captum
captum copied to clipboard
pytorch
Model interpretability and understanding for PyTorch
Without the packaging library, statements like: `"1.8.0" > "1.10.0"` will be equal to True, despite v1.10 being a later version that v1.8.0. The `packaging` library will in some cases not...
The warning messages take up a lot of space on the console log, and it was really simple to resolve them. The common.rst file was also incorrectly pointing to the...
## 🐛 Bug Running the tutorial [Interpret regression models using Boston House Prices Dataset](https://captum.ai/tutorials/House_Prices_Regression_Interpret) results in an `IndexError` in the last cell ## To Reproduce Steps to reproduce the behavior:...
This switches usage of full backward hooks to instead apply forward hooks which then add tensor backward hooks, as suggested in #914 . We initially did not choose this approach...
I think that the amount of Conda install timeouts can be lessened by removing unnecessary installs: * Sphinx is not used for the Conda tests, so it doesn't make sense...
This PR does the following: * Implements the suggested changed from: https://github.com/pytorch/captum/pull/831. * Ensures that variable names are consistent across objectives. * Lets `batch_index` support 2 values in addition to...
## 📚 Documentation In the Titanic Data Analysis tutorial, the use of CrossEntropyLoss is inappropriate. From the documentation of pytorch, CrossEntropyLoss is a combination of LogSoftmax and NLLLoss, so the...
Thank you for providing this package for explaining models! I used lime to explain a FM-based model for recommendation.And the code is as follows: `explainer = Lime(model_initial) attr = explainer.attribute(inputs=model_input_lime,...
Summary: the tests for `test_tracin_show_progress` were failing we check that the progress reaches 100% X times, but sometimes, the progress reaches 100% more than X times. As aobo-y pointed out,...
