From 567998d8f780bdac3412d729a6d183a99832f7fb Mon Sep 17 00:00:00 2001
From: Sun Jin Kim <sk2521@ic.ac.uk>
Date: Mon, 25 Apr 2022 18:48:50 +0100
Subject: [PATCH] fully implement `exclude_method` in aa_learners
---
.../autoaugment_learners/aa_learner.py | 54 +++++++++----------
.../autoaugment_learners/evo_learner.py | 30 +++++------
.../autoaugment_learners/gru_learner.py | 24 ++-------
.../randomsearch_learner.py | 42 +++++----------
.../autoaugment_learners/ucb_learner.py | 14 +++--
5 files changed, 66 insertions(+), 98 deletions(-)
diff --git a/MetaAugment/autoaugment_learners/aa_learner.py b/MetaAugment/autoaugment_learners/aa_learner.py
index 44a86aa3..561222a5 100644
--- a/MetaAugment/autoaugment_learners/aa_learner.py
+++ b/MetaAugment/autoaugment_learners/aa_learner.py
@@ -1,4 +1,3 @@
-from numpy import isin
import torch
import torch.nn as nn
import torch.optim as optim
@@ -7,31 +6,10 @@ from MetaAugment.autoaugment_learners.autoaugment import AutoAugment
import torchvision.transforms as transforms
-from pprint import pprint
-import matplotlib.pyplot as plt
import copy
import types
-# We will use this augmentation_space temporarily. Later on we will need to
-# make sure we are able to add other image functions if the users want.
-augmentation_space = [
- # (function_name, do_we_need_to_specify_magnitude)
- ("ShearX", True),
- ("ShearY", True),
- ("TranslateX", True),
- ("TranslateY", True),
- ("Rotate", True),
- ("Brightness", True),
- ("Color", True),
- ("Contrast", True),
- ("Sharpness", True),
- ("Posterize", True),
- ("Solarize", True),
- ("AutoContrast", False),
- ("Equalize", False),
- ("Invert", False),
- ]
class aa_learner:
@@ -96,8 +74,30 @@ class aa_learner:
# TODO: We should probably use a different way to store results than self.history
self.history = []
- self.augmentation_space = [x for x in augmentation_space if x not in exclude_method]
- self.fun_num = len(augmentation_space)
+
+ # this is the full augmentation space. We take out some image functions
+ # if the user specifies so in the exclude_method parameter
+ augmentation_space = [
+ # (function_name, do_we_need_to_specify_magnitude)
+ ("ShearX", True),
+ ("ShearY", True),
+ ("TranslateX", True),
+ ("TranslateY", True),
+ ("Rotate", True),
+ ("Brightness", True),
+ ("Color", True),
+ ("Contrast", True),
+ ("Sharpness", True),
+ ("Posterize", True),
+ ("Solarize", True),
+ ("AutoContrast", False),
+ ("Equalize", False),
+ ("Invert", False),
+ ]
+ self.exclude_method = exclude_method
+ self.augmentation_space = [x for x in augmentation_space if x[0] not in exclude_method]
+
+ self.fun_num = len(self.augmentation_space)
self.op_tensor_length = self.fun_num + p_bins + m_bins if discrete_p_m else self.fun_num +2
@@ -174,7 +174,7 @@ class aa_learner:
prob_idx = torch.multinomial(prob_t, 1).item() # 0 <= p <= 10
mag = torch.multinomial(mag_t, 1).item() # 0 <= m <= 9
- function = augmentation_space[fun_idx][0]
+ function = self.augmentation_space[fun_idx][0]
prob = prob_idx/(self.p_bins-1)
indices = (fun_idx, prob_idx, mag)
@@ -203,13 +203,13 @@ class aa_learner:
prob = round(prob, 1) # round to nearest first decimal digit
mag = round(mag) # round to nearest integer
- function = augmentation_space[fun_idx][0]
+ function = self.augmentation_space[fun_idx][0]
assert 0 <= prob <= 1, prob
assert 0 <= mag <= self.m_bins-1, (mag, self.m_bins)
# if the image function does not require a magnitude, we set the magnitude to None
- if augmentation_space[fun_idx][1] == True: # if the image function has a magnitude
+ if self.augmentation_space[fun_idx][1] == True: # if the image function has a magnitude
operation = (function, prob, mag)
else:
operation = (function, prob, None)
diff --git a/MetaAugment/autoaugment_learners/evo_learner.py b/MetaAugment/autoaugment_learners/evo_learner.py
index 682061ef..34cc2d44 100644
--- a/MetaAugment/autoaugment_learners/evo_learner.py
+++ b/MetaAugment/autoaugment_learners/evo_learner.py
@@ -1,4 +1,3 @@
-from cgi import test
import torch
torch.manual_seed(0)
import torch.nn as nn
@@ -7,19 +6,18 @@ import pygad.torchga as torchga
import copy
import torch
-from MetaAugment.autoaugment_learners.aa_learner import aa_learner, augmentation_space
-import MetaAugment.child_networks as cn
+from MetaAugment.autoaugment_learners.aa_learner import aa_learner
class evo_learner(aa_learner):
def __init__(self,
# search space settings
- discrete_p_m=False,
- exclude_method=[],
sp_num=5,
p_bins=10,
m_bins=10,
+ discrete_p_m=False,
+ exclude_method=[],
# child network settings
learning_rate=1e-1,
max_epochs=float('inf'),
@@ -32,16 +30,18 @@ class evo_learner(aa_learner):
controller=None
):
- super().__init__(sp_num=sp_num,
- p_bins=p_bins,
- m_bins=m_bins,
- discrete_p_m=discrete_p_m,
- batch_size=batch_size,
- toy_size=toy_size,
- learning_rate=learning_rate,
- max_epochs=max_epochs,
- early_stop_num=early_stop_num,
- exclude_method=exclude_method)
+ super().__init__(
+ sp_num=sp_num,
+ p_bins=p_bins,
+ m_bins=m_bins,
+ discrete_p_m=discrete_p_m,
+ batch_size=batch_size,
+ toy_size=toy_size,
+ learning_rate=learning_rate,
+ max_epochs=max_epochs,
+ early_stop_num=early_stop_num,
+ exclude_method=exclude_method
+ )
self.num_solutions = num_solutions
self.controller = controller
diff --git a/MetaAugment/autoaugment_learners/gru_learner.py b/MetaAugment/autoaugment_learners/gru_learner.py
index 6955257f..5c15a4a4 100644
--- a/MetaAugment/autoaugment_learners/gru_learner.py
+++ b/MetaAugment/autoaugment_learners/gru_learner.py
@@ -9,25 +9,6 @@ import pickle
-# We will use this augmentation_space temporarily. Later on we will need to
-# make sure we are able to add other image functions if the users want.
-augmentation_space = [
- # (function_name, do_we_need_to_specify_magnitude)
- ("ShearX", True),
- ("ShearY", True),
- ("TranslateX", True),
- ("TranslateY", True),
- ("Rotate", True),
- ("Brightness", True),
- ("Color", True),
- ("Contrast", True),
- ("Sharpness", True),
- ("Posterize", True),
- ("Solarize", True),
- ("AutoContrast", False),
- ("Equalize", False),
- ("Invert", False),
- ]
class gru_learner(aa_learner):
@@ -50,6 +31,7 @@ class gru_learner(aa_learner):
p_bins=11,
m_bins=10,
discrete_p_m=False,
+ exclude_method=[],
# hyperparameters for when training the child_network
batch_size=8,
toy_size=1,
@@ -85,7 +67,9 @@ class gru_learner(aa_learner):
toy_size=toy_size,
learning_rate=learning_rate,
max_epochs=max_epochs,
- early_stop_num=early_stop_num,)
+ early_stop_num=early_stop_num,
+ exclude_method=exclude_method,
+ )
# GRU-specific attributes that aren't in general aa_learner's
self.alpha = alpha
diff --git a/MetaAugment/autoaugment_learners/randomsearch_learner.py b/MetaAugment/autoaugment_learners/randomsearch_learner.py
index 71d8bc1a..2c35fb80 100644
--- a/MetaAugment/autoaugment_learners/randomsearch_learner.py
+++ b/MetaAugment/autoaugment_learners/randomsearch_learner.py
@@ -10,25 +10,7 @@ import pickle
-# We will use this augmentation_space temporarily. Later on we will need to
-# make sure we are able to add other image functions if the users want.
