Merge branch 'main' of github.com:spfrommer/torchexplorer

README.md

@@ -39,11 +39,26 @@ torchexplorer.watch(model, backend='wandb') # Or 'standalone'
 ### Install
 Installing requires one external `graphviz` dependency, which should be available on most package managers.
 
+_Linux_.
 ```bash
 sudo apt-get install libgraphviz-dev graphviz
 pip install torchexplorer
 ```
-For Mac, `brew install graphviz` should suffice. If the `pygraphviz` wheel build fails because it can't find `Python.h`, you must install the python header files as described [here](https://stackoverflow.com/a/22077790/4864247).
+If the `pygraphviz` wheel build fails because it can't find `Python.h`, you must install the python header files as described [here](https://stackoverflow.com/a/22077790/4864247).
+
+_Mac_.
+```bash
+brew install graphviz
+pip install torchexplorer
+```
+If there's an error regarding `#include "graphviz/cgraph.h"`, [the following](https://github.com/pygraphviz/pygraphviz/issues/11#issuecomment-1038479834) worked for me on Apple silicon:
+```bash
+python -m pip install \
+ --global-option=build_ext \
+ --global-option="-I$(brew --prefix graphviz)/include/" \
+ --global-option="-L$(brew --prefix graphviz)/lib/" \
+ pygraphviz
+```
 
 ### Usage
 

torchexplorer/api/api.py

@@ -49,8 +49,8 @@ def watch(
  log_freq: int = 500,
  ignore_io_grad_classes: list[type] = [],
  disable_inplace: bool = False,
- bins: int = 20,
- sample_n: Optional[int] = 100,
+ bins: int = 30,
+ sample_n: Optional[int] = 1000,
  reject_outlier_proportion: float = 0.1,
  time_log: tuple[str, Callable] = ('step', lambda module, step: step),
  delay_log_multi_backward: bool = False,

torchexplorer/vega/mock_vega_data.py

@@ -48,7 +48,7 @@ def main():
  model = torchvision.models.resnet18()
  inplace_classes = [torchvision.models.resnet.BasicBlock]
  structure_wrapper = api.watch(
- model, log_freq=1, backend='none',
+ model, log_freq=1, backend='none', bins=20, sample_n=500,
  ignore_io_grad_classes=inplace_classes, disable_inplace=True
  )
  X, y = torch.randn(5, 3, 32, 32), torch.randn(5, 1000)
@@ -65,7 +65,7 @@ def main():
  optimizer = optim.Adam(model.parameters(), lr=1e-2)
  loss_fn = torch.nn.MSELoss()
 
- for step in range(2):
+ for step in range(10):
  y_hat = model(X)
  loss = loss_fn(y_hat, y)
  loss.backward()

torchexplorer/vega/vega_dataless.json

@@ -1,3 +1,4 @@
+
 {
  "$schema": "https://vega.github.io/schema/vega/v5.json",
  "description": "A neural network visualizer.",
@@ -1448,10 +1449,15 @@
  }
  },
  "transform": [
+ {
+ "type": "formula",
+ "as": "x",
+ "expr": "datum.x+1"
+ },
  {
  "type": "formula",
  "as": "width",
- "expr": "datum.width+1"
+ "expr": "datum.width-2"
  },
  {
  "type": "formula",
@@ -1556,7 +1562,8 @@
  "enter": {
  "stroke": {"value": "#000"},
  "strokeWidth": {"value": 2.5},
- "strokeDash": {"value": [8,2]}
+ "strokeDash": {"value": [7,4]},
+ "strokeOpacity": {"value": 1.0}
  },
  "update": {
  "x": {"signal": "datum.histogram_x + histogram_margin + histogram_inner_width"},