ACE-Step/save_checkpoint.ipynb

In [5]:

import torch

checkpoint_path = "/root/sag_train/checkpoints/aceflow3_250401/epoch=22-step=460k_pretrained_ft_240k_small_repa.ckpt"

checkpoint = torch.load(checkpoint_path, map_location="cpu")

/tmp/ipykernel_3367140/2080906558.py:5: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
  checkpoint = torch.load(checkpoint_path, map_location="cpu")

In [6]:

checkpoint.keys()

Out [6]:

odict_keys(['resampler_mert.kernel', 'resampler_mhubert.kernel', 'vae.mel_transform.spectrogram.window', 'vae.mel_transform.mel_scale.fb', 'vae.vocoder.mel_transform.spectrogram.window', 'vae.vocoder.mel_transform.mel_scale.fb', 'vae.resampler.kernel', 'lyric_processor.lyric_text_model.shared.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.0.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.1.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.2.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.3.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.4.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.5.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.6.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.7.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.8.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.9.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.10.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.0.SelfAttention.q.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.0.SelfAttention.k.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.0.SelfAttention.v.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.0.SelfAttention.o.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.0.SelfAttention.relative_attention_bias.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.0.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.1.DenseReluDense.wi_0.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.1.DenseReluDense.wi_1.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.1.DenseReluDense.wo.weight', 'lyric_processor.lyric_text_model.encoder.block.11.layer.1.layer_norm.weight', 'lyric_processor.lyric_text_model.encoder.final_layer_norm.weight', 'mert_model.masked_spec_embed', 'mert_model.feature_extractor.conv_layers.0.conv.weight', 'mert_model.feature_extractor.conv_layers.0.layer_norm.weight', 'mert_model.feature_extractor.conv_layers.0.layer_norm.bias', 'mert_model.feature_extractor.conv_layers.1.conv.weight', 'mert_model.feature_extractor.conv_layers.2.conv.weight', 'mert_model.feature_extractor.conv_layers.3.conv.weight', 'mert_model.feature_extractor.conv_layers.4.conv.weight', 'mert_model.feature_extractor.conv_layers.5.conv.weight', 'mert_model.feature_extractor.conv_layers.6.conv.weight', 'mert_model.feature_projection.layer_norm.weight', 'mert_model.feature_projection.layer_norm.bias', 'mert_model.feature_projection.projection.weight', 'mert_model.feature_projection.projection.bias', 'mert_model.encoder.pos_conv_embed.conv.bias', 'mert_model.encoder.pos_conv_embed.conv.parametrizations.weight.original0', 'mert_model.encoder.pos_conv_embed.conv.parametrizations.weight.original1', 'mert_model.encoder.layer_norm.weight', 'mert_model.encoder.layer_norm.bias', 'mert_model.encoder.layers.0.attention.k_proj.weight', 'mert_model.encoder.layers.0.attention.k_proj.bias', 'mert_model.encoder.layers.0.attention.v_proj.weight', 'mert_model.encoder.layers.0.attention.v_proj.bias', 'mert_model.encoder.layers.0.attention.q_proj.weight', 'mert_model.encoder.layers.0.attention.q_proj.bias', 'mert_model.encoder.layers.0.attention.out_proj.weight', 'mert_model.encoder.layers.0.attention.out_proj.bias', 'mert_model.encoder.layers.0.layer_norm.weight', 'mert_model.encoder.layers.0.layer_norm.bias', 'mert_model.encoder.layers.0.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.0.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.0.feed_forward.output_dense.weight', 'mert_model.encoder.layers.0.feed_forward.output_dense.bias', 'mert_model.encoder.layers.0.final_layer_norm.weight', 'mert_model.encoder.layers.0.final_layer_norm.bias', 'mert_model.encoder.layers.1.attention.k_proj.weight', 'mert_model.encoder.layers.1.attention.k_proj.bias', 'mert_model.encoder.layers.1.attention.v_proj.weight', 'mert_model.encoder.layers.1.attention.v_proj.bias', 'mert_model.encoder.layers.1.attention.q_proj.weight', 'mert_model.encoder.layers.1.attention.q_proj.bias', 'mert_model.encoder.layers.1.attention.out_proj.weight', 'mert_model.encoder.layers.1.attention.out_proj.bias', 'mert_model.encoder.layers.1.layer_norm.weight', 'mert_model.encoder.layers.1.layer_norm.bias', 'mert_model.encoder.layers.1.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.1.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.1.feed_forward.output_dense.weight', 'mert_model.encoder.layers.1.feed_forward.output_dense.bias', 'mert_model.encoder.layers.1.final_layer_norm.weight', 'mert_model.encoder.layers.1.final_layer_norm.bias', 'mert_model.encoder.layers.2.attention.k_proj.weight', 'mert_model.encoder.layers.2.attention.k_proj.bias', 'mert_model.encoder.layers.2.attention.v_proj.weight', 'mert_model.encoder.layers.2.attention.v_proj.bias', 'mert_model.encoder.layers.2.attention.q_proj.weight', 'mert_model.encoder.layers.2.attention.q_proj.bias', 'mert_model.encoder.layers.2.attention.out_proj.weight', 'mert_model.encoder.layers.2.attention.out_proj.bias', 'mert_model.encoder.layers.2.layer_norm.weight', 'mert_model.encoder.layers.2.layer_norm.bias', 'mert_model.encoder.layers.2.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.2.