diff --git a/JayRnn.py b/JayRnn.py
index 002b0ac..e605e82 100644
--- a/JayRnn.py
+++ b/JayRnn.py
@@ -24,10 +24,11 @@ from rbm.train import train
 from rbm.status import Status
 
 # ── Hyper-parameters ──────────────────────────────────────────────────────────
-TEXT    = "CHAPTER 1 LOOMINGS CALL ME ISHMAEL SOME YEARS AGO NEVER MIND HOW LONG PRECISELY HAVING LITTLE OR NO" # the character sequence to learn (matches diagram example)
-WIN    = 2         # sliding-window width  (= N in the diagram)
+#TEXT    = "CHAPTER 1 LOOMINGS CALL ME ISHMAEL SOME YEARS AGO NEVER MIND HOW LONG PRECISELY HAVING LITTLE OR NO" # the character sequence to learn (matches diagram example)
+TEXT = "0123456789 "
+WIN    = 3         # sliding-window width  (= N in the diagram)
 UNITS  = 5
-H_SIZE = 128         # hidden units per RBM cell
+H_SIZE = 256         # hidden units per RBM cell
 NUM_EPOCHS = 1000
 
 TRAIN_PARAMS = TrainingParams(
@@ -114,7 +115,7 @@ class RnnModel(Model):
 				print(f"train[{index}:{i:03}]: {unit.name}:{vc2char(_vc[i,:])}")
 			train(unit, _vc, status)
 			# For the next unit: Update context portion of vc
-			_c = unit.forward(_vc)
+#			_c = unit.forward(_vc)
 
 
 	def forward_step(self, _ch: str, _state: str):
@@ -122,13 +123,13 @@ class RnnModel(Model):
 		_c = np.zeros([len(_ch), H_SIZE])
 		_ch_predict = '?'
 		for index, unit in enumerate(self.units):
-			_s = _state[index] + SPACE*(WIN-1)
+			_s = _state[index:index+WIN-1] + SPACE
 			print(f"forward_step in : {unit.name}:{_s}")
 			_v = str2vec(_s).reshape(1, WIN*vocab_size())
 			_vc = concat(_v, _c, axis=1)
-			_c = unit.forward(_vc)
-			_vc = unit.reconstruct(_c)
-			_r, _ = split(_vc, H_SIZE, axis=1)
+#			_c = unit.forward(_vc)
+			_vc_r = unit.reconstruct(_c)
+			_r, _ = split(_vc_r, H_SIZE, axis=1)
 			_r_cl = clamp(_r.reshape([WIN, vocab_size()]), axis=1)
 
 			_ch_predict = v2char(_r_cl)
@@ -143,7 +144,7 @@ if __name__ == "__main__":
 	if 1:
 		# Test of helper function
 		# ToDo: move to tests
-		vec = idx2vec(ch2idx("JENS"))
+		vec = idx2vec(ch2idx("1234"))
 		vec = clamp(vec, axis=1)
 		indices = vec2idx(vec, axis=1)
 		ch_str = idx2ch(indices)
@@ -162,20 +163,15 @@ if __name__ == "__main__":
 
 	# vc contains vis + context
 	# context will be updated after training
-	model.train(win_text, status=Status())
-	model.save()
+#	model.train(win_text, status=Status())
+#	model.save()
 
 	# test the model
-	seed_str = 'CHAP'
-	state = SPACE*UNITS
+	state = '01234'
 
-	result = seed_str
-	ch = SPACE
+	result = state
+	ch = '5'
 	# Prime
-	for s in seed_str:
-		state, ch = model.forward_step(s, state)
-		print(f'Prime: {ch}')
-
 	for i in range(len(TEXT)):
 		state, ch = model.forward_step(ch, state)
 		print(f'Predict: {ch}')
diff --git a/src/stack/rnn_helper.py b/src/stack/rnn_helper.py
index 692c740..c2f7d0f 100644
--- a/src/stack/rnn_helper.py
+++ b/src/stack/rnn_helper.py
@@ -1,7 +1,8 @@
 from rbm.matrix import Mat, np
 
 SPACE=' '
-VOCAB = SPACE + 'ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789'
+#VOCAB = SPACE + 'ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789'
+VOCAB = SPACE + '0123456789'
 _CH2IDX = {ch: i for i, ch in enumerate(VOCAB)}
 
 def clamp(vec: Mat, axis=0):