that generates a
+[dot file](https://en.wikipedia.org/wiki/DOT_(graph_description_language))
+from a [pytorch](http://pytorch.org) autograd graph.
# Usage #
### agtree2dot.save_dot(variable, variable_labels, result_file) ###
-Saves into `result_file` a dot file corresponding to the auto-grad graph for `variable`, which can be either a single `Variable` or a set of `Variable`s. The dictionary `variable_labels` associates strings to some variables, which will be used in the resulting graph.
+Saves into `result_file` a dot file corresponding to the autograd
+graph for the `Variable` `variable`. The dictionary `variable_labels`
+associates strings to some variables, which will be used in the
+resulting graph.
## Example ##
-A typical use would be:
+A typical use is provided in [mlp.py](https://fleuret.org/git-extract/agtree2dot/mlp.py):
```python
-import torch
+import subprocess
from torch import nn
from torch.nn import functional as fn
loss = criterion(output, target)
agtree2dot.save_dot(loss,
- { input: 'input', target: 'target', loss: 'loss' },
+ {
+ input: 'input',
+ target: 'target',
+ loss: 'loss',
+ mlp.fc1.weight: 'weight1',
+ mlp.fc1.bias: 'bias1',
+ mlp.fc2.weight: 'weight2',
+ mlp.fc2.bias: 'bias2',
+ },
open('./mlp.dot', 'w'))
-```
-which would generate a file mlp.dot, which can then be translated to pdf with
+print('Generated mlp.dot')
-```
-dot mlp.dot -Lg -T pdf -o mlp.pdf
+try:
+ subprocess.check_call(["dot", "mlp.dot", "-Lg", "-T", "pdf", "-o", "mlp.pdf" ])
+except subprocess.CalledProcessError:
+ print('Calling the dot command failed. Is Graphviz installed?')
+ sys.exit(1)
+
+print('Generated mlp.pdf')
```
-to produce [mlp.pdf](https://fleuret.org/git-extract/agtree2dot/mlp.pdf).
+which would generate a file mlp.dot and try to generate
+[mlp.pdf](https://fleuret.org/git-extract/agtree2dot/mlp.pdf) from it
+with [Graphviz tools.](http://www.graphviz.org/)