python/FunASR-XL.git

			@@ -248,7 +248,7 @@
			concat_after: bool = False,
			att_layer_num: int = 6,
			kernel_size: int = 21,
			sanm_shift: int = 0,
			sanm_shfit: int = 0,
			lora_list: List[str] = None,
			lora_rank: int = 8,
			lora_alpha: int = 16,
			@@ -298,14 +298,14 @@

			self.att_layer_num = att_layer_num
			self.num_blocks = num_blocks
			if sanm_shift is None:
			sanm_shift = (kernel_size - 1) // 2
			if sanm_shfit is None:
			sanm_shfit = (kernel_size - 1) // 2
			self.decoders = repeat(
			att_layer_num,
			lambda lnum: DecoderLayerSANM(
			attention_dim,
			MultiHeadedAttentionSANMDecoder(
			attention_dim, self_attention_dropout_rate, kernel_size, sanm_shift=sanm_shift
			attention_dim, self_attention_dropout_rate, kernel_size, sanm_shfit=sanm_shfit
			),
			MultiHeadedAttentionCrossAtt(
			attention_heads,
			@@ -330,7 +330,7 @@
			lambda lnum: DecoderLayerSANM(
			attention_dim,
			MultiHeadedAttentionSANMDecoder(
			attention_dim, self_attention_dropout_rate, kernel_size, sanm_shift=0
			attention_dim, self_attention_dropout_rate, kernel_size, sanm_shfit=0
			),
			None,
			PositionwiseFeedForwardDecoderSANM(attention_dim, linear_units, dropout_rate),
			@@ -785,20 +785,20 @@
			for _ in range(cache_num)
			]
			return (tgt, memory, pre_acoustic_embeds, cache)


			def is_optimizable(self):
			return True


			def get_input_names(self):
			cache_num = len(self.model.decoders) + len(self.model.decoders2)
			return ['tgt', 'memory', 'pre_acoustic_embeds'] \
			+ ['cache_%d' % i for i in range(cache_num)]


			def get_output_names(self):
			cache_num = len(self.model.decoders) + len(self.model.decoders2)
			return ['y'] \
			+ ['out_cache_%d' % i for i in range(cache_num)]


			def get_dynamic_axes(self):
			ret = {
			'tgt': {