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.2.feed_forward.output_dense.weight', 'mert_model.encoder.layers.2.feed_forward.output_dense.bias', 'mert_model.encoder.layers.2.final_layer_norm.weight', 'mert_model.encoder.layers.2.final_layer_norm.bias', 'mert_model.encoder.layers.3.attention.k_proj.weight', 'mert_model.encoder.layers.3.attention.k_proj.bias', 'mert_model.encoder.layers.3.attention.v_proj.weight', 'mert_model.encoder.layers.3.attention.v_proj.bias', 'mert_model.encoder.layers.3.attention.q_proj.weight', 'mert_model.encoder.layers.3.attention.q_proj.bias', 'mert_model.encoder.layers.3.attention.out_proj.weight', 'mert_model.encoder.layers.3.attention.out_proj.bias', 'mert_model.encoder.layers.3.layer_norm.weight', 'mert_model.encoder.layers.3.layer_norm.bias', 'mert_model.encoder.layers.3.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.3.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.3.feed_forward.output_dense.weight', 'mert_model.encoder.layers.3.feed_forward.output_dense.bias', 'mert_model.encoder.layers.3.final_layer_norm.weight', 'mert_model.encoder.layers.3.final_layer_norm.bias', 'mert_model.encoder.layers.4.attention.k_proj.weight', 'mert_model.encoder.layers.4.attention.k_proj.bias', 'mert_model.encoder.layers.4.attention.v_proj.weight', 'mert_model.encoder.layers.4.attention.v_proj.bias', 'mert_model.encoder.layers.4.attention.q_proj.weight', 'mert_model.encoder.layers.4.attention.q_proj.bias', 'mert_model.encoder.layers.4.attention.out_proj.weight', 'mert_model.encoder.layers.4.attention.out_proj.bias', 'mert_model.encoder.layers.4.layer_norm.weight', 'mert_model.encoder.layers.4.layer_norm.bias', 'mert_model.encoder.layers.4.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.4.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.4.feed_forward.output_dense.weight', 'mert_model.encoder.layers.4.feed_forward.output_dense.bias', 'mert_model.encoder.layers.4.final_layer_norm.weight', 'mert_model.encoder.layers.4.final_layer_norm.bias', 'mert_model.encoder.layers.5.attention.k_proj.weight', 'mert_model.encoder.layers.5.attention.k_proj.bias', 'mert_model.encoder.layers.5.attention.v_proj.weight', 'mert_model.encoder.layers.5.attention.v_proj.bias', 'mert_model.encoder.layers.5.attention.q_proj.weight', 'mert_model.encoder.layers.5.attention.q_proj.bias', 'mert_model.encoder.layers.5.attention.out_proj.weight', 'mert_model.encoder.layers.5.attention.out_proj.bias', 'mert_model.encoder.layers.5.layer_norm.weight', 'mert_model.encoder.layers.5.layer_norm.bias', 'mert_model.encoder.layers.5.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.5.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.5.feed_forward.output_dense.weight', 'mert_model.encoder.layers.5.feed_forward.output_dense.bias', 'mert_model.encoder.layers.5.final_layer_norm.weight', 'mert_model.encoder.layers.5.final_layer_norm.bias', 'mert_model.encoder.layers.6.attention.k_proj.weight', 'mert_model.encoder.layers.6.attention.k_proj.bias', 'mert_model.encoder.layers.6.attention.v_proj.weight', 'mert_model.encoder.layers.6.attention.v_proj.bias', 'mert_model.encoder.layers.6.attention.q_proj.weight', 'mert_model.encoder.layers.6.attention.q_proj.bias', 'mert_model.encoder.layers.6.attention.out_proj.weight', 'mert_model.encoder.layers.6.attention.out_proj.bias', 'mert_model.encoder.layers.6.layer_norm.weight', 'mert_model.encoder.layers.6.layer_norm.bias', 'mert_model.encoder.layers.6.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.6.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.6.feed_forward.output_dense.weight', 'mert_model.encoder.layers.6.feed_forward.output_dense.bias', 'mert_model.encoder.layers.6.final_layer_norm.weight', 'mert_model.encoder.layers.6.final_layer_norm.bias', 'mert_model.encoder.layers.7.attention.k_proj.weight', 'mert_model.encoder.layers.7.attention.k_proj.bias', 'mert_model.encoder.layers.7.attention.v_proj.weight', 'mert_model.encoder.layers.7.attention.v_proj.bias', 'mert_model.encoder.layers.7.attention.q_proj.weight', 'mert_model.encoder.layers.7.attention.q_proj.bias', 'mert_model.encoder.layers.7.attention.out_proj.weight', 'mert_model.encoder.layers.7.attention.out_proj.bias', 'mert_model.encoder.layers.7.layer_norm.weight', 'mert_model.encoder.layers.7.layer_norm.bias', 'mert_model.encoder.layers.7.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.7.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.7.feed_forward.output_dense.weight', 'mert_model.encoder.layers.7.feed_forward.output_dense.bias', 'mert_model.encoder.layers.7.final_layer_norm.weight', 'mert_model.encoder.layers.7.final_layer_norm.bias', 'mert_model.encoder.layers.8.attention.k_proj.weight', 'mert_model.encoder.layers.8.attention.k_proj.bias', 'mert_model.encoder.layers.8.attention.v_proj.weight', 'mert_model.encoder.layers.8.attention.v_proj.bias', 'mert_model.encoder.layers.8.attention.q_proj.weight', 'mert_model.encoder.layers.8.attention.q_proj.bias', 'mert_model.encoder.layers.8.attention.out_proj.weight', 'mert_model.encoder.layers.8.attention.out_proj.bias', 'mert_model.encoder.layers.8.layer_norm.weight', 'mert_model.encoder.layers.8.layer_norm.bias', 'mert_model.encoder.layers.8.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.8.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.8.feed_forward.output_dense.weight', 'mert_model.encoder.layers.8.feed_forward.output_dense.bias', 'mert_model.encoder.layers.8.final_layer_norm.weight', 'mert_model.encoder.layers.8.final_layer_norm.bias', 'mert_model.encoder.layers.9.attention.k_proj.weight', 'mert_model.encoder.layers.9.attention.k_proj.bias', 'mert_model.encoder.layers.9.attention.v_proj.weight', 'mert_model.encoder.layers.9.attention.v_proj.bias', 'mert_model.encoder.layers.9.attention.q_proj.weight', 'mert_model.encoder.layers.9.attention.q_proj.bias', 'mert_model.encoder.layers.9.attention.out_proj.weight', 'mert_model.encoder.layers.9.attention.out_proj.bias', 'mert_model.encoder.layers.9.layer_norm.weight', 'mert_model.encoder.layers.9.layer_norm.bias', 'mert_model.encoder.layers.9.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.9.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.9.feed_forward.output_dense.weight', 'mert_model.encoder.layers.9.feed_forward.output_dense.bias', 'mert_model.encoder.layers.9.final_layer_norm.weight', 'mert_model.encoder.layers.9.final_layer_norm.bias', 'mert_model.encoder.layers.10.attention.k_proj.weight', 'mert_model.encoder.layers.10.attention.k_proj.bias', 'mert_model.encoder.layers.10.attention.v_proj.weight', 'mert_model.encoder.layers.10.attention.v_proj.bias', 'mert_model.encoder.layers.10.attention.q_proj.weight', 'mert_model.encoder.layers.10.attention.q_proj.bias', 'mert_model.encoder.layers.10.attention.out_proj.weight', 'mert_model.encoder.layers.10.attention.out_proj.bias', 'mert_model.encoder.layers.10.layer_norm.weight', 'mert_model.encoder.layers.10.layer_norm.bias', 'mert_model.encoder.layers.10.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.10.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.10.feed_forward.output_dense.weight', 'mert_model.encoder.layers.10.feed_forward.output_dense.bias', 'mert_model.encoder.layers.10.final_layer_norm.weight', 'mert_model.encoder.layers.10.final_layer_norm.bias', 'mert_model.encoder.layers.11.attention.k_proj.weight', 'mert_model.encoder.layers.11.attention.k_proj.bias', 'mert_model.encoder.layers.11.attention.v_proj.weight', 'mert_model.encoder.layers.11.attention.v_proj.bias', 'mert_model.encoder.layers.11.attention.q_proj.weight', 'mert_model.encoder.layers.11.attention.q_proj.bias', 'mert_model.encoder.layers.11.attention.out_proj.weight', 'mert_model.encoder.layers.11.attention.out_proj.bias', 'mert_model.encoder.layers.11.layer_norm.weight', 'mert_model.encoder.layers.11.layer_norm.bias', 'mert_model.encoder.layers.11.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.11.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.11.feed_forward.output_dense.weight', 'mert_model.encoder.layers.11.feed_forward.output_dense.bias', 'mert_model.encoder.layers.11.final_layer_norm.weight', 'mert_model.encoder.layers.11.final_layer_norm.bias', 'mert_model.encoder.layers.12.attention.k_proj.weight', 'mert_model.encoder.layers.12.attention.k_proj.bias', 'mert_model.encoder.layers.12.attention.v_proj.weight', 'mert_model.encoder.layers.12.attention.v_proj.bias', 'mert_model.encoder.layers.12.attention.q_proj.weight', 'mert_model.encoder.layers.12.attention.q_proj.bias', 'mert_model.encoder.layers.12.attention.out_proj.weight', 'mert_model.encoder.layers.12.attention.out_proj.bias', 'mert_model.encoder.layers.12.layer_norm.weight', 'mert_model.encoder.layers.12.layer_norm.bias', 'mert_model.encoder.layers.12.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.12.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.12.feed_forward.output_dense.weight', 'mert_model.encoder.layers.12.feed_forward.output_dense.bias', 'mert_model.encoder.layers.12.final_layer_norm.weight', 'mert_model.encoder.layers.12.final_layer_norm.bias', 'mert_model.encoder.layers.13.attention.k_proj.weight', 'mert_model.encoder.layers.13.attention.k_proj.bias', 'mert_model.encoder.layers.13.attention.v_proj.weight', 'mert_model.encoder.layers.13.attention.v_proj.bias', 'mert_model.encoder.layers.13.attention.q_proj.weight', 'mert_model.encoder.layers.13.attention.q_proj.bias', 'mert_model.encoder.layers.13.attention.out_proj.weight', 'mert_model.encoder.layers.13.attention.out_proj.bias', 'mert_model.encoder.layers.13.layer_norm.weight', 'mert_model.encoder.layers.13.layer_norm.bias', 'mert_model.encoder.layers.13.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.13.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.13.feed_forward.output_dense.weight', 'mert_model.encoder.layers.13.feed_forward.output_dense.bias', 'mert_model.encoder.layers.13.final_layer_norm.weight', 'mert_model.encoder.layers.13.final_layer_norm.bias', 'mert_model.encoder.layers.14.attention.k_proj.weight', 'mert_model.encoder.layers.14.attention.k_proj.bias', 'mert_model.encoder.layers.14.attention.v_proj.weight', 'mert_model.encoder.layers.14.attention.v_proj.bias', 'mert_model.encoder.layers.14.attention.q_proj.weight', 'mert_model.encoder.layers.14.attention.q_proj.bias', 'mert_model.encoder.layers.14.attention.out_proj.weight', 'mert_model.encoder.layers.14.attention.out_proj.bias', 'mert_model.encoder.layers.14.layer_norm.weight', 'mert_model.encoder.layers.14.layer_norm.bias', 'mert_model.encoder.layers.14.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.14.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.14.feed_forward.output_dense.weight', 'mert_model.encoder.layers.14.feed_forward.output_dense.bias', 'mert_model.encoder.layers.14.final_layer_norm.weight', 'mert_model.encoder.layers.14.final_layer_norm.bias', 'mert_model.encoder.layers.15.attention.k_proj.weight', 'mert_model.encoder.layers.15.attention.k_proj.bias', 'mert_model.encoder.layers.15.attention.v_proj.weight', 'mert_model.encoder.layers.15.attention.v_proj.bias', 'mert_model.encoder.layers.15.attention.q_proj.weight', 'mert_model.encoder.layers.15.attention.q_proj.bias', 'mert_model.encoder.layers.15.attention.out_proj.weight', 'mert_model.encoder.layers.15.attention.out_proj.bias', 'mert_model.encoder.layers.15.layer_norm.weight', 'mert_model.encoder.layers.15.layer_norm.bias', 'mert_model.encoder.layers.15.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.15.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.15.feed_forward.output_dense.weight', 'mert_model.encoder.layers.15.feed_forward.output_dense.bias', 'mert_model.encoder.layers.15.final_layer_norm.weight', 'mert_model.encoder.layers.15.final_layer_norm.bias', 'mert_model.encoder.layers.16.attention.k_proj.weight', 'mert_model.encoder.layers.16.attention.k_proj.bias', 'mert_model.encoder.layers.16.attention.v_proj.weight', 'mert_model.encoder.layers.16.attention.v_proj.bias', 'mert_model.encoder.layers.16.attention.q_proj.weight', 'mert_model.encoder.layers.16.attention.q_proj.bias', 'mert_model.encoder.layers.16.attention.out_proj.weight', 'mert_model.encoder.layers.16.attention.out_proj.bias', 'mert_model.encoder.layers.16.layer_norm.weight', 'mert_model.encoder.layers.16.layer_norm.bias', 'mert_model.encoder.layers.16.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.16.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.16.feed_forward.output_dense.weight', 'mert_model.encoder.layers.16.feed_forward.output_dense.bias', 'mert_model.encoder.layers.16.final_layer_norm.weight', 'mert_model.encoder.layers.16.final_layer_norm.bias', 'mert_model.encoder.layers.17.attention.k_proj.weight', 'mert_model.encoder.layers.17.attention.k_proj.bias', 'mert_model.encoder.layers.17.attention.v_proj.weight', 'mert_model.encoder.layers.17.attention.v_proj.bias', 'mert_model.encoder.layers.17.attention.q_proj.weight', 'mert_model.encoder.layers.17.attention.q_proj.bias', 'mert_model.encoder.layers.17.attention.out_proj.weight', 'mert_model.encoder.layers.17.attention.out_proj.bias', 'mert_model.encoder.layers.17.layer_norm.weight', 'mert_model.encoder.layers.17.layer_norm.bias', 'mert_model.encoder.layers.17.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.17.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.17.feed_forward.output_dense.weight', 'mert_model.encoder.layers.17.feed_forward.output_dense.bias', 'mert_model.encoder.layers.17.final_layer_norm.weight', 'mert_model.encoder.layers.17.final_layer_norm.bias', 'mert_model.encoder.layers.18.attention.k_proj.weight', 'mert_model.encoder.layers.18.attention.k_proj.bias', 'mert_model.encoder.layers.18.attention.v_proj.weight', 'mert_model.encoder.layers.18.attention.v_proj.bias', 'mert_model.encoder.layers.18.attention.q_proj.weight', 'mert_model.encoder.layers.18.attention.q_proj.bias', 'mert_model.encoder.layers.18.attention.out_proj.weight', 'mert_model.encoder.layers.18.attention.out_proj.bias', 'mert_model.encoder.layers.18.layer_norm.weight', 'mert_model.encoder.layers.18.layer_norm.bias', 'mert_model.encoder.layers.18.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.18.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.18.feed_forward.output_dense.weight', 'mert_model.encoder.layers.18.feed_forward.output_dense.bias', 'mert_model.encoder.layers.18.final_layer_norm.weight', 'mert_model.encoder.layers.18.final_layer_norm.bias', 'mert_model.encoder.layers.19.attention.k_proj.weight', 'mert_model.encoder.layers.19.attention.k_proj.bias', 'mert_model.encoder.layers.19.attention.v_proj.weight', 'mert_model.encoder.layers.19.attention.v_proj.bias', 'mert_model.encoder.layers.19.attention.q_proj.weight', 'mert_model.encoder.layers.19.attention.q_proj.bias', 'mert_model.encoder.layers.19.attention.out_proj.weight', 'mert_model.encoder.layers.19.attention.out_proj.bias', 'mert_model.encoder.layers.19.layer_norm.weight', 'mert_model.encoder.layers.19.layer_norm.bias', 'mert_model.encoder.layers.19.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.19.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.19.feed_forward.output_dense.weight', 'mert_model.encoder.layers.19.feed_forward.output_dense.bias', 'mert_model.encoder.layers.19.final_layer_norm.weight', 'mert_model.encoder.layers.19.final_layer_norm.bias', 'mert_model.encoder.layers.20.attention.k_proj.weight', 'mert_model.encoder.layers.20.attention.k_proj.bias', 'mert_model.encoder.layers.20.attention.v_proj.weight', 'mert_model.encoder.layers.20.attention.v_proj.bias', 'mert_model.encoder.layers.20.attention.q_proj.weight', 'mert_model.encoder.layers.20.attention.q_proj.bias', 'mert_model.encoder.layers.20.attention.out_proj.weight', 'mert_model.encoder.layers.20.attention.out_proj.bias', 'mert_model.encoder.layers.20.layer_norm.weight', 'mert_model.encoder.layers.20.layer_norm.bias', 'mert_model.encoder.layers.20.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.20.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.20.feed_forward.output_dense.weight', 'mert_model.encoder.layers.20.feed_forward.output_dense.bias', 'mert_model.encoder.layers.20.final_layer_norm.weight', 'mert_model.encoder.layers.20.final_layer_norm.bias', 'mert_model.encoder.layers.21.attention.k_proj.weight', 'mert_model.encoder.layers.21.attention.k_proj.bias', 'mert_model.encoder.layers.21.attention.v_proj.weight', 'mert_model.encoder.layers.21.attention.v_proj.bias', 'mert_model.encoder.layers.21.attention.q_proj.weight', 'mert_model.encoder.layers.21.attention.q_proj.bias', 'mert_model.encoder.layers.21.attention.out_proj.weight', 'mert_model.encoder.layers.21.attention.out_proj.bias', 'mert_model.encoder.layers.21.layer_norm.weight', 'mert_model.encoder.layers.21.layer_norm.bias', 'mert_model.encoder.layers.21.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.21.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.21.feed_forward.output_dense.weight', 'mert_model.encoder.layers.21.feed_forward.output_dense.bias', 'mert_model.encoder.layers.21.final_layer_norm.weight', 'mert_model.encoder.layers.21.final_layer_norm.bias', 'mert_model.encoder.layers.22.attention.k_proj.weight', 'mert_model.encoder.layers.22.attention.k_proj.bias', 'mert_model.encoder.layers.22.attention.v_proj.weight', 'mert_model.encoder.layers.22.attention.v_proj.bias', 'mert_model.encoder.layers.22.attention.q_proj.weight', 'mert_model.encoder.layers.22.attention.q_proj.bias', 'mert_model.encoder.layers.22.attention.out_proj.weight', 'mert_model.encoder.layers.22.attention.out_proj.bias', 'mert_model.encoder.layers.22.layer_norm.weight', 'mert_model.encoder.layers.22.layer_norm.bias', 'mert_model.encoder.layers.22.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.22.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.22.feed_forward.output_dense.weight', 'mert_model.encoder.layers.22.feed_forward.output_dense.bias', 'mert_model.encoder.layers.22.final_layer_norm.weight', 'mert_model.encoder.layers.22.final_layer_norm.bias', 'mert_model.encoder.layers.23.attention.k_proj.weight', 'mert_model.encoder.layers.23.attention.k_proj.bias', 'mert_model.encoder.layers.23.attention.v_proj.weight', 'mert_model.encoder.layers.23.attention.v_proj.bias', 'mert_model.encoder.layers.23.attention.q_proj.weight', 'mert_model.encoder.layers.23.attention.q_proj.bias', 'mert_model.encoder.layers.23.attention.out_proj.weight', 'mert_model.encoder.layers.23.attention.out_proj.bias', 'mert_model.encoder.layers.23.layer_norm.weight', 'mert_model.encoder.layers.23.layer_norm.bias', 'mert_model.encoder.layers.23.feed_forward.intermediate_dense.weight', 'mert_model.encoder.layers.23.feed_forward.intermediate_dense.bias', 'mert_model.encoder.layers.23.feed_forward.output_dense.weight', 'mert_model.encoder.layers.23.feed_forward.output_dense.bias', 'mert_model.encoder.layers.23.final_layer_norm.weight', 'mert_model.encoder.layers.23.final_layer_norm.bias', 'hubert_model.masked_spec_embed', 'hubert_model.feature_extractor.conv_layers.0.conv.weight', 'hubert_model.feature_extractor.conv_layers.0.layer_norm.weight', 'hubert_model.feature_extractor.conv_layers.0.layer_norm.bias', 'hubert_model.feature_extractor.conv_layers.1.conv.weight', 'hubert_model.feature_extractor.conv_layers.2.conv.weight', 'hubert_model.feature_extractor.conv_layers.3.conv.weight', 'hubert_model.feature_extractor.conv_layers.4.conv.weight', 'hubert_model.feature_extractor.conv_layers.5.conv.weight', 'hubert_model.feature_extractor.conv_layers.6.conv.weight', 'hubert_model.feature_projection.layer_norm.weight', 'hubert_model.feature_projection.layer_norm.bias', 'hubert_model.feature_projection.projection.weight', 'hubert_model.feature_projection.projection.bias', 'hubert_model.encoder.pos_conv_embed.conv.bias', 'hubert_model.encoder.pos_conv_embed.conv.parametrizations.weight.original0', 'hubert_model.encoder.pos_conv_embed.conv.parametrizations.weight.original1', 'hubert_model.encoder.layer_norm.weight', 'hubert_model.encoder.layer_norm.bias', 'hubert_model.encoder.layers.0.attention.k_proj.weight', 'hubert_model.encoder.layers.0.attention.k_proj.bias', 'hubert_model.encoder.layers.0.attention.v_proj.weight', 'hubert_model.encoder.layers.0.attention.v_proj.bias', 'hubert_model.encoder.layers.0.attention.q_proj.weight', 'hubert_model.encoder.layers.0.attention.q_proj.bias', 'hubert_model.encoder.layers.0.attention.out_proj.weight', 'hubert_model.encoder.layers.0.attention.out_proj.bias', 'hubert_model.encoder.layers.0.layer_norm.weight', 'hubert_model.encoder.layers.0.layer_norm.bias', 'hubert_model.encoder.layers.0.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.0.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.0.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.0.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.0.final_layer_norm.weight', 'hubert_model.encoder.layers.0.final_layer_norm.bias', 'hubert_model.encoder.layers.1.attention.k_proj.weight', 'hubert_model.encoder.layers.1.attention.k_proj.bias', 'hubert_model.encoder.layers.1.attention.v_proj.weight', 'hubert_model.encoder.layers.1.attention.v_proj.bias', 'hubert_model.encoder.layers.1.attention.q_proj.weight', 'hubert_model.encoder.layers.1.attention.q_proj.bias', 'hubert_model.encoder.layers.1.attention.out_proj.weight', 'hubert_model.encoder.layers.1.attention.out_proj.bias', 'hubert_model.encoder.layers.1.layer_norm.weight', 'hubert_model.encoder.layers.1.layer_norm.bias', 'hubert_model.encoder.layers.1.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.1.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.1.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.1.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.1.final_layer_norm.weight', 'hubert_model.encoder.layers.1.final_layer_norm.bias', 'hubert_model.encoder.layers.2.attention.k_proj.weight', 'hubert_model.encoder.layers.2.attention.k_proj.bias', 'hubert_model.encoder.layers.2.attention.v_proj.weight', 'hubert_model.encoder.layers.2.attention.v_proj.bias', 'hubert_model.encoder.layers.2.attention.q_proj.weight', 'hubert_model.encoder.layers.2.attention.q_proj.bias', 'hubert_model.encoder.layers.2.attention.out_proj.weight', 'hubert_model.encoder.layers.2.attention.out_proj.bias', 'hubert_model.encoder.layers.2.layer_norm.weight', 'hubert_model.encoder.layers.2.layer_norm.bias', 'hubert_model.encoder.layers.2.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.2.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.2.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.2.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.2.final_layer_norm.weight', 'hubert_model.encoder.layers.2.final_layer_norm.bias', 'hubert_model.encoder.layers.3.attention.k_proj.weight', 'hubert_model.encoder.layers.3.attention.k_proj.bias', 'hubert_model.encoder.layers.3.attention.v_proj.weight', 'hubert_model.encoder.layers.3.attention.v_proj.bias', 'hubert_model.encoder.layers.3.attention.q_proj.weight', 'hubert_model.encoder.layers.3.attention.q_proj.bias', 'hubert_model.encoder.layers.3.attention.out_proj.weight', 'hubert_model.encoder.layers.3.attention.out_proj.bias', 'hubert_model.encoder.layers.3.layer_norm.weight', 'hubert_model.encoder.layers.3.layer_norm.bias', 'hubert_model.encoder.layers.3.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.3.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.3.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.3.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.3.final_layer_norm.weight', 'hubert_model.encoder.layers.3.final_layer_norm.bias', 'hubert_model.encoder.layers.4.attention.k_proj.weight', 'hubert_model.encoder.layers.4.attention.k_proj.bias', 'hubert_model.encoder.layers.4.attention.v_proj.weight', 'hubert_model.encoder.layers.4.attention.v_proj.bias', 'hubert_model.encoder.layers.4.attention.q_proj.weight', 'hubert_model.encoder.layers.4.attention.q_proj.bias', 'hubert_model.encoder.layers.4.attention.out_proj.weight', 'hubert_model.encoder.layers.4.attention.out_proj.bias', 'hubert_model.encoder.layers.4.layer_norm.weight', 'hubert_model.encoder.layers.4.layer_norm.bias', 'hubert_model.encoder.layers.4.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.4.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.4.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.4.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.4.final_layer_norm.weight', 'hubert_model.encoder.layers.4.final_layer_norm.bias', 'hubert_model.encoder.layers.5.attention.k_proj.weight', 'hubert_model.encoder.layers.5.attention.k_proj.bias', 'hubert_model.encoder.layers.5.attention.v_proj.weight', 'hubert_model.encoder.layers.5.attention.v_proj.bias', 'hubert_model.encoder.layers.5.attention.q_proj.weight', 'hubert_model.encoder.layers.5.attention.q_proj.bias', 'hubert_model.encoder.layers.5.attention.out_proj.weight', 'hubert_model.encoder.layers.5.attention.out_proj.bias', 'hubert_model.encoder.layers.5.layer_norm.weight', 'hubert_model.encoder.layers.5.layer_norm.bias', 'hubert_model.encoder.layers.5.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.5.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.5.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.5.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.5.final_layer_norm.weight', 'hubert_model.encoder.layers.5.final_layer_norm.bias', 'hubert_model.encoder.layers.6.attention.k_proj.weight', 'hubert_model.encoder.layers.6.attention.k_proj.bias', 'hubert_model.encoder.layers.6.attention.v_proj.weight', 'hubert_model.encoder.layers.6.attention.v_proj.bias', 'hubert_model.encoder.layers.6.attention.q_proj.weight', 'hubert_model.encoder.layers.6.attention.q_proj.bias', 'hubert_model.encoder.layers.6.attention.out_proj.weight', 'hubert_model.encoder.layers.6.attention.out_proj.bias', 'hubert_model.encoder.layers.6.layer_norm.weight', 'hubert_model.encoder.layers.6.layer_norm.bias', 'hubert_model.encoder.layers.6.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.6.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.6.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.6.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.6.final_layer_norm.weight', 'hubert_model.encoder.layers.6.final_layer_norm.bias', 'hubert_model.encoder.layers.7.attention.k_proj.weight', 'hubert_model.encoder.layers.7.attention.k_proj.bias', 'hubert_model.encoder.layers.7.attention.v_proj.weight', 'hubert_model.encoder.layers.7.attention.v_proj.bias', 'hubert_model.encoder.layers.7.attention.q_proj.weight', 'hubert_model.encoder.layers.7.attention.q_proj.bias', 'hubert_model.encoder.layers.7.attention.out_proj.weight', 'hubert_model.encoder.layers.7.attention.out_proj.bias', 'hubert_model.encoder.layers.7.layer_norm.weight', 'hubert_model.encoder.layers.7.layer_norm.bias', 'hubert_model.encoder.layers.7.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.7.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.7.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.7.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.7.final_layer_norm.weight', 'hubert_model.encoder.layers.7.final_layer_norm.bias', 'hubert_model.encoder.layers.8.attention.k_proj.weight', 'hubert_model.encoder.layers.8.attention.k_proj.bias', 'hubert_model.encoder.layers.8.attention.v_proj.weight', 'hubert_model.encoder.layers.8.attention.v_proj.bias', 'hubert_model.encoder.layers.8.attention.q_proj.weight', 'hubert_model.encoder.layers.8.attention.q_proj.bias', 'hubert_model.encoder.layers.8.attention.out_proj.weight', 'hubert_model.encoder.layers.8.attention.out_proj.bias', 'hubert_model.encoder.layers.8.layer_norm.weight', 'hubert_model.encoder.layers.8.layer_norm.bias', 'hubert_model.encoder.layers.8.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.8.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.8.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.8.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.8.final_layer_norm.weight', 'hubert_model.encoder.layers.8.final_layer_norm.bias', 'hubert_model.encoder.layers.9.attention.k_proj.weight', 'hubert_model.encoder.layers.9.attention.k_proj.bias', 'hubert_model.encoder.layers.9.attention.v_proj.weight', 'hubert_model.encoder.layers.9.attention.v_proj.bias', 'hubert_model.encoder.layers.9.attention.q_proj.weight', 'hubert_model.encoder.layers.9.attention.q_proj.bias', 'hubert_model.encoder.layers.9.attention.out_proj.weight', 'hubert_model.encoder.layers.9.attention.out_proj.bias', 'hubert_model.encoder.layers.9.layer_norm.weight', 'hubert_model.encoder.layers.9.layer_norm.bias', 'hubert_model.encoder.layers.9.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.9.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.9.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.9.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.9.final_layer_norm.weight', 'hubert_model.encoder.layers.9.final_layer_norm.bias', 'hubert_model.encoder.layers.10.attention.k_proj.weight', 'hubert_model.encoder.layers.10.attention.k_proj.bias', 'hubert_model.encoder.layers.10.attention.v_proj.weight', 'hubert_model.encoder.layers.10.attention.v_proj.bias', 'hubert_model.encoder.layers.10.attention.q_proj.weight', 'hubert_model.encoder.layers.10.attention.q_proj.bias', 'hubert_model.encoder.layers.10.attention.out_proj.weight', 'hubert_model.encoder.layers.10.attention.out_proj.bias', 'hubert_model.encoder.layers.10.layer_norm.weight', 'hubert_model.encoder.layers.10.layer_norm.bias', 'hubert_model.encoder.layers.10.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.10.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.10.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.10.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.10.final_layer_norm.weight', 'hubert_model.encoder.layers.10.final_layer_norm.bias', 'hubert_model.encoder.layers.11.attention.k_proj.weight', 'hubert_model.encoder.layers.11.attention.k_proj.bias', 'hubert_model.encoder.layers.11.attention.v_proj.weight', 'hubert_model.encoder.layers.11.attention.v_proj.bias', 'hubert_model.encoder.layers.11.attention.q_proj.weight', 'hubert_model.encoder.layers.11.attention.q_proj.bias', 'hubert_model.encoder.layers.11.attention.out_proj.weight', 'hubert_model.encoder.layers.11.attention.out_proj.bias', 'hubert_model.encoder.layers.11.layer_norm.weight', 'hubert_model.encoder.layers.11.layer_norm.bias', 'hubert_model.encoder.layers.11.feed_forward.intermediate_dense.weight', 'hubert_model.encoder.layers.11.feed_forward.intermediate_dense.bias', 'hubert_model.encoder.layers.11.feed_forward.output_dense.weight', 'hubert_model.encoder.layers.11.feed_forward.output_dense.bias', 'hubert_model.encoder.layers.11.final_layer_norm.weight', 'hubert_model.encoder.layers.11.final_layer_norm.bias', 'vae.encoder.conv_in.weight', 'vae.encoder.conv_in.bias', 'vae.encoder.down_blocks.0.0.conv1.weight', 'vae.encoder.down_blocks.0.0.conv1.bias', 'vae.encoder.down_blocks.0.0.conv2.weight', 'vae.encoder.down_blocks.0.0.norm.weight', 'vae.encoder.down_blocks.0.0.norm.bias', 'vae.encoder.down_blocks.0.1.conv1.weight', 'vae.encoder.down_blocks.0.1.conv1.bias', 'vae.encoder.down_blocks.0.1.conv2.weight', 'vae.encoder.down_blocks.0.1.norm.weight', 'vae.encoder.down_blocks.0.1.norm.bias', 'vae.encoder.down_blocks.0.2.conv.weight', 'vae.encoder.down_blocks.0.2.conv.bias', 'vae.encoder.down_blocks.1.0.conv1.weight', 'vae.encoder.down_blocks.1.0.conv1.bias', 'vae.encoder.down_blocks.1.0.conv2.weight', 'vae.encoder.down_blocks.1.0.norm.weight', 'vae.encoder.down_blocks.1.0.norm.bias', 'vae.encoder.down_blocks.1.1.conv1.weight', 'vae.encoder.down_blocks.1.1.conv1.bias', 'vae.encoder.down_blocks.1.1.conv2.weight', 'vae.encoder.down_blocks.1.1.norm.weight', 'vae.encoder.down_blocks.1.1.norm.bias', 'vae.encoder.down_blocks.1.2.conv.weight', 'vae.encoder.down_blocks.1.2.conv.bias', 'vae.encoder.down_blocks.2.0.conv1.weight', 'vae.encoder.down_blocks.2.0.conv1.bias', 'vae.encoder.down_blocks.2.0.conv2.weight', 'vae.encoder.down_blocks.2.0.norm.weight', 'vae.encoder.down_blocks.2.0.norm.bias', 'vae.encoder.down_blocks.2.1.conv1.weight', 'vae.encoder.down_blocks.2.1.conv1.bias', 'vae.encoder.down_blocks.2.1.conv2.weight', 'vae.encoder.down_blocks.2.1.norm.weight', 'vae.encoder.down_blocks.2.1.norm.bias', 'vae.encoder.down_blocks.2.2.conv1.weight', 'vae.encoder.down_blocks.2.2.conv1.bias', 'vae.encoder.down_blocks.2.2.conv2.weight', 'vae.encoder.down_blocks.2.2.norm.weight', 'vae.encoder.down_blocks.2.2.norm.bias', 'vae.encoder.down_blocks.2.3.conv.weight', 'vae.encoder.down_blocks.2.3.conv.bias', 'vae.encoder.down_blocks.3.0.attn.to_q.weight', 'vae.encoder.down_blocks.3.0.attn.to_k.weight', 'vae.encoder.down_blocks.3.0.attn.to_v.weight', 'vae.encoder.down_blocks.3.0.attn.to_qkv_multiscale.0.proj_in.weight', 'vae.encoder.down_blocks.3.0.attn.to_qkv_multiscale.0.proj_out.weight', 'vae.encoder.down_blocks.3.0.attn.to_out.weight', 'vae.encoder.down_blocks.3.0.attn.norm_out.weight', 'vae.encoder.down_blocks.3.0.attn.norm_out.bias', 'vae.encoder.down_blocks.3.0.conv_out.conv_inverted.weight', 'vae.encoder.down_blocks.3.0.conv_out.conv_inverted.bias', 'vae.encoder.down_blocks.3.0.conv_out.conv_depth.weight', 'vae.encoder.down_blocks.3.0.conv_out.conv_depth.bias', 'vae.encoder.down_blocks.3.0.conv_out.conv_point.weight', 'vae.encoder.down_blocks.3.0.conv_out.norm.weight', 'vae.encoder.down_blocks.3.0.conv_out.norm.bias', 'vae.encoder.down_blocks.3.1.attn.to_q.weight', 'vae.encoder.down_blocks.3.1.attn.to_k.weight', 'vae.encoder.down_blocks.3.1.attn.to_v.weight', 'vae.encoder.down_blocks.3.1.attn.to_qkv_multiscale.0.proj_in.weight', 'vae.encoder.down_blocks.3.1.attn.to_qkv_multiscale.0.proj_out.weight', 'vae.encoder.down_blocks.3.1.attn.to_out.weight', 'vae.encoder.down_blocks.3.1.attn.norm_out.weight', 'vae.encoder.down_blocks.3.1.attn.norm_out.bias', 'vae.encoder.down_blocks.3.1.conv_out.conv_inverted.weight', 'vae.encoder.down_blocks.3.1.conv_out.conv_inverted.bias', 'vae.encoder.down_blocks.3.1.conv_out.conv_depth.weight', 'vae.encoder.down_blocks.3.1.conv_out.conv_depth.bias', 'vae.encoder.down_blocks.3.1.conv_out.conv_point.weight', 'vae.encoder.down_blocks.3.1.conv_out.norm.weight', 'vae.encoder.down_blocks.3.1.conv_out.norm.bias', 'vae.encoder.down_blocks.3.2.attn.to_q.weight', 'vae.encoder.down_blocks.3.2.attn.to_k.weight', 'vae.encoder.down_blocks.3.2.attn.to_v.weight', 'vae.encoder.down_blocks.3.2.attn.to_qkv_multiscale.0.proj_in.weight', 'vae.encoder.down_blocks.3.2.attn.to_qkv_multiscale.0.proj_out.weight', 'vae.encoder.down_blocks.3.2.attn.to_out.weight', 'vae.encoder.down_blocks.3.2.attn.norm_out.weight', 'vae.encoder.down_blocks.3.2.attn.norm_out.bias', 'vae.encoder.down_blocks.3.2.conv_out.conv_inverted.weight', 'vae.encoder.down_blocks.3.2.conv_out.conv_inverted.bias', 'vae.encoder.down_blocks.3.2.conv_out.conv_depth.weight', 'vae.encoder.down_blocks.3.2.conv_out.conv_depth.bias', 'vae.encoder.down_blocks.3.2.conv_out.conv_point.weight', 'vae.encoder.down_blocks.3.2.conv_out.norm.weight', 'vae.encoder.down_blocks.3.2.conv_out.norm.bias', 'vae.encoder.conv_out.weight', 'vae.encoder.conv_out.bias', 'vae.decoder.conv_in.weight', 'vae.decoder.conv_in.bias', 'vae.decoder.up_blocks.0.0.conv.weight', 'vae.decoder.up_blocks.0.0.conv.bias', 'vae.decoder.up_blocks.0.1.conv1.weight', 'vae.decoder.up_blocks.0.1.conv1.bias', 'vae.decoder.up_blocks.0.1.conv2.weight', 'vae.decoder.up_blocks.0.1.norm.weight', 'vae.decoder.up_blocks.0.1.norm.bias', 'vae.decoder.up_blocks.0.2.conv1.weight', 'vae.decoder.up_blocks.0.2.conv1.bias', 'vae.decoder.up_blocks.0.2.conv2.weight', 'vae.decoder.up_blocks.0.2.norm.weight', 'vae.decoder.up_blocks.0.2.norm.bias', 'vae.decoder.up_blocks.0.3.conv1.weight', 'vae.decoder.up_blocks.0.3.conv1.bias', 'vae.decoder.up_blocks.0.3.conv2.weight', 'vae.decoder.up_blocks.0.3.norm.weight', 'vae.decoder.up_blocks.0.3.norm.bias', 'vae.decoder.up_blocks.1.0.conv.weight', 'vae.decoder.up_blocks.1.0.conv.bias', 'vae.decoder.up_blocks.1.1.conv1.weight', 'vae.decoder.up_blocks.1.1.conv1.bias', 'vae.decoder.up_blocks.1.1.conv2.weight', 'vae.decoder.up_blocks.1.1.norm.weight', 'vae.decoder.up_blocks.1.1.norm.bias', 'vae.decoder.up_blocks.1.2.conv1.weight', 'vae.decoder.up_blocks.1.2.conv1.bias', 'vae.decoder.up_blocks.1.2.conv2.weight', 'vae.decoder.up_blocks.1.2.norm.weight', 'vae.decoder.up_blocks.1.2.norm.bias', 'vae.decoder.up_blocks.1.3.conv1.weight', 'vae.decoder.up_blocks.1.3.conv1.bias', 'vae.decoder.up_blocks.1.3.conv2.weight', 'vae.decoder.up_blocks.1.3.norm.weight', 'vae.decoder.up_blocks.1.3.norm.bias', 'vae.decoder.up_blocks.2.0.conv.weight', 'vae.decoder.up_blocks.2.0.conv.bias', 'vae.decoder.up_blocks.2.1.conv1.weight', 'vae.decoder.up_blocks.2.1.conv1.bias', 'vae.decoder.up_blocks.2.1.conv2.weight', 'vae.decoder.up_blocks.2.1.norm.weight', 'vae.decoder.up_blocks.2.1.norm.bias', 'vae.decoder.up_blocks.2.2.conv1.weight', 'vae.decoder.up_blocks.2.2.conv1.bias', 'vae.decoder.up_blocks.2.2.conv2.weight', 'vae.decoder.up_blocks.2.2.norm.weight', 'vae.decoder.up_blocks.2.2.norm.bias', 'vae.decoder.up_blocks.2.3.conv1.weight', 'vae.decoder.up_blocks.2.3.conv1.bias', 'vae.decoder.up_blocks.2.3.conv2.weight', 'vae.decoder.up_blocks.2.3.norm.weight', 'vae.decoder.up_blocks.2.3.norm.bias', 'vae.decoder.up_blocks.3.0.attn.to_q.weight', 'vae.decoder.up_blocks.3.0.attn.to_k.weight', 'vae.decoder.up_blocks.3.0.attn.to_v.weight', 'vae.decoder.up_blocks.3.0.attn.to_qkv_multiscale.0.proj_in.weight', 'vae.decoder.up_blocks.3.0.attn.to_qkv_multiscale.0.proj_out.weight', 'vae.decoder.up_blocks.3.0.attn.to_out.weight', 'vae.decoder.up_blocks.3.0.attn.norm_out.weight', 'vae.decoder.up_blocks.3.0.attn.norm_out.bias', 'vae.decoder.up_blocks.3.0.conv_out.conv_inverted.weight', 'vae.decoder.up_blocks.3.0.conv_out.conv_inverted.bias', 'vae.decoder.up_blocks.3.0.conv_out.conv_depth.weight', 'vae.decoder.up_blocks.3.0.conv_out.conv_depth.bias', 'vae.decoder.up_blocks.3.0.conv_out.conv_point.weight', 'vae.decoder.up_blocks.3.0.conv_out.norm.weight', 'vae.decoder.up_blocks.3.0.conv_out.norm.bias', 'vae.decoder.up_blocks.3.1.attn.to_q.weight', 'vae.decoder.up_blocks.3.1.attn.to_k.weight', 'vae.decoder.up_blocks.3.1.attn.to_v.weight', 'vae.decoder.up_blocks.3.1.attn.to_qkv_multiscale.0.proj_in.weight', 'vae.decoder.up_blocks.3.1.attn.to_qkv_multiscale.0.proj_out.weight', 'vae.decoder.up_blocks.3.1.attn.to_out.weight', 'vae.decoder.up_blocks.3.1.attn.norm_out.weight', 'vae.decoder.up_blocks.3.1.attn.norm_out.bias', 'vae.decoder.up_blocks.3.1.conv_out.conv_inverted.weight', 'vae.decoder.up_blocks.3.1.conv_out.conv_inverted.bias', 'vae.decoder.up_blocks.3.1.conv_out.conv_depth.weight', 'vae.decoder.up_blocks.3.1.conv_out.conv_depth.bias', 'vae.decoder.up_blocks.3.1.conv_out.conv_point.weight', 'vae.decoder.up_blocks.3.1.conv_out.norm.weight', 'vae.decoder.up_blocks.3.1.conv_out.norm.bias', 'vae.decoder.up_blocks.3.2.attn.to_q.weight', 'vae.decoder.up_blocks.3.2.attn.to_k.weight', 'vae.decoder.up_blocks.3.2.attn.to_v.weight', 'vae.decoder.up_blocks.3.2.attn.to_qkv_multiscale.0.proj_in.weight', 'vae.decoder.up_blocks.3.2.attn.to_qkv_multiscale.0.proj_out.weight', 'vae.decoder.up_blocks.3.2.attn.to_out.weight', 'vae.decoder.up_blocks.3.2.attn.norm_out.weight', 'vae.decoder.up_blocks.3.2.attn.norm_out.bias', 'vae.decoder.up_blocks.3.2.conv_out.conv_inverted.weight', 'vae.decoder.up_blocks.3.2.conv_out.conv_inverted.bias', 'vae.decoder.up_blocks.3.2.conv_out.conv_depth.weight', 'vae.decoder.up_blocks.3.2.conv_out.conv_depth.bias', 'vae.decoder.up_blocks.3.2.conv_out.conv_point.weight', 'vae.decoder.up_blocks.3.2.conv_out.norm.weight', 'vae.decoder.up_blocks.3.2.conv_out.norm.bias', 'vae.decoder.norm_out.weight', 'vae.decoder.norm_out.bias', 'vae.decoder.conv_out.weight', 'vae.decoder.conv_out.bias', 'vae.vocoder.backbone.channel_layers.0.0.weight', 'vae.vocoder.backbone.channel_layers.0.0.bias', 'vae.vocoder.backbone.channel_layers.0.1.weight', 'vae.vocoder.backbone.channel_layers.0.1.bias', 'vae.vocoder.backbone.channel_layers.1.0.weight', 'vae.vocoder.backbone.channel_layers.1.0.bias', 'vae.vocoder.backbone.channel_layers.1.1.weight', 'vae.vocoder.backbone.channel_layers.1.1.bias', 'vae.vocoder.backbone.channel_layers.2.0.weight', 'vae.vocoder.backbone.channel_layers.2.0.bias', 'vae.vocoder.backbone.channel_layers.2.1.weight', 'vae.vocoder.backbone.channel_layers.2.1.bias', 'vae.vocoder.backbone.channel_layers.3.0.weight', 'vae.vocoder.backbone.channel_layers.3.0.bias', 'vae.vocoder.backbone.channel_layers.3.1.weight', 'vae.vocoder.backbone.channel_layers.3.1.bias', 'vae.vocoder.backbone.stages.0.0.gamma', 'vae.vocoder.backbone.stages.0.0.dwconv.weight', 'vae.vocoder.backbone.stages.0.0.dwconv.bias', 'vae.vocoder.backbone.stages.0.0.norm.weight', 'vae.vocoder.backbone.stages.0.0.norm.bias', 'vae.vocoder.backbone.stages.0.0.pwconv1.weight', 'vae.vocoder.backbone.stages.0.0.pwconv1.bias', 'vae.vocoder.backbone.stages.0.0.pwconv2.weight', 'vae.vocoder.backbone.stages.0.0.pwconv2.bias', 'vae