rework forward_step to use char-by-char state priming with SPACE token

Replace space ' ' with explicit SPACE='_' vocab token to disambiguate
padding from real spaces, and rewrite forward_step to take a single
character plus rolling state string instead of a reconstructed vc matrix,
producing predictions one character at a time per unit.

JayRnn.py

@@ -19,12 +19,12 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), 'src'))
 from rbm.entity import EntityParams, TrainingParams, Entity
 from rbm.matrix import np, Mat
 from model.model import Model
-from stack.rnn_helper import vocab_size, shift_left, clamp, concat, split, vec2idx, idx2ch, ch2idx, idx2vec
+from stack.rnn_helper import vocab_size, shift_left, clamp, concat, split, vec2idx, idx2ch, ch2idx, idx2vec, SPACE
 from rbm.train import train
 from rbm.status import Status
 
 # ── Hyper-parameters ──────────────────────────────────────────────────────────
-TEXT    = "Call me Ishmael. Some years ago" # the character sequence to learn (matches diagram example)
+TEXT    = "123456789_Call_me_Ishmael._Some_years_ago" # the character sequence to learn (matches diagram example)
 WIN    = 2         # sliding-window width  (= N in the diagram)
 STRIDE = 1     	   # sliding-window step size
 UNITS  = 5
@@ -55,7 +55,7 @@ def str2vec(ch_str: str) -> Mat:
 
 def to_window(text: str) -> list[str]:
 	result = []
-	_win = [' ']*UNITS
+	_win = [SPACE]*UNITS
 	for _i in range(WIN - 1):
 		for ch in text:
 			_win = _win[1:WIN] + [ch]
@@ -88,7 +88,8 @@ def batch_delay(_batch: Mat, delay=0):
 	return result
 
 def vc2char(_vc: Mat):
-	_v, _ = split(_vc.reshape(1, WIN*vocab_size()+H_SIZE), H_SIZE, axis=1)
+	__vc = _vc.reshape([1, WIN*vocab_size()+H_SIZE])
+	_v, _ = split(__vc, H_SIZE, axis=1)
 	return v2char(_v)
 
 def v2char(_v: Mat):
@@ -112,25 +113,29 @@ class RnnModel(Model):
 			_batch = to_batch(_win_text, index)
 			_vc = concat(_batch, _c, axis=1)
 			for i in range(_vc.shape[0]):
-				print(f"train: {unit.name}:{vc2char(_vc[i,:])}")
+				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)
 
 
-	def forward_step(self, _vc: Mat):
+	def forward_step(self, _ch: str, _state: str):
-		text = ''
+		_state = _state[len(_ch):] + _ch
-		for unit in self.units:
+		_c = np.zeros([len(_ch), H_SIZE])
-			print(f"forward_step in : {unit.name}:{vc2char(_vc)}")
+		_ch_predict = '?'
+		for index, unit in enumerate(self.units):
+			_s = _state[index] + SPACE*(WIN-1)
+			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)
-			print(f"forward_step out: {unit.name}:{vc2char(_vc)}")
+			_r, _ = split(_vc, H_SIZE, axis=1)
-			text += vc2char(_vc)[WIN-1]
+			_r_cl = clamp(_r.reshape([WIN, vocab_size()]), axis=1)
-			_v, _ = split(_vc, H_SIZE, axis=1)
+
-			_v_cl = clamp(_v.reshape([WIN, vocab_size()]), axis=1)
+			_ch_predict = v2char(_r_cl)
-			_v = shift_left(_v_cl.reshape([1, WIN*vocab_size()]), vocab_size())
+			print(f"forward_step out: {unit.name}:{_ch_predict}")
-			_vc = concat(_v, _c, axis=1)
+		return _state, _ch_predict[-1]
-		return _vc, text
 
 	def forward(self, x: Mat):
 		pass
@@ -163,14 +168,14 @@ if __name__ == "__main__":
 	model.save()
 
 	# test the model
-	seed_str = ' '
+	seed_str = '1234'
-	seed_padded = seed_str + ' '*(WIN-len(seed_str))
+	state = SPACE*UNITS
-	v_test = str2vec(seed_padded).reshape([1, WIN*vocab_size()])
+
-	c_test = np.zeros([1, H_SIZE])
+	# Prime
-	vc_test = concat(v_test, c_test, axis=1)
+	for s in seed_str:
-	text = ''
+		state, ch = model.forward_step(s, state)
-	for i in range(len(TEXT)):
+		print(f'Predict: {ch}')
-		vc_test, text_frag = model.forward_step(vc_test)
+
-		text += text_frag
+	state, ch = model.forward_step('_', state)
+	print(f'Predict: {ch}')
 
-	print(text)

src/stack/rnn_helper.py

@@ -1,6 +1,7 @@
 from rbm.matrix import Mat, np
 
-VOCAB = ' -.!?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789'
+SPACE='_'
+VOCAB = SPACE + '-.!?ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789'
 _CH2IDX = {ch: i for i, ch in enumerate(VOCAB)}
 
 def clamp(vec: Mat, axis=0):