Uploading files

.gitignore

@@ -0,0 +1,141 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+cfgs_aba/
+scratch/
+features
+wandb 
+# C extensions
+*.so
+output
+ckpt
+data
+*.pkl
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+*.ipynb
+*.sh
+classification/scripts/
+classification/output*
+classification_test/
+classification/visualize.ipynb

README.md

@@ -0,0 +1,102 @@
+# ReservoirTTA
+
+<a href="https://arxiv.org/pdf/2505.14511?"><img src="https://img.shields.io/badge/arxiv-orange"></a>
+
+This repository contains the official Pytorch implementation of under review paper: "ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains"
+
+## Abstract
+This paper introduces **ReservoirTTA**, a novel plug–in framework designed for prolonged test–time adaptation (TTA) in scenarios where the test domain continuously shifts over time, including cases where domains recur or evolve gradually. At its core, ReservoirTTA maintains a reservoir of domain-specialized models—an adaptive test-time model ensemble—that both detects new domains via online clustering over style features of incoming samples and routes each sample to the appropriate specialized model, and thereby enables domain-specific adaptation. This multi-model strategy overcomes key limitations of single model adaptation, such as catastrophic forgetting, inter-domain interference, and error accumulation, ensuring robust and stable performance on sustained non-stationary test distributions. Our theoretical analysis reveals key components that bound parameter variance and prevent model collapse, while our plug–in TTA module mitigates catastrophic forgetting of previously encountered domains. Extensive experiments on the classification corruption benchmarks, including ImageNet-C and CIFAR-10/100-C, as well as the Cityscapes→ACDC semantic segmentation task, covering recurring and continuously evolving domain shifts, demonstrate that ReservoirTTA significantly improves adaptation accuracy and maintains stable performance across prolonged, recurring shifts, outperforming state-of-the-art methods.
+
+## Overview
+
+<img src="figs/introduction.png" alt="image" style="width:auto;height:auto;">
+
+<b>Recurring test-time adaptation scenarios. Left: </b> Visual domains can recur over time; ETA, lacking regularization, steadily degrades under these repeated shifts. <b>Right:</b> A zoom-in on the snow corruption across 20 recurrences shows that EATA remains overall stable but still exhibits error spikes on returning to the same corruption across recurrences. <b> ReservoirTTA </b> detects returning domains and reuses specialized models to preserve learned knowledge, delivering improved robustness and faster (re-)adaptation over successive recurrences.
+
+<br>
+<br>
+<br>
+
+<img src="figs/method.png" alt="image" style="width:auto;height:auto;">
+
+<b>Overview of ReservoirTTA.</b> ReservoirTTA operates in four stages: (1) <b>Style Characterization and Domain Identification</b> extracts early convolutional features and assigns incoming test batches to a style cluster via an online clustering mechanism; (2) <b>Model Reservoir Initialization</b> adds a new model for a detected domain, initializing it with parameters that maximize prediction mutual information; (3) <b>Model Reservoir Adaptation</b> selectively adapts the most relevant model using TTA methods; and (4) <b>Model Prediction</b> is then obtained via the ensemble’s parameters.
+
+## Prerequisites
+To use the repository please use the following conda environment
+
+```
+conda update conda
+conda env create -f environment.yml
+conda activate reservoirtta 
+```
+
+## Run
+To execute the code, use the following command:
+
+```
+python test_time.py --cfg <config_filename>
+```
+
+Replace `<config_filename>` with the appropriate configuration file located in the `cfgs` directory.
+
+### Examples
+To run experiments on CIFAR100-C with the CSC setting:
+
+- **ETA**:
+  ```
+  python test_time.py --cfg cfgs/cifar100_c/Standard/eta.yaml
+  ```
+- **EATA**:
+  ```
+  python test_time.py --cfg cfgs/cifar100_c/Standard/eata.yaml
+  ```
+- **EATA+ReservoirTTA**:
+  ```
+  python test_time.py --cfg cfgs/cifar100_c/Standard/eata_reservoir.yaml
+  ```
+
+To run experiments on CIFAR10-C with the CDC setting:
+
+- **ROID**:
+  ```
+  python test_time.py --cfg cfgs/cifar10_c/Standard/roid.yaml SETTING continual_cdc
+  ```
+- **ROID+ReservoirTTA**:
+  ```
+  python test_time.py --cfg cfgs/cifar10_c/Standard/roid_reservoir.yaml SETTING continual_cdc
+  ```
+
+### Parameters to Tune
+The default configurations are defined in `conf.py`. Below is a detailed explanation of the key parameters you can adjust:
+
+#### General Parameters
+- **`DATA_DIR`**: Specifies the path to the data directory where datasets are stored.
+- **`SETTING`**: Defines the Test-Time Adaptation (TTA) protocol. Options include:
+  - `"continual"`: For the Continual Shifting Corruption (CSC) setting.
+  - `"continual_cdc"`: For the Continual Domain Corruption (CDC) setting.
+- **`CORRUPTION.RECUR`**: Indicates the number of recurrences for the corruption.
+
+#### ReservoirTTA-Specific Parameters
+- **`RESERVOIRTTA.MAX_NUM_MODELS`**: The maximum number of models allowed in the reservoir.
+- **`RESERVOIRTTA.SIZE_OF_BUFFER`**: The size of the style reservoir buffer used for storing style features.
+- **`RESERVOIRTTA.QUANTILE_THR`**: The quantile threshold used to set the new domain detector threshold.
+- **`RESERVOIRTTA.ENSEMBLING`**: A boolean flag to enable or disable weight ensembling for predictions.
+- **`RESERVOIRTTA.SAMPLING`**: Specifies the method used to sample features from the style reservoir.
+- **`RESERVOIRTTA.INIT`**: Defines the method for initializing a new model in the reservoir.
+- **`RESERVOIRTTA.STYLE_IDX`**: A list of layers from the frozen VGG model to use as style extractors.
+- **`RESERVOIRTTA.STYLE_FORMAT`**: Specifies the function type used to compute the style features from the frozen VGG model.
+
+These parameters allow you to customize the behavior of the framework to suit your specific use case or dataset.
+
+## Thanks
+Our code is derived from https://github.com/mariodoebler/test-time-adaptation. Please follow this repository to download datasets under `data` for CIFAR10-C, CIFAR100-C, ImageNet-C, and CCC.
+
+## Cite
+```
+@article{vray2025reservoirtta,
+  title={ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains},
+  author={Vray, Guillaume and Tomar, Devavrat and Gao, Xufeng and Thiran, Jean-Philippe and Shelhamer, Evan and Bozorgtabar, Behzad},
+  journal={arXiv preprint arXiv:2505.14511},
+  year={2025}
+}
+```

