Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
File size: 4,212 Bytes
1f53a4c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 |
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import pandas as pd
import torch
from typing import List, Dict
class Likelihood:
"""
Class for making likelihood.
"""
def __init__(self, task: str, num_outputs_for_label: Dict[str, int]) -> None:
"""
Args:
task (str): task
num_outputs_for_label (Dict[str, int]): number of classes for each label
"""
self.task = task
self.num_outputs_for_label = num_outputs_for_label
self.base_column_list = self._set_base_columns(self.task)
self.pred_column_list = self._make_pred_columns(self.task, self.num_outputs_for_label)
def _set_base_columns(self, task: str) -> List[str]:
"""
Return base columns.
Args:
task (str): task
Returns:
List[str]: base columns except columns of label and prediction
"""
if (task == 'classification') or (task == 'regression'):
base_columns = ['uniqID', 'group', 'imgpath', 'split']
return base_columns
elif task == 'deepsurv':
base_columns = ['uniqID', 'group', 'imgpath', 'split', 'periods']
return base_columns
else:
raise ValueError(f"Invalid task: {task}.")
def _make_pred_columns(self, task: str, num_outputs_for_label: Dict[str, int]) -> Dict[str, List[str]]:
"""
Make column names of predictions with label name and its number of classes.
Args:
task (str): task
num_outputs_for_label (Dict[str, int]): number of classes for each label
Returns:
Dict[str, List[str]]: label and list of columns of predictions with its class number
eg.
{label_A: 2, label_B: 2} -> {label_A: [pred_label_A_0, pred_label_A_1], label_B: [pred_label_B_0, pred_label_B_1]}
{label_A: 1, label_B: 1} -> {label_A: [pred_label_A], label_B: [pred_label_B]}
"""
pred_columns = dict()
if task == 'classification':
for label_name, num_classes in num_outputs_for_label.items():
pred_columns[label_name] = ['pred_' + label_name + '_' + str(i) for i in range(num_classes)]
return pred_columns
elif (task == 'regression') or (task == 'deepsurv'):
for label_name, num_classes in num_outputs_for_label.items():
pred_columns[label_name] = ['pred_' + label_name]
return pred_columns
else:
raise ValueError(f"Invalid task: {task}.")
def make_format(self, data: Dict, output: Dict[str, torch.Tensor]) -> pd.DataFrame:
"""
Make a new DataFrame of likelihood every batch.
Args:
data (Dict): batch data from dataloader
output (Dict[str, torch.Tensor]): output of model
"""
_likelihood = {column_name: data[column_name] for column_name in self.base_column_list}
df_likelihood = pd.DataFrame(_likelihood)
if any(data['labels']):
for label_name, pred in output.items():
_df_label = pd.DataFrame({label_name: data['labels'][label_name].tolist()})
pred = pred.to('cpu').detach().numpy().copy()
_df_pred = pd.DataFrame(pred, columns=self.pred_column_list[label_name])
df_likelihood = pd.concat([df_likelihood, _df_label, _df_pred], axis=1)
return df_likelihood
else:
for label_name, pred in output.items():
pred = pred.to('cpu').detach().numpy().copy()
_df_pred = pd.DataFrame(pred, columns=self.pred_column_list[label_name])
df_likelihood = pd.concat([df_likelihood, _df_pred], axis=1)
return df_likelihood
def set_likelihood(task: str, num_outputs_for_label: Dict[str, int]) -> Likelihood:
"""
Set likelihood.
Args:
task (str): task
num_outputs_for_label (Dict[str, int]): number of classes for each label
Returns:
Likelihood: instance of class Likelihood
"""
return Likelihood(task, num_outputs_for_label)
|