-
Needed for out-of-training test data for the models and humans.
Aim for roughly half lower order and half higher order questions.
4 options each.
15 questions Josh
15 questions Rongbo
-
The Markov Attribution implemented uses a first order markov chain model to compute the removal effects for each channel. This underlying model assumes that the probability of the next future state de…
-
Thanks for the great work on pathpy! I do research on host intrusion detection with Linux system calls which can be nicely modeled with pathpy. However, a feature I am missing so far is the concept of…
-
The syntax for template names is kind of like a single method interface.
The method parameter types are based on known identifiers.
The method signature itself can be a call on an interface but inst…
-
### 🐛 Describe the bug
Hi,
I'm trying to make my [MoE Triton kernel](https://github.com/RobertCsordas/moe_layer/blob/master/triton_src/moe_layer/cvmm.py) work with torch.compile(). I know that thi…
-
Has anyone tried downscaling the K and/or Q matrices for repeated layers in franken-merges? This should act like changing the temperature of the softmax and effectively smooth the distribution:
**H…
-
# Plotting partial pooling in mixed-effects models - Higher Order Functions
There are a lot of bilabial sounds in that title
[https://tjmahr.github.io/plotting-partial-pooling-in-mixed-effects-model…
-
#### Summary:
Add testing for higher order autodiff. We need to be able to determine where we aren't able to compile models in higher order.
This goes along with stan-dev/math#403.
#### Descrip…
-
**Application**
The class naming of the polynomial classes seems inconsistent to me ("model" vs. "polynomial"). For example, consider the higher order polynomial class (`BinaryPolynomial`) vs. the qu…
-
### Is this a new bug in dbt-core?
- [X] I believe this is a new bug in dbt-core
- [X] I have searched the existing issues, and I could not find an existing issue for this bug
### Current Behav…