god it would be nice to know the best way to handle audio embeddings, because I genuinely don't know without skimming through papers or devoting X amount of GPU hours in training

vall_e/config.py

@@ -190,6 +190,7 @@ class Model:
 	p_ar_level: float | str = "auto" # determines odds of selecting the AR (level 0) when training, "auto" for default behavior
 	frozen_params: list[str] = field(default_factory=lambda: []) # frozen parameters that are not updated when training
 	attention: str = "eager" # or flash_attention_2
+	audio_embedding_sums: bool = True
 
 	def get(self, name=None):
 		return [ self ] if not name or self.name == name else []

vall_e/models/base.py

@@ -174,11 +174,8 @@ class Embedding(nn.Embedding):
 		return super().forward(torch.cat(x_list)).split([*map(len, x_list)])
 """
 
+# Deprecated implementation
 class MultiEmbedding(nn.Module):
-	"""
-	This embedding sums embeddings on different levels.
-	"""
-
 	def __init__(self, max_n_levels, n_tokens, token_dim, monolithic=False):
 		super().__init__()
 		self.monolithic = monolithic
@@ -216,21 +213,41 @@ class MultiEmbedding(nn.Module):
 
 # Embedding that sums each RVQ-bin level within a given input acoustic prompt
 class AudioEmbedding(nn.Module):
-	def __init__(self, l_tokens, token_dim, levels=None):
+	def __init__(
+		self,
+		l_tokens: int, # list of number of tokens (needed because AR resps includes stop token)
+		token_dim: int, # dimensionality of the embedding
+		levels: int | None = None, # number of RVQ-bins (I don't remember the specifics)
+		sums: bool = True # whether to sum all previous layers of embeddings to factor in other RVQ bin levels (I do not know which way is better)
+	):
 		super().__init__()
+		# array of embeddings
+		#   proms are [0, prom_levels]
+		#   resp are split to where [0] is for the AR, and [1:] are reserved for NAR
 		self.embeddings = nn.ModuleList([nn.Embedding(n_tokens, token_dim) for n_tokens in l_tokens])
+		# weight influencer for the influence for each level (desu this should be really useless because the weights in the embedding themselves should factor this)
 		self.weight = nn.ParameterList([nn.Parameter( torch.Tensor([1]) ) for i in range(levels)]) if levels is not None else None
+		# 
+		self.sums = sums
 	
 	def forward(self, xi: Tensor, quant_levels: Tensor | None = None ) -> Tensor:
 		# prom
 		if quant_levels is None and xi.shape[-1] > 1:
-			x = sum( [ self.embeddings[k]( xi[:, k] ) * (self.weight[k] if self.weight is not None else 1) for k in range(xi.shape[-1]) ] )
+			if self.sums:
+				x = sum( [ self.embeddings[k]( xi[:, k] ) * (self.weight[k] if self.weight is not None else 1) for k in range(xi.shape[-1]) ] )
+			else:
+				k = 0 # only use the most significant RVQ bin level for the input prom
+				x = self.embeddings[k]( xi[:, k] ) * (self.weight[k] if self.weight is not None else 1)
 		# AR resp
 		elif quant_levels is None or quant_levels == 0:
 			x = self.embeddings[0]( xi[:, 0] )
 		# NAR resp
 		else:
-			x = sum( [ self.embeddings[k+1]( xi[:, k] ) * (self.weight[k+1] if self.weight is not None else 1) for k in range(xi.shape[-1]) ] )
+			if self.sums:
+				x = sum( [ self.embeddings[k+1]( xi[:, k] ) * (self.weight[k+1] if self.weight is not None else 1) for k in range(xi.shape[-1]) ] )
+			else:
+				k = xi.shape[-1] - 1 # only use the previous RVQ bin level for the current resp embedding
+				x = self.embeddings[k+1]( xi[:, k] ) * (self.weight[k+1] if self.weight is not None else 1)
 		
 		return x
 
@@ -345,9 +362,17 @@ class Base(nn.Module):
 			self.resps_emb = MultiEmbedding(self.n_resp_levels, n_resp_tokens, d_model, monolithic=self.monolithic)
 		else:
 			# [1024] * 8
-			self.proms_emb = AudioEmbedding([n_prom_tokens] * self.n_prom_levels, d_model, self.n_prom_levels if self.version > 3 else None)
+			self.proms_emb = AudioEmbedding(
+				[n_prom_tokens] * self.n_prom_levels, d_model,
+				levels=self.n_prom_levels if self.version > 3 else None,
+				sums=self.config.audio_embedding_sums
+			)
 			# [1025] + [1024] * 8
-			self.resps_emb = AudioEmbedding([n_resp_tokens] + [n_resp_tokens - 1] * (self.n_resp_levels - 1), d_model, self.n_resp_levels if self.version > 3 else None)
+			self.resps_emb = AudioEmbedding(
+				[n_resp_tokens] + [n_resp_tokens - 1] * (self.n_resp_levels - 1), d_model,
+				levels=self.n_resp_levels if self.version > 3 else None,
+				sums=self.config.audio_embedding_sums
+			)
 
 		
 		if self.version >= 3: