y-ren16/MCLP-RPTTS

1flow: !new:cosyvoice2.flow.flow.CausalMaskedDiffWithXvec2    input_size: 5123    output_size: 804    spk_embed_dim: 1925    output_type: 'mel'6    vocab_size: 65617    encoder: !new:cosyvoice2.transformer.upsample_encoder_v2.UpsampleConformerEncoderV28        input_size: 5129        output_size: 51210        input_layer: 'linear'11        pre_lookahead_len: 312        num_blocks: 613        num_up_blocks: 414        up_stride: 215        up_scale_factor: 216        attention_heads: 817        pos_enc_layer_type: 'rel_pos_espnet'18        selfattention_layer_type: 'rel_selfattn'19        key_bias: true20        linear_units: 204821        dropout_rate: 0.122        positional_dropout_rate: 0.123        attention_dropout_rate: 0.124        normalize_before: True25    decoder: !new:cosyvoice2.flow.flow_matching.CausalConditionalCFM26        inference_cfg_rate: 0.727        estimator: !new:cosyvoice2.flow.decoder_dit.DiT28            in_channels: 32029            out_channels: 8030            mlp_ratio: 4.031            depth: 1632            num_heads: 833            head_dim: 6434            hidden_size: 51235