scale up model, expand vocab, collect predicted text in forward pass

- Increase H_SIZE to 128, lr to 0.2, disable Rao-Blackwell
- Expand VOCAB to include lowercase and hyphen/period
- Fix batch_delay to always use delay=1 per unit
- forward_step now returns (vc, text) with predicted char per unit step
- Collect and print generated text in main inference loop

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

JayRnn.py

@@ -24,19 +24,19 @@ from rbm.train import train
 from rbm.status import Status
 
 # ── Hyper-parameters ──────────────────────────────────────────────────────────
-TEXT    = "HALLO MAUSI! SUPER HASE!" # the character sequence to learn (matches diagram example)
+TEXT    = "Call me Ishmael. Some years ago" # the character sequence to learn (matches diagram example)
 WIN    = 3         # sliding-window width  (= N in the diagram)
 STRIDE = 1     	   # sliding-window step size
 UNITS  = 3
-H_SIZE = 64         # hidden units per RBM cell
+H_SIZE = 128         # hidden units per RBM cell
 NUM_EPOCHS = 1000
 NUM_ITERATIONS = 1
 
 TRAIN_PARAMS = TrainingParams(
-	learning_rate    = 0.1,
+	learning_rate    = 0.2,
 	momentum         = 0.5,
 	num_epochs       = NUM_EPOCHS,
-	do_rao_blackwell = True,
+	do_rao_blackwell = False,
 	num_gibbs_samples= 1
 )
 #WIN=3
@@ -92,27 +92,29 @@ class RnnModel(Model):
 
 	def train(self, vc: Mat, status: Status = None):
 		_v, _c = split(vc, H_SIZE, axis=1)
-		for delay, unit in enumerate(self.units):
-			_vd = batch_delay(_v, delay)
+		for unit in self.units:
+			_vd = batch_delay(_v, 1)
 			_vc = concat(_vd, _c, axis=1)
 			for i in range(vc.shape[0]):
 				print(f"train: {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, _vc: Mat):
+		text = ''
 		for unit in self.units:
 #			print(f"forward_step in : {unit.name}:{vc2char(_vc)}")
 			_c = unit.forward(_vc)
 			_vc = unit.reconstruct(_c)
-			print(f"forward_step out: {unit.name}:{vc2char(_vc)}")
+#			print(f"forward_step out: {unit.name}:{vc2char(_vc)}")
+			text += vc2char(_vc)[WIN-1]
 			_v, _ = split(_vc, H_SIZE, axis=1)
 			_v_cl = clamp(_v.reshape([WIN, vocab_size()]), axis=1)
 			_v = shift_left(_v_cl.reshape([1, WIN*vocab_size()]), vocab_size())
 			_vc = concat(_v, _c, axis=1)
-		return _vc
+		return _vc, text
 
 	def forward(self, x: Mat):
 		pass
@@ -147,10 +149,14 @@ if __name__ == "__main__":
 	model.save()
 
 	# test the model
-	seed_str = 'HA'
+	seed_str = 'Cal'
 	seed_padded = seed_str + '^'*(WIN-len(seed_str))
 	v_test = str2vec(seed_padded).reshape([1, WIN*vocab_size()])
 	c_test = np.zeros([1, H_SIZE])
 	vc_test = concat(v_test, c_test, axis=1)
+	text = ''
 	for i in range(len(TEXT)):
-		vc_test = model.forward_step(vc_test)
+		vc_test, text_frag = model.forward_step(vc_test)
+		text += text_frag
+
+	print(text)

src/stack/rnn_helper.py

@@ -1,6 +1,6 @@
 from rbm.matrix import Mat, np
 
-VOCAB = '^ .!?ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789'
+VOCAB = '^ -.!?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789'
 _CH2IDX = {ch: i for i, ch in enumerate(VOCAB)}
 
 def clamp(vec: Mat, axis=0):