pulsatrix 1.1.0
Loading...
Searching...
No Matches
pulsatrix::RelevanceDiff Struct Reference

The two explanations of the same prediction, and their difference. More...

#include <fine_tune_diff.hpp>

Collaboration diagram for pulsatrix::RelevanceDiff:

Public Attributes

Tensor tuned
 The fine-tuned model's input relevance, as its propagate_relevance() returns it ((N, L) per token for a language model).
 
Tensor base
 The base model's, for the same class at the same position.
 
Tensor difference
 tuned - base.
 
std::vector< int64_t > classes
 The class explained for each sample.
 
std::vector< double > tuned_scores
 Each model's output at that class, per sample: the relevance each one distributes.
 
std::vector< double > base_scores
 

Detailed Description

The two explanations of the same prediction, and their difference.

Member Data Documentation

◆ base

Tensor pulsatrix::RelevanceDiff::base

The base model's, for the same class at the same position.

◆ base_scores

std::vector<double> pulsatrix::RelevanceDiff::base_scores

◆ classes

std::vector<int64_t> pulsatrix::RelevanceDiff::classes

The class explained for each sample.

◆ difference

Tensor pulsatrix::RelevanceDiff::difference

tuned - base.

◆ tuned

Tensor pulsatrix::RelevanceDiff::tuned

The fine-tuned model's input relevance, as its propagate_relevance() returns it ((N, L) per token for a language model).

◆ tuned_scores

std::vector<double> pulsatrix::RelevanceDiff::tuned_scores

Each model's output at that class, per sample: the relevance each one distributes.


The documentation for this struct was generated from the following file: