- changed sample text

- changed model params

JayRnn.py

@@ -24,17 +24,20 @@ from rbm.train import train
 from rbm.status import Status
 
 # ── Hyper-parameters ──────────────────────────────────────────────────────────
-TEXT    = "  HALLO SUPER JENS. SUPER!    " # the character sequence to learn (matches diagram example)
+TEXT    = "  HALLO SUPER JENS. SUPPE!    " # the character sequence to learn (matches diagram example)
 WIN    = 3         # sliding-window width  (= N in the diagram)
 STRIDE = 1     	   # sliding-window step size
 UNITS  = 1
-H_SIZE = 120         # hidden units per RBM cell
+H_SIZE = 128         # hidden units per RBM cell
+NUM_EPOCHS = 100
+NUM_ITERATIONS = 100
 
 TRAIN_PARAMS = TrainingParams(
-	learning_rate    = 0.05,
+	learning_rate    = 0.25,
-	momentum         = 0.9,
+	momentum         = 0.5,
-	num_epochs       = 1000,
+	num_epochs       = NUM_EPOCHS,
 	do_rao_blackwell = True,
+	num_gibbs_samples= 1
 )
 #WIN=3
 #STRIDE=1
@@ -78,7 +81,8 @@ class RnnModel(Model):
 		for unit in self.units:
 			train(unit, vc, status)
 			# Update context portion of vc
-			vc[1:, WIN*vocab_size():] = unit.forward(vc)[0:-1,:]
+			_c = unit.forward(vc)
+			vc[1:, WIN*vocab_size():] = _c[0:-1,:]
 
 	def forward_step(self, _vc: Mat):
 		for unit in self.units:
@@ -115,8 +119,9 @@ if __name__ == "__main__":
 	# context will be updated after training
 	c_train = np.zeros([len(batch), H_SIZE])
 	vc_train = concat(batch, c_train, axis=1)
-	model.train(vc_train, status=Status())
+	for k in range(NUM_ITERATIONS):
-	model.save()
+		model.train(vc_train, status=Status())
+		model.save()
 
 	# test the model
 	seed = str2vec('HA^')