-
I was looking through the code the other day and some of it appears to assume the dataset can fit in memory as a numpy array. My dataset is 0.5 TB so unfortunately that won't work. Am I mistaken or wi…
-
I'm implementing a classifier[1] which does *concept learning*, i.e. unsymmetrical binary classification with a (sometimes user-defined) *target/positive class* and the learned model describing that c…
-
Hello there, as i cloned your repo and added my own train.csv file and trained my data with reddit fastText now i am testing my data using ./intent_classifier.py config.json , but when i gave my pre-d…
-
Using wfdb as a python package to read the waveforms, data preprocess is required.
Following steps should be taken:
- Noise cleaning: removing data spikes that reach zero, and search how to approach…
-
Post your response to our challenge questions.
Think about how Large Language Models (LLMs) may assist with your planned final project. For example, they could help in labeling data, searching for …
lkcao updated
8 months ago
-
Bad documenttaion. not very long errors
Detecting toxicity in outputs generated by Large Language Models (LLMs) is crucial for ensuring that these models produce safe, respectful, and appropriate con…
-
Dear developers,
I was looking for a Semi-Supervised ML method in R and found your excellent package. I tried your example code adapting it to my input data, and after some reformating it works app…
-
This package is very useful, thanks a lot for providing it! I would like to use it in a large open-source project ([3D Slicer](https://github.com/Slicer/Slicer)) but I'm not sure if I can rely on it i…
-
I recently migrated from metwork v2.1 to v2.2. So I rebuilt only the plugin with embedded python packages (`local/lib/python3.X/site-packages`) because of the python upgrade from v3.10 to v3.11. It se…
-
make sure in the maintainer section that we include discussions of what versions of python to support. the scientific python spec 00 is what we want to ideally promote here. see comment thread below:
…