libritts/model.txt

VALLE(
 (ar_text_embedding): TokenEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 (word_embeddings): Embedding(512, 1024)
 )
 (nar_text_embedding): TokenEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 (word_embeddings): Embedding(512, 1024)
 )
 (ar_audio_embedding): TokenEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 (word_embeddings): Embedding(1025, 1024)
 )
 (ar_text_prenet): Identity()
 (ar_audio_prenet): Identity()
 (ar_text_position): SinePositionalEmbedding(
 (dropout): Dropout(p=0.1, inplace=False)
 )
 (ar_audio_position): SinePositionalEmbedding(
 (dropout): Dropout(p=0.1, inplace=False)
 )
 (ar_decoder): TransformerEncoder(
 (layers): ModuleList(
 (0-11): 12 x TransformerEncoderLayer(
 (self_attn): MultiheadAttention(
 (out_proj): NonDynamicallyQuantizableLinear(in_features=1024, out_features=1024, bias=True)
 )
 (linear1): Linear(in_features=1024, out_features=4096, bias=True)
 (dropout): Dropout(p=0.1, inplace=False)
 (linear2): Linear(in_features=4096, out_features=1024, bias=True)
 (dropout1): Dropout(p=0.1, inplace=False)
 (dropout2): Dropout(p=0.1, inplace=False)
 (norm1): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
 (norm2): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
 )
 )
 (norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
 )
 (ar_predict_layer): Linear(in_features=1024, out_features=1025, bias=False)
 (ar_accuracy_metric): MulticlassAccuracy()
 (nar_audio_embeddings): ModuleList(
 (0): TokenEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 (word_embeddings): Embedding(1025, 1024)
 )
 (1-7): 7 x TokenEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 (word_embeddings): Embedding(1024, 1024)
 )
 )
 (nar_text_prenet): Identity()
 (nar_audio_prenet): Identity()
 (nar_text_position): SinePositionalEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 )
 (nar_audio_position): SinePositionalEmbedding(
 (dropout): Dropout(p=0.1, inplace=False)
 )
 (nar_decoder): TransformerEncoder(
 (layers): ModuleList(
 (0-11): 12 x TransformerEncoderLayer(
 (self_attn): MultiheadAttention(
 (out_proj): NonDynamicallyQuantizableLinear(in_features=1024, out_features=1024, bias=True)
 )
 (linear1): Linear(in_features=1024, out_features=4096, bias=True)
 (dropout): Dropout(p=0.1, inplace=False)
 (linear2): Linear(in_features=4096, out_features=1024, bias=True)
 (dropout1): Dropout(p=0.1, inplace=False)
 (dropout2): Dropout(p=0.1, inplace=False)
 (norm1): AdaptiveLayerNorm(
 (project_layer): Linear(in_features=1024, out_features=2048, bias=True)
 (norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
 )
 (norm2): AdaptiveLayerNorm(
 (project_layer): Linear(in_features=1024, out_features=2048, bias=True)
 (norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
 )
 )
 )
 (norm): AdaptiveLayerNorm(
 (project_layer): Linear(in_features=1024, out_features=2048, bias=True)
 (norm): LayerNorm((1024,), eps=1e-05, elementwise_affine=True)
 )
 )
 (nar_predict_layers): ModuleList(
 (0-6): 7 x Linear(in_features=1024, out_features=1024, bias=False)
 )
 (nar_stage_embeddings): ModuleList(
 (0-6): 7 x TokenEmbedding(
 (dropout): Dropout(p=0.0, inplace=False)
 (word_embeddings): Embedding(1, 1024)
 )
 )
 (nar_accuracy_metric): MulticlassAccuracy()
)