-augmentation_space = [
- # (function_name, do_we_need_to_specify_magnitude)
- ("ShearX", True),
- ("ShearY", True),
- ("TranslateX", True),
- ("TranslateY", True),
- ("Rotate", True),
- ("Brightness", True),
- ("Color", True),
- ("Contrast", True),
- ("Sharpness", True),
- ("Posterize", True),
- ("Solarize", True),
- ("AutoContrast", False),
- ("Equalize", False),
- ("Invert", False),
- ]
+
class randomsearch_learner(aa_learner):
"""
@@ -41,6 +23,7 @@ class randomsearch_learner(aa_learner):
p_bins=11,
m_bins=10,
discrete_p_m=True,
+ exclude_method=[],
# hyperparameters for when training the child_network
batch_size=8,
toy_size=1,
@@ -49,15 +32,18 @@ class randomsearch_learner(aa_learner):
early_stop_num=30,
):
- super().__init__(sp_num=sp_num,
- p_bins=p_bins,
- m_bins=m_bins,
- discrete_p_m=discrete_p_m,
- batch_size=batch_size,
- toy_size=toy_size,
- learning_rate=learning_rate,
- max_epochs=max_epochs,
- early_stop_num=early_stop_num,)
+ super().__init__(
+ sp_num=sp_num,
+ p_bins=p_bins,
+ m_bins=m_bins,
+ discrete_p_m=discrete_p_m,
+ batch_size=batch_size,
+ toy_size=toy_size,
+ learning_rate=learning_rate,
+ max_epochs=max_epochs,
+ early_stop_num=early_stop_num,
+ exclude_method=exclude_method
+ )
def generate_new_discrete_operation(self):
diff --git a/MetaAugment/autoaugment_learners/ucb_learner.py b/MetaAugment/autoaugment_learners/ucb_learner.py
index aa6cd685..fdf735be 100644
--- a/MetaAugment/autoaugment_learners/ucb_learner.py
+++ b/MetaAugment/autoaugment_learners/ucb_learner.py
@@ -1,15 +1,9 @@
import numpy as np
-import torch
-import torch.nn as nn
-import torch.optim as optim
-import torchvision
from tqdm import trange
from ..child_networks import *
-from ..main import train_child_network
from .randomsearch_learner import randomsearch_learner
-from .aa_learner import augmentation_space
class ucb_learner(randomsearch_learner):
@@ -23,6 +17,7 @@ class ucb_learner(randomsearch_learner):
p_bins=11,
m_bins=10,
discrete_p_m=True,
+ exclude_method=[],
# hyperparameters for when training the child_network
batch_size=8,
toy_size=1,
@@ -33,7 +28,8 @@ class ucb_learner(randomsearch_learner):
num_policies=100
):
- super().__init__(sp_num=sp_num,
+ super().__init__(
+ sp_num=sp_num,
p_bins=p_bins,
m_bins=m_bins,
discrete_p_m=discrete_p_m,
@@ -41,7 +37,9 @@ class ucb_learner(randomsearch_learner):
toy_size=toy_size,
learning_rate=learning_rate,
max_epochs=max_epochs,
- early_stop_num=early_stop_num,)
+ early_stop_num=early_stop_num,
+ exclude_method=exclude_method,
+ )
--
GitLab