LAP/torchlap/interpreting/attention_interpreter_smooth_integrator.py

from typing import Dict
from collections import OrderedDict as OrdDict

import torch
import torch.nn.functional as F

from .attention_interpreter import AttentionInterpreter


AttentionDict = Dict[torch.Size, torch.Tensor]


class AttentionInterpreterSmoothIntegrator(AttentionInterpreter):

    def _process_attentions(self) -> Dict[str, torch.Tensor]:
        sorted_attentions = {
            group_name: OrdDict(sorted(group_attention.items(),
                                           key=lambda pair: pair[0]))
            for group_name, group_attention in self._attention_groups.items()
        }

        results: Dict[str, torch.Tensor] = {group_name: None for group_name in sorted_attentions}
        for group_name, group_attention in sorted_attentions.items():
            self._impact = 1.0
            # from smallest to largest
            sum_ = 0
            for attention in group_attention.values():
                if torch.is_tensor(sum_):
                    sum_ = F.interpolate(sum_, size=attention.shape[-2:])
                sum_ += attention
                attention = (attention - 0.5).relu()
                results[group_name] = self._process_attention(results[group_name], attention)
            results[group_name] = torch.where(results[group_name] > 0, sum_, sum_ / (2 * len(group_attention)))

        interpretations = {}
        for group_name, group_result in results.items():
            group_result = F.interpolate(group_result, self._input_size, mode='bilinear', align_corners=True)
            interpretations[group_name] = group_result
            '''
            if group_result.shape[1] == 1:
                interpretations[group_name] = group_result
            else:
                interpretations.update({
                    f"{group_name}_{i}": group_result[:, i:i+1]
                    for i in range(group_result.shape[1])
                })
            '''
        return interpretations