too brainlet to diagnose why low temp / greedy sampling is randomly unstable some of the time

vall_e/models/ar_nar.py

@@ -266,15 +266,17 @@ class AR_NAR(Base):
 		] * batch_size if sampling_mirostat_tau > 0.0 else None
 
 		scores = [ 1.0 ] * sampling_beam_width
-		entropies = []
+		metrics = []
 
 		# ick
-		low_temperature = False # sampling_repetition_penalty == 1.0 and sampling_temperature == 0.0 #
+		"""
+		low_temperature = False # sampling_temperature < 0.6 # sampling_repetition_penalty == 1.0 and sampling_temperature == 0.0 #
 		low_temperature_range = cfg.dataset.frames_per_second * 5
 		
 		original_sampling_temperature = sampling_temperature 
 		original_sampling_repetition_penalty = sampling_repetition_penalty 
 		original_sampling_repetition_penalty_decay = sampling_repetition_penalty_decay
+		"""
 
 		for i, sequence in enumerate( sequence_list ):
 			# add <bos> to text for STT
@@ -300,11 +302,13 @@ class AR_NAR(Base):
 			# naturally, rep pen wrangles this initial burst of noise, but naively relying on rep_pen is no good, as it fails after ~6 seconds of audio
 			# however, switching to a default sampling temperature with "clean greedy sampled codes" will make the rest of sequence sound as if it were greedy sampled
 			# to-do: tune these values, maybe have it factor based on confidence scores or something
+			"""
 			if low_temperature:
 				enabled = n < low_temperature_range
-				sampling_repetition_penalty = 1.125 if enabled else original_sampling_repetition_penalty
-				sampling_repetition_penalty_decay = 0.0 if enabled else original_sampling_repetition_penalty_decay
-				sampling_temperature = original_sampling_temperature if enabled else 1.0
+				sampling_repetition_penalty = 1.125 if enabled else 1.25
+				#sampling_repetition_penalty_decay = 0.0 if enabled else original_sampling_repetition_penalty_decay
+				#sampling_temperature = original_sampling_temperature if enabled else 1.0
+			"""
 
 			inputs = self.inputs(
 				text_list=text_list,
@@ -351,7 +355,14 @@ class AR_NAR(Base):
 			r = sampled[0]
 
 			if sampled.entropy:
-				entropies.append( sampled.entropy )
+				metrics.append( sampled.entropy )
+				"""
+				elif sampled.confidence:
+					metrics.append( sampled.confidence )
+				"""
+			elif False:
+				p = [ { "p": torch.nn.functional.softmax(logit[-1, :].cpu(), dim=0)[token.item()].item() } for logit, token in zip(logits, r) ]
+				metrics.append( p )
 
 			if mirostat is not None:
 				mirostat = sampled.scores
@@ -381,9 +392,9 @@ class AR_NAR(Base):
 			if stopped.all().item():
 				break
 
-		if entropies:
-			from ..plot import plot_entropies
-			plot_entropies( entropies )
+		if metrics:
+			from ..plot import plot_sample_metrics
+			plot_sample_metrics( metrics )
 
 		# pick the best scoring candidate
 		# desu this is always going to be candidate 0

vall_e/plot.py

@@ -93,13 +93,13 @@ def plot(paths, args):
 		#bbox_to_anchor=(1.04, 0.5),
 	)
 
-def plot_entropies( entropies ):
+def plot_sample_metrics( metrics ):
 	"""
 	fig = plt.figure()
-	fig.set_figwidth( 16 * len(entropies) // cfg.dataset.frames_per_second )
+	fig.set_figwidth( 16 * len(metrics) // cfg.dataset.frames_per_second )
 	"""
 
-	data = { key: [ e[0][key] for e in entropies ] for key in entropies[0][0].keys() }
+	data = { key: [ e[0][key] for e in metrics ] for key in metrics[0][0].keys() }
 
 	df = pd.DataFrame(data)
 	df.plot()

vall_e/webui.py

@@ -134,8 +134,8 @@ def do_inference_tts( progress=gr.Progress(track_tqdm=True), *args, **kwargs ):
 		raise Exception("No YAML loaded.")
 
 	if kwargs.pop("dynamic-sampling", False):
-		kwargs['min-ar-temp'] = 0.85 if kwargs['ar-temp'] > 0.85 else 0.0
-		kwargs['min-nar-temp'] = 0.85 if kwargs['nar-temp'] > 0.85 else 0.0 # should probably disable it for the NAR
+		kwargs['min-ar-temp'] = 0.01 if kwargs['ar-temp'] > 0.01 else 0.0
+		kwargs['min-nar-temp'] = 0.0 # 0.85 if kwargs['nar-temp'] > 0.85 else 0.0 # should probably disable it for the NAR
 	else:
 		kwargs['min-ar-temp'] = -1
 		kwargs['min-nar-temp'] = -1