AdityaKane2001

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
/tmp/ipykernel_17/2958370653.py in <module>
      1 for DATASET in POSSIBLE_DATASETS:
      2     for MODEL_ALIAS in POSSIBLE_MODEL_ALIASES:
----> 3         train_eval_test(MODEL_ALIAS, DATASET)

/tmp/ipykernel_17/1094019785.py in train_eval_test(MODEL_ALIAS, DATASET)
     35         model.train()
     36         for batch in train_dl:

AdityaKane2001

from scipy.stats import norm as dist_model
import numpy as np
import torch

cfg = None # training config
device = None # training device object
seen_classes = list(range(5)) # list of seen classes
OOD_CLASS_NUMBER = -1

def fit(prob_pos_X):

AdityaKane2001

    def _get_random_dims(self, area, height, width, max_iter=10):
        """
        Working logic:
        1. Initialize values, start for loop and generate required random values.
        2. If h_crop and w_crop (i.e. the generated values) are lesser than image dimensions,
            then generate random x and y values.
        3. Cache these x and y values if they are useful (BOTH MUST BE USEFUL)
        4. If cached values exist, return them, else return bad values (atleast one of x and y will contain -1).
        Run cases:
        1. Cached values can be filled multiple times. Any time, they will be useful. 

AdityaKane2001

    def _pca_jitter(self, image, target):
        """
        Applies PCA jitter to images.
        Args:
            image: Batch of images to perform random rotation on.
            target: Target tensor.
        Returns:
            Augmented example with batch of images and targets with same dimensions.
        """

AdityaKane2001

    def _mixup(self, image, label, alpha=0.2) -> Tuple:
        """
        Function to apply mixup augmentation. To be applied after
        one hot encoding and before batching.
        Args:
            entry1: Entry from first dataset. Should be one hot encoded and batched.
            entry2: Entry from second dataset. Must be one hot encoded and batched.
        Returns:
            Tuple with same structure as the entries.
        """

AdityaKane2001

    def random_flip(self, image: tf.Tensor, target: tf.Tensor) -> tuple:
        """
        Returns randomly flipped batch of images. Only horizontal flip
        is available
        Args:
            image: Batch of images to perform random rotation on.
            target: Target tensor.
        Returns:
            Augmented example with batch of images and targets with same dimensions.
        """

AdityaKane2001

    def random_flip(self, image: tf.Tensor, target: tf.Tensor) -> tuple:
        """
        Returns randomly flipped batch of images. Only horizontal flip
        is available
        Args:
            image: Batch of images to perform random rotation on.
            target: Target tensor.
        Returns:
            Augmented example with batch of images and targets with same dimensions.
        """

AdityaKane2001

bazel test keras/layers/convolutional_test

AdityaKane2001

# Make sure you run the following command from the root of your Keras fork
# Note: The version 2.8.0 can change with changes in Keras versions.

bazel build //keras/tools/pip_package:build_pip_package

~/keras_installation/keras/bazel-bin/keras/tools/pip_package/build_pip_package ~/keras_installation/keras_pkg

pip3 install - force-reinstall - user ~/keras_installation/keras_pkg/keras-2.8.0-py2.py3-none-any.whl

AdityaKane2001

.
└── keras/
    ├── api/
    │   ├── BUILD
    │   └── api_init_files.bzl
    ├── ...
    └── <submodule name>/
        └── BUILD