jaccard, temp, etc.

speak_broadcast.py

@@ -7,8 +7,9 @@ import examine.metric
 logging.basicConfig(level=logging.INFO)
 
 UTTERANCE_LEN = 16				#<--------------- these should be in config
-NUM_METRIC_GEN = 50
+NUM_METRIC_GEN = 75
 NUM_SAMPLE_VOICES = 3
+RANDOM_SEED_TIMER_MIN = 2
 broadcast = None
 metric = None
 exit = False
@@ -52,7 +53,11 @@ def broadcast_utterance(v, utterance):
 		temp = format(v.temp, '.2f')
 		text = f"{v.name.upper()}: {temp}\n"
 	else:
-		text = f"{v.name.upper()}:\n"
+		text = f""
+
+	# Send all text to server to calculate bounds in advance
+	broadcast.utterance(utterance, v.calculate)
+	
 
 	broadcast.utterance(text, v.channel)
 	terminal_utterance(text)
@@ -113,7 +118,7 @@ def main() -> int:
 	p = argparse.ArgumentParser()
 	p.add_argument("-c", "--config", type=str, default="voice.config.json", help="configuratin file")
 	p.add_argument("-i", "--iterations", type=int, default=10, help="number of iterations")
-	p.add_argument("-t", "--terminal", action=argparse.BooleanOptionalAction, help="print to terminal")
+	p.add_argument("-t", "--terminal", action='store_true', help="print to terminal")
 	args = p.parse_args()	
 
 	logging.info(f"INIT::loading config file - {args.config}")
@@ -135,6 +140,8 @@ def main() -> int:
 		model = v['model']
 		voice = utterance.voice.Voice(name=v["name"].upper(), model=model['model_dir'], tokenizer=model['tokeniser_file'], temp=float(model["temperature"]), lenght=UTTERANCE_LEN)
 		voice.set_channel(v['osc_channel']['root'], v['osc_channel']['utterance'])
+		voice.set_calculate(v['osc_channel']['root'], v['osc_channel']['calculate'])
+		voice.set_temperature(v['osc_channel']['root'], v['osc_channel']['temperature'])
 		voices.append(voice)
 
 
@@ -152,11 +159,12 @@ def main() -> int:
 		if type(temp) == str and temp == "DEBUG":
 			debug = name
 			logging.info(f"DEBUG MODE: {debug}")
-		else:
+		else: # temperature
 			for v in voices:
 				if v.name == name:
 					print(f'{name} - {temp}')
 					v.temp = temp
+					broadcast.temperature(temp, v.temperature)
 
 	receiver = utterance.osc.OscReceiver(name="osc_receiver", host=conf['host_machinespeak'], port=conf['port_machinespeak'], callback_fn=receiver_cb)
 
@@ -176,10 +184,10 @@ def main() -> int:
 		logging.info("INIT::SEEDING RANDOM")
 		random.seed(time.time())
 		if not exit:
-			t_random_seed = threading.Timer(random.randint(60, 60 * 4), random_seed)
+			t_random_seed = threading.Timer(random.randint(60, 60 * RANDOM_SEED_TIMER_MIN), random_seed)
 			t_random_seed.start()
 
-	t_random_seed = threading.Timer(random.randint(60, 60 * 4), random_seed)
+	t_random_seed = threading.Timer(random.randint(60, 60 * RANDOM_SEED_TIMER_MIN), random_seed)
 	t_random_seed.start()
 
 	#--------------------#
@@ -206,7 +214,19 @@ def main() -> int:
 
 		# ok here we need to randomise maybe...?!
 		# ([d, t, v])
+
+
 		choice = results[0]
+		# makse sure we don't say the same thing over and over again
+		for r in results:
+			v = r[2]
+			u = r[1]
+			if v.select(u):
+				choice = r
+				break
+			else:
+				logging.info(f"LOOP::reduncancy {v.name}")
+
 		v = choice[2]
 		uv = choice[1]
 		uv_vec = metric.vector(uv)

utterance/osc.py

@@ -37,6 +37,11 @@ class OscBroadcaster:
 		osc_send(msg, self.name)
 		update()
 
+	def temperature(self, temp: float, channel: str):
+		msg = oscbuildparse.OSCMessage(channel, None, [temp])
+		osc_send(msg, self.name)
+		update()
+
 class OscReceiver:
 
 	def __init__(self, name: str, host: str, port: str, callback_fn=None):

utterance/utils.py

@@ -16,7 +16,7 @@ def clean(text: str) -> str:
 
 def format(text: str) -> str:
 
-	return text.replace('\r\n', '\n').replace('\n\n', '\n')
+	return text.replace('\r\n', '\n').replace('\n\n', '\n').replace('‘', "'").replace("’", "'")
 
 def fragments(utterance: str):
 	frags = []

utterance/voice.py

@@ -1,6 +1,9 @@
 from aitextgen import aitextgen
 import utterance.utils
-import regex, string, time
+import gensim, regex, string, time
+
+UTTERANCE_MEMORY_LEN = 15
+UTTERANCE_MEMORY_MIN_DIST = 0.2
 
 class Voice:
 
@@ -9,6 +12,7 @@ class Voice:
 		self.temp = temp
 		self.lenght = lenght
 		self.v = aitextgen(model_folder=model, tokenizer_file=tokenizer)
+		self.utterances = []
 
 	def utter_n(self, n: int, temp: float = None, lenght: int = None):
 		t = self.temp if temp == None else temp
@@ -20,12 +24,49 @@ class Voice:
 		l = self.lenght if lenght == None else lenght		
 		return self.utter_n(n=1, temp=temp, lenght=lenght)[0]
 
+	def select(self, utterance: str) -> bool:
+
+		# fuction making sure to not say the same thing
+
+		print("select;")
+
+		toks = set(gensim.utils.tokenize(utterance))
+		
+		i = 0
+		for u in self.utterances:
+			d = gensim.matutils.jaccard_distance(toks, u)
+			print(f"{i} - d: {d}")
+			if d < UTTERANCE_MEMORY_MIN_DIST:
+				return False
+
+		self.utterances.append(toks)
+
+		print(len(self.utterances))
+
+		if len(self.utterances) > UTTERANCE_MEMORY_LEN:
+			self.utterances.pop(0)
+
+		return True
+
+
 	def set_channel(self, root: str, endpoint: str):
 		self.channel = root + endpoint
 
 	def channel(self):
 		return self.channel
 
+	def set_calculate(self, root: str, endpoint: str):
+		self.calculate = root + endpoint
+
+	def calculate(self):
+		return self.calculate
+
+	def set_temperature(self, root: str, endpoint: str):
+		self.temperature = root + endpoint
+
+	def calculate(self):
+		return self.temperature
+
 	def fragments(self, utt: str):
 		self.utterance = utt
 		self.utterance_fragments = utterance.utils.fragments(utt)