updated setup.sh

debugging.py

@@ -0,0 +1,4 @@
+import tensorflow as tf
+
+print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU')))
+print("Is built with CUDA: ", tf.test.is_built_with_cuda())

setup.sh

@@ -1,5 +1,8 @@
 #!/bin/bash
 
+git config --global user.email "nathan@hintonclan.org"
+git config --global user.name "bionickatana"
+
 # Make virtual env
 
 python3 -m venv venv

train.py

@@ -25,8 +25,8 @@ import numpy as np
 # ==========================================
 # 1. ARCHITECTURE CONFIGURATION
 # ==========================================
-EMBEDDING_DIM = 512
-RNN_UNITS = 512
+EMBEDDING_DIM = 64
+RNN_UNITS = 256
 
 # Pipeline & Training Hyperparameters
 TEXT_FILE = "combined_training_data.txt"
@@ -113,7 +113,7 @@ if os.path.exists(TEXT_FILE):
 # ==========================================
 
 class CharacterTextModel(tf.keras.Model):
-    def __init__(self, vocab_size, embedding_dim, rnn_units, num_layers=2, dropout_rate=0.2):
+    def __init__(self, vocab_size, embedding_dim, rnn_units, num_layers=2, dropout_rate=0.4):
         super().__init__()
         self.embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim)
 
@@ -201,6 +201,13 @@ class GenerationCallback(tf.keras.callbacks.Callback):
 checkpoint_prefix = os.path.join(CHECKPOINT_DIR, "ckpt_{epoch}.weights.h5")
 checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_prefix, save_weights_only=True)
 
+# Stop training when validation loss stops improving for 3 epochs
+early_stopping = tf.keras.callbacks.EarlyStopping(
+    monitor='val_loss',
+    patience=3,
+    restore_best_weights=True # Automatically rolls back to epoch 15 weights!
+)
+
 # ==========================================
 # 6. START RUN
 # ==========================================