Relevance per residue for one explained value.
More...
#include <protein_explanations.hpp>
|
| double | total () const |
| | The relevance of every token, <cls> and <eos> included.
|
| |
|
| std::string | sequence |
| |
| std::string | target |
| |
| float | value = 0.0f |
| | The explained value: a logit, a log-odds or a head's output.
|
| |
| std::vector< float > | residues |
| | One value per residue (the input tokens between <cls> and <eos>).
|
| |
| float | cls = 0.0f |
| |
| float | eos = 0.0f |
| |
| std::vector< std::vector< float > > | layers |
| | Relevance per token (<cls>, residues, <eos>) at the embeddings and after each layer: (layers + 1) rows of L + 2.
|
| |
Relevance per residue for one explained value.
◆ total()
| double pulsatrix::ResidueRelevance::total |
( |
| ) |
const |
The relevance of every token, <cls> and <eos> included.
◆ cls
| float pulsatrix::ResidueRelevance::cls = 0.0f |
◆ eos
| float pulsatrix::ResidueRelevance::eos = 0.0f |
◆ layers
| std::vector<std::vector<float> > pulsatrix::ResidueRelevance::layers |
Relevance per token (<cls>, residues, <eos>) at the embeddings and after each layer: (layers + 1) rows of L + 2.
◆ residues
| std::vector<float> pulsatrix::ResidueRelevance::residues |
One value per residue (the input tokens between <cls> and <eos>).
◆ sequence
| std::string pulsatrix::ResidueRelevance::sequence |
◆ target
| std::string pulsatrix::ResidueRelevance::target |
◆ value
| float pulsatrix::ResidueRelevance::value = 0.0f |
The explained value: a logit, a log-odds or a head's output.
The documentation for this struct was generated from the following file: