-
I wonder whether torchtune can support traditional tasks such as translation or more general text generation tasks which have a input and output column. I have read the datasets doc at [here](https://…
-
It should be possible to specify a top level node that should be displayed in the class hierarchy. For example, to show all three classifications in iCAT, the top-level node should be "Classification …
-
## Overview
As per discussion on #2893, it looks the shape of `shap_values` returned by TreeExplainer is inconsistent. It has:
- shape `(n_samples, n_features)` if the background dataset is passed…
-
in the demo notebook I implemented a simple majority vote based on what has been done in HRL, but this still requires further testing and tuning.
In HRL in the end we used a more advanced technique ba…
-
Similar to #695 (closed). In the original discussion, participants did not seem to understand the difference between "learning-to-rank" and "ordinal classification". "learning-to-rank" produce relativ…
-
### Have I written custom code (as opposed to using a stock example script provided in MediaPipe)
No
### OS Platform and Distribution
Ubuntu 22.04
### Python Version
3.10
### MediaPi…
-
I tried using your script for zero shot classification together with the pretrained weights (both resnet18 and resnet50). The calssification results I got are very random (accuracy 17-22% for each cla…
-
Hello, I noticed that in the slide level fine-tuning implementation, the `CrossEntropyLoss` loss function and the `Softmax` activation function are used in binary classification. This seems to be diff…
-
-
Hello, Thank you for sharing the code. I am using your code for the classification of my custom data. The image size of the dataset is just 64x64 and has 8 classes.
I just want to know, if can you p…