augmentations/augmix_ops.py

@@ -0,0 +1,146 @@
+# Copyright 2019 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+"""Base augmentations operators."""
+
+import numpy as np
+from PIL import Image, ImageOps, ImageEnhance
+
+
+def int_parameter(level, maxval):
+  """Helper function to scale `val` between 0 and maxval .
+
+  Args:
+    level: Level of the operation that will be between [0, `PARAMETER_MAX`].
+    maxval: Maximum value that the operation can have. This will be scaled to
+      level/PARAMETER_MAX.
+
+  Returns:
+    An int that results from scaling `maxval` according to `level`.
+  """
+  return int(level * maxval / 10)
+
+
+def float_parameter(level, maxval):
+  """Helper function to scale `val` between 0 and maxval.
+
+  Args:
+    level: Level of the operation that will be between [0, `PARAMETER_MAX`].
+    maxval: Maximum value that the operation can have. This will be scaled to
+      level/PARAMETER_MAX.
+
+  Returns:
+    A float that results from scaling `maxval` according to `level`.
+  """
+  return float(level) * maxval / 10.
+
+
+def sample_level(n):
+  return np.random.uniform(low=0.1, high=n)
+
+
+def autocontrast(pil_img, *args):
+  return ImageOps.autocontrast(pil_img)
+
+
+def equalize(pil_img, *args):
+  return ImageOps.equalize(pil_img)
+
+
+def posterize(pil_img, level, _):
+  level = int_parameter(sample_level(level), 4)
+  return ImageOps.posterize(pil_img, 4 - level)
+
+
+def rotate(pil_img, level, _):
+  degrees = int_parameter(sample_level(level), 30)
+  if np.random.uniform() > 0.5:
+    degrees = -degrees
+  return pil_img.rotate(degrees, resample=Image.BILINEAR)
+
+
+def solarize(pil_img, level, _):
+  level = int_parameter(sample_level(level), 256)
+  return ImageOps.solarize(pil_img, 256 - level)
+
+
+def shear_x(pil_img, level, img_size):
+  level = float_parameter(sample_level(level), 0.3)
+  if np.random.uniform() > 0.5:
+    level = -level
+  return pil_img.transform((img_size, img_size),
+                           Image.AFFINE, (1, level, 0, 0, 1, 0),
+                           resample=Image.BILINEAR)
+
+
+def shear_y(pil_img, level, img_size):
+  level = float_parameter(sample_level(level), 0.3)
+  if np.random.uniform() > 0.5:
+    level = -level
+  return pil_img.transform((img_size, img_size),
+                           Image.AFFINE, (1, 0, 0, level, 1, 0),
+                           resample=Image.BILINEAR)
+
+
+def translate_x(pil_img, level, img_size):
+  level = int_parameter(sample_level(level), img_size / 3)
+  if np.random.random() > 0.5:
+    level = -level
+  return pil_img.transform((img_size, img_size),
+                           Image.AFFINE, (1, 0, level, 0, 1, 0),
+                           resample=Image.BILINEAR)
+
+
+def translate_y(pil_img, level, img_size):
+  level = int_parameter(sample_level(level), img_size / 3)
+  if np.random.random() > 0.5:
+    level = -level
+  return pil_img.transform((img_size, img_size),
+                           Image.AFFINE, (1, 0, 0, 0, 1, level),
+                           resample=Image.BILINEAR)
+
+
+# operation that overlaps with ImageNet-C's test set
+def color(pil_img, level, _):
+    level = float_parameter(sample_level(level), 1.8) + 0.1
+    return ImageEnhance.Color(pil_img).enhance(level)
+
+
+# operation that overlaps with ImageNet-C's test set
+def contrast(pil_img, level, _):
+    level = float_parameter(sample_level(level), 1.8) + 0.1
+    return ImageEnhance.Contrast(pil_img).enhance(level)
+
+
+# operation that overlaps with ImageNet-C's test set
+def brightness(pil_img, level, _):
+    level = float_parameter(sample_level(level), 1.8) + 0.1
+    return ImageEnhance.Brightness(pil_img).enhance(level)
+
+
+# operation that overlaps with ImageNet-C's test set
+def sharpness(pil_img, level, _):
+    level = float_parameter(sample_level(level), 1.8) + 0.1
+    return ImageEnhance.Sharpness(pil_img).enhance(level)
+
+
+augmentations = [
+    autocontrast, equalize, posterize, rotate, solarize, shear_x, shear_y,
+    translate_x, translate_y
+]
+
+augmentations_all = [
+    autocontrast, equalize, posterize, rotate, solarize, shear_x, shear_y,
+    translate_x, translate_y, color, contrast, brightness, sharpness
+]

