Toby Roseman
Toby Roseman
With this change, test_api_visibilities.py passes when run in isolation.
Don't assume source predictions are always stored (in the data frame) in a column called `prediction`. Instead, assume that column name matches the key in the dictionary returned by `predict`,...
One TensorFlow 1.X unit test that has been consistently failing on our CI: `coremltools/converters/mil/frontend/tensorflow/test/test_ops.py::TestSliceByIndex::test_slice_by_index[[use_cpu_only=False]-mlprogram-fp16-[rank=1]-[masking=True]]` Logs: https://gitlab.com/zach_nation/coremltools/-/jobs/2254707322 Looking at when this started to failure, it doesn't seem related to any change...
Currently there is no easy way to remove a row in an SFrame/SArray, given the row number. I think we should update our `__del__` methods: `del my_sframe[]` If `value` is...
If a pandas data frame has an object column that contains NaN value, we can not convert it to an SFrame, and we get an unhelpful error message. Turicreate version:...
SArray has [a `filter` method](https://apple.github.io/turicreate/docs/api/generated/turicreate.SArray.filter.html#turicreate.SArray.filter) which takes a function to filter its content. SFrame should have this functionality too. Simple example: ``` sf = tc.SFrame({'num': range(3), 'letter': ['a','b','c']}) sf.filter(lambda row:...
The `evaluate` method has a number of related issues: - [ ] The `metric` parameter is largely ignored. Many metrics are calculated regardless of the value of `metric`. For example...
[The method](https://github.com/apple/turicreate/blob/c8be17cb80322ff568f1facd76168077aacca536/src/python/turicreate/toolkits/_evaluate_utils.py#L21) for calculating a confusion matrix is quite slow. Especially when there is a large number of classes. This seems to primarily be caused by [appending](https://github.com/apple/turicreate/blob/c8be17cb80322ff568f1facd76168077aacca536/src/python/turicreate/toolkits/_evaluate_utils.py#L53) potentially large SFrames...