Pinned

1. poke-env poke-env Public

   A python interface for training Reinforcement Learning bots to battle on pokemon showdown

   Python 276 93

2. Speech-to-maths Speech-to-maths Public

   Speech-recognition for Latex generation

   CSS 7 2

3. Pokemon-Showdown Pokemon-Showdown Public

   Forked from smogon/pokemon-showdown

   Pokemon-Showdown fork optimized for RL training performance.

   TypeScript 11 7

4. This function takes a tensorflow dat... This function takes a tensorflow dataset and returns a corresponding dataset implementing cutmix

   1

   def tf_ds_cutmix(ds, shuffling=1024):

   2

       ds_shuffled = ds.shuffle(shuffling)

   3

   4

       def cutmix(p1, p2):

   5

           img_1, label_1 = p1

5. This snippet was used to recover a p... This snippet was used to recover a proper keras model from a saved model which contained a submodel (ie., one of its layers was actually another model), in order to apply model optimization a posteriori (quantization, pruning). It can be extended to handle more type of layers.

   1

   from tensorflow.keras.models import Sequential

   2

   import tensorflow.keras.layers as keras_layers

   3

   4

   5

   clone = Sequential()

6. A custom tensorflow / keras loss imp... A custom tensorflow / keras loss implementing OHEM (https://arxiv.org/abs/1604.03540) with cross-entropy.

   1

   import tensorflow as tf

   2

   from tensorflow.keras.losses import categorical_crossentropy

   3

   4

   @tf.function

   5

   def ohem_crossentropy_loss(y_true, y_pred):