augmentations/transforms_cotta.py

@@ -0,0 +1,166 @@
+# KATANA: Simple Post-Training Robustness Using Test Time Augmentations
+# https://arxiv.org/pdf/2109.08191v1.pdf
+import PIL
+import torch
+import torchvision.transforms.functional as F
+import torchvision.transforms as transforms
+from torchvision.transforms import ColorJitter, Compose, Lambda
+from numpy import random
+
+
+class GaussianNoise(torch.nn.Module):
+    def __init__(self, mean=0., std=1.):
+        super().__init__()
+        self.std = std
+        self.mean = mean
+
+    def forward(self, img):
+        noise = torch.randn(img.size()) * self.std + self.mean
+        noise = noise.to(img.device)
+        return img + noise
+
+    def __repr__(self):
+        return self.__class__.__name__ + '(mean={0}, std={1})'.format(self.mean, self.std)
+
+
+class Clip(torch.nn.Module):
+    def __init__(self, min_val=0., max_val=1.):
+        super().__init__()
+        self.min_val = min_val
+        self.max_val = max_val
+
+    def forward(self, img):
+        return torch.clip(img, self.min_val, self.max_val)
+
+    def __repr__(self):
+        return self.__class__.__name__ + '(min_val={0}, max_val={1})'.format(self.min_val, self.max_val)
+
+
+class ColorJitterPro(ColorJitter):
+    """Randomly change the brightness, contrast, saturation, and gamma correction of an image."""
+
+    def __init__(self, brightness=0, contrast=0, saturation=0, hue=0, gamma=0):
+        super().__init__(brightness, contrast, saturation, hue)
+        self.gamma = self._check_input(gamma, 'gamma')
+
+    @staticmethod
+    @torch.jit.unused
+    def get_params(brightness, contrast, saturation, hue, gamma):
+        """Get a randomized transform to be applied on image.
+
+        Arguments are same as that of __init__.
+
+        Returns:
+            Transform which randomly adjusts brightness, contrast and
+            saturation in a random order.
+        """
+        transforms = []
+
+        if brightness is not None:
+            brightness_factor = random.uniform(brightness[0], brightness[1])
+            transforms.append(Lambda(lambda img: F.adjust_brightness(img, brightness_factor)))
+
+        if contrast is not None:
+            contrast_factor = random.uniform(contrast[0], contrast[1])
+            transforms.append(Lambda(lambda img: F.adjust_contrast(img, contrast_factor)))
+
+        if saturation is not None:
+            saturation_factor = random.uniform(saturation[0], saturation[1])
+            transforms.append(Lambda(lambda img: F.adjust_saturation(img, saturation_factor)))
+
+        if hue is not None:
+            hue_factor = random.uniform(hue[0], hue[1])
+            transforms.append(Lambda(lambda img: F.adjust_hue(img, hue_factor)))
+
+        if gamma is not None:
+            gamma_factor = random.uniform(gamma[0], gamma[1])
+            transforms.append(Lambda(lambda img: F.adjust_gamma(img, gamma_factor)))
+
+        random.shuffle(transforms)
+        transform = Compose(transforms)
+
+        return transform
+
+    def forward(self, img):
+        """
+        Args:
+            img (PIL Image or Tensor): Input image.
+
+        Returns:
+            PIL Image or Tensor: Color jittered image.
+        """
+        fn_idx = torch.randperm(5)
+        for fn_id in fn_idx:
+            if fn_id == 0 and self.brightness is not None:
+                brightness = self.brightness
+                brightness_factor = torch.tensor(1.0).uniform_(brightness[0], brightness[1]).item()
+                img = F.adjust_brightness(img, brightness_factor)
+
+            if fn_id == 1 and self.contrast is not None:
+                contrast = self.contrast
+                contrast_factor = torch.tensor(1.0).uniform_(contrast[0], contrast[1]).item()
+                img = F.adjust_contrast(img, contrast_factor)
+
+            if fn_id == 2 and self.saturation is not None:
+                saturation = self.saturation
+                saturation_factor = torch.tensor(1.0).uniform_(saturation[0], saturation[1]).item()
+                img = F.adjust_saturation(img, saturation_factor)
+
+            if fn_id == 3 and self.hue is not None:
+                hue = self.hue
+                hue_factor = torch.tensor(1.0).uniform_(hue[0], hue[1]).item()
+                img = F.adjust_hue(img, hue_factor)
+
+            if fn_id == 4 and self.gamma is not None:
+                gamma = self.gamma
+                gamma_factor = torch.tensor(1.0).uniform_(gamma[0], gamma[1]).item()
+                img = img.clamp(1e-8, 1.0)  # to fix Nan values in gradients, which happens when applying gamma
+                                            # after contrast
+                img = F.adjust_gamma(img, gamma_factor)
+
+        return img
+
+    def __repr__(self):
+        format_string = self.__class__.__name__ + '('
+        format_string += 'brightness={0}'.format(self.brightness)
+        format_string += ', contrast={0}'.format(self.contrast)
+        format_string += ', saturation={0}'.format(self.saturation)
+        format_string += ', hue={0})'.format(self.hue)
+        format_string += ', gamma={0})'.format(self.gamma)
+        return format_string
+
+
+def get_tta_transforms(img_size, gaussian_std: float=0.005, soft=False, padding_mode='edge', cotta_augs=True):
+    n_pixels = img_size[0] if isinstance(img_size, (list, tuple)) else img_size
+
+    tta_transforms = [
+        Clip(0.0, 1.0),
+        ColorJitterPro(
+            brightness=[0.8, 1.2] if soft else [0.6, 1.4],
+            contrast=[0.85, 1.15] if soft else [0.7, 1.3],
+            saturation=[0.75, 1.25] if soft else [0.5, 1.5],
+            hue=[-0.03, 0.03] if soft else [-0.06, 0.06],
+            gamma=[0.85, 1.15] if soft else [0.7, 1.3]
+        ),
+        transforms.Pad(padding=int(n_pixels / 2), padding_mode=padding_mode),
+        transforms.RandomAffine(
+            degrees=[-8, 8] if soft else [-15, 15],
+            translate=(1/16, 1/16),
+            scale=(0.95, 1.05) if soft else (0.9, 1.1),
+            shear=None,
+            interpolation=PIL.Image.BILINEAR,
+            fill=0
+        )
+    ]
+    if cotta_augs:
+        tta_transforms += [transforms.GaussianBlur(kernel_size=5, sigma=[0.001, 0.25] if soft else [0.001, 0.5]),
+                           transforms.CenterCrop(size=n_pixels),
+                           transforms.RandomHorizontalFlip(p=0.5),
+                           GaussianNoise(0, gaussian_std),
+                           Clip(0.0, 1.0)]
+    else:
+        tta_transforms += [transforms.CenterCrop(size=n_pixels),
+                           transforms.RandomHorizontalFlip(p=0.5),
+                           Clip(0.0, 1.0)]
+
+    return transforms.Compose(tta_transforms)

cfgs/ccc/resnet50/eata.yaml

@@ -0,0 +1,22 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 2000.0
+  D_MARGIN: 0.05
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/ccc/resnet50/eata_reservoir.yaml

@@ -0,0 +1,26 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 2000.0
+  D_MARGIN: 0.05
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+  

cfgs/ccc/resnet50/eta.yaml

@@ -0,0 +1,20 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.05

cfgs/ccc/resnet50/eta_reservoir.yaml

@@ -0,0 +1,26 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.05
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+  

cfgs/ccc/resnet50/rdumb.yaml

@@ -0,0 +1,22 @@
+MODEL:
+  ADAPTATION: eata
+  RESET_AFTER_NUM_UPDATES: 1000
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.05
+

cfgs/ccc/resnet50/roid.yaml

@@ -0,0 +1,20 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: roid
+  ARCH: resnet50
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False

cfgs/ccc/resnet50/roid_reservoir.yaml

@@ -0,0 +1,25 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: roid_reservoirtta
+  ARCH: resnet50
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 4096

cfgs/ccc/resnet50/sar.yaml

@@ -0,0 +1,17 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: sar
+  ARCH: resnet50
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64

cfgs/ccc/resnet50/sar_reservoir.yaml

@@ -0,0 +1,22 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: sar_reservoirtta
+  ARCH: resnet50
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024

cfgs/ccc/resnet50/source.yaml

@@ -0,0 +1,11 @@
+MODEL:
+  ADAPTATION: source
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44

cfgs/ccc/resnet50/tent.yaml

@@ -0,0 +1,18 @@
+MODEL:
+  ADAPTATION: tent
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+

cfgs/ccc/resnet50/tent_reservoir.yaml

@@ -0,0 +1,23 @@
+MODEL:
+  ADAPTATION: tent_reservoirtta
+  ARCH: resnet50
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 0.00025
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+  

cfgs/ccc/vit_b_16/dpcore.yaml

@@ -0,0 +1,27 @@
+MODEL:
+  ADAPTATION: dpcore
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  METHOD: AdamW
+  STEPS: 1
+  MOMENTUM: 0.9
+  LR: 0.01
+  WD: 0.01
+  BETA: 0.9
+DPCORE:
+  TEMP_TAU: 3.0
+  EMA_ALPHA: 0.999
+  THR_RHO: 0.8
+  E_ID: 1
+  E_OOD: 50
+  NUM_PROMPTS: 8
+  SRC_NUM_SAMPLES: 300
+  MAX_PROTOTYPES: 2000

cfgs/ccc/vit_b_16/eata.yaml

@@ -0,0 +1,22 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 2000.0
+  D_MARGIN: 0.05
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/ccc/vit_b_16/eata_reservoir.yaml

@@ -0,0 +1,26 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 2000.0
+  D_MARGIN: 0.05
+  MOMENTUM_SRC: 1.0
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024

cfgs/ccc/vit_b_16/eta.yaml

@@ -0,0 +1,21 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.05
+

cfgs/ccc/vit_b_16/eta_reservoir.yaml

@@ -0,0 +1,24 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.05
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+  NUM_SAMPLES: 2000

cfgs/ccc/vit_b_16/prompt_reservoir.yaml

@@ -0,0 +1,36 @@
+MODEL:
+  ADAPTATION: prompt_reservoirtta
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  METHOD: AdamW
+  STEPS: 1
+  MOMENTUM: 0.9
+  LR: 0.1
+  WD: 0.01
+  BETA: 0.9
+DPCORE:
+  TEMP_TAU: 3.0
+  EMA_ALPHA: 0.999
+  THR_RHO: 0.8
+  E_ID: 1
+  E_OOD: 50
+  NUM_PROMPTS: 8
+  SRC_NUM_SAMPLES: 300
+  MAX_PROTOTYPES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 64
+  SIZE_OF_BUFFER: 1024
+  ENSEMBLING: False
+  SOURCE_BUFFER: True
+  CENTROID_REG: True 
+  NUM_SAMPLES: 2000
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/ccc/vit_b_16/rdumb.yaml

@@ -0,0 +1,21 @@
+MODEL:
+  ADAPTATION: eata
+  RESET_AFTER_NUM_UPDATES: 1000
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.05

cfgs/ccc/vit_b_16/roid.yaml

@@ -0,0 +1,20 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: roid
+  ARCH: Standard_VITB
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False

cfgs/ccc/vit_b_16/roid_reservoir.yaml

@@ -0,0 +1,25 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: roid_reservoirtta
+  ARCH: Standard_VITB
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 4096
+SOURCE:
+  NUM_SAMPLES: 2000
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False

cfgs/ccc/vit_b_16/sar.yaml

@@ -0,0 +1,17 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: sar
+  ARCH: Standard_VITB
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64

cfgs/ccc/vit_b_16/sar_reservoir.yaml

@@ -0,0 +1,22 @@
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+MODEL:
+  ADAPTATION: sar_reservoirtta
+  ARCH: Standard_VITB
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+TEST:
+  BATCH_SIZE: 64
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/ccc/vit_b_16/source_vit.yaml

@@ -0,0 +1,11 @@
+MODEL:
+  ADAPTATION: source
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44

cfgs/ccc/vit_b_16/tent.yaml

@@ -0,0 +1,17 @@
+MODEL:
+  ADAPTATION: tent
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0

cfgs/ccc/vit_b_16/tent_reservoir.yaml

@@ -0,0 +1,22 @@
+MODEL:
+  ADAPTATION: tent_reservoirtta
+  ARCH: Standard_VITB
+TEST:
+  BATCH_SIZE: 64
+CORRUPTION:
+  DATASET: ccc
+  SEVERITY:
+    - 5
+  TYPE:
+    - baseline_20_transition+speed_1000_seed_44
+OPTIM:
+  BETA: 0.9
+  LR: 1e-4
+  METHOD: SGD
+  STEPS: 1
+  WD: 0.0
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar100_c/Standard/eata.yaml

@@ -0,0 +1,36 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 2000.
+  D_MARGIN: 0.1
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar100_c/Standard/eata_reservoir.yaml

@@ -0,0 +1,39 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: Hendrycks2020AugMix_ResNeXt
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 2000.
+  D_MARGIN: 0.1
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024

cfgs/cifar100_c/Standard/eta.yaml

@@ -0,0 +1,34 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.1

cfgs/cifar100_c/Standard/eta_reservoir.yaml

@@ -0,0 +1,39 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: Hendrycks2020AugMix_ResNeXt
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.1
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar100_c/Standard/rdumb.yaml

@@ -0,0 +1,36 @@
+MODEL:
+  ADAPTATION: eata
+  RESET_AFTER_NUM_UPDATES: 1000
+  ARCH: Hendrycks2020AugMix_ResNeXt
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  RECUR: 20
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.1

cfgs/cifar100_c/Standard/roid.yaml

@@ -0,0 +1,34 @@
+MODEL:
+  ADAPTATION: roid
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False

cfgs/cifar100_c/Standard/roid_reservoir.yaml

@@ -0,0 +1,39 @@
+MODEL:
+  ADAPTATION: roid_reservoirtta
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar100_c/Standard/sar.yaml

@@ -0,0 +1,35 @@
+MODEL:
+  ADAPTATION: sar
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+  
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.1

cfgs/cifar100_c/Standard/sar_reservoir.yaml

@@ -0,0 +1,36 @@
+MODEL:
+  ADAPTATION: sar_reservoirtta
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar100_c/Standard/source.yaml

@@ -0,0 +1,27 @@
+MODEL:
+  ADAPTATION: source
+  ARCH: Hendrycks2020AugMix_ResNeXt
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  RECUR: 1
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - motion_blur
+    - snow
+    - fog
+    - shot_noise
+    - defocus_blur
+    - contrast
+    - zoom_blur
+    - brightness
+    - frost
+    - elastic_transform
+    - glass_blur
+    - gaussian_noise
+    - pixelate
+    - jpeg_compression
+    - impulse_noise
+

cfgs/cifar100_c/Standard/tent.yaml

@@ -0,0 +1,31 @@
+MODEL:
+  ADAPTATION: tent
+  ARCH: Hendrycks2020AugMix_ResNeXt 
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.

cfgs/cifar100_c/Standard/tent_reservoir.yaml

@@ -0,0 +1,36 @@
+MODEL:
+  ADAPTATION: tent_reservoirtta
+  ARCH: Hendrycks2020AugMix_ResNeXt
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar100_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar10_c/Standard/eata.yaml

@@ -0,0 +1,36 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 2000.
+  D_MARGIN: 0.4
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar10_c/Standard/eata_reservoir.yaml

@@ -0,0 +1,39 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 2000.
+  D_MARGIN: 0.4
+SOURCE:
+  NUM_SAMPLES: 2000
+RESERVOITTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024

cfgs/cifar10_c/Standard/eta.yaml

@@ -0,0 +1,34 @@
+MODEL:
+  ADAPTATION: eata
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.4

cfgs/cifar10_c/Standard/eta_reservoir.yaml

@@ -0,0 +1,39 @@
+MODEL:
+  ADAPTATION: eata_reservoirtta
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.4
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar10_c/Standard/rdumb.yaml

@@ -0,0 +1,35 @@
+MODEL:
+  ADAPTATION: eata
+  RESET_AFTER_NUM_UPDATES: 1000
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+EATA:
+  FISHER_ALPHA: 0.0
+  D_MARGIN: 0.4

cfgs/cifar10_c/Standard/roid.yaml

@@ -0,0 +1,31 @@
+MODEL:
+  ADAPTATION: roid
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.

cfgs/cifar10_c/Standard/roid_reservoir.yaml

@@ -0,0 +1,39 @@
+MODEL:
+  ADAPTATION: roid_reservoirtta
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.
+RESERVOIRTTA:
+  MAX_NUM_MODELS: 16
+  SIZE_OF_BUFFER: 1024
+ROID:
+  USE_CONSISTENCY: False
+  USE_PRIOR_CORRECTION: False
+SOURCE:
+  NUM_SAMPLES: 2000

cfgs/cifar10_c/Standard/sar.yaml

@@ -0,0 +1,31 @@
+MODEL:
+  ADAPTATION: sar
+  ARCH: Standard
+TEST:
+  BATCH_SIZE: 200
+CORRUPTION:
+  DATASET: cifar10_c
+  SEVERITY:
+    - 5
+  TYPE:
+    - gaussian_noise
+    - shot_noise
+    - impulse_noise
+    - defocus_blur
+    - glass_blur
+    - motion_blur
+    - zoom_blur
+    - snow
+    - frost
+    - fog
+    - brightness
+    - contrast
+    - elastic_transform
+    - pixelate
+    - jpeg_compression
+OPTIM:
+  METHOD: Adam
+  STEPS: 1
+  BETA: 0.9
+  LR: 1e-3
+  WD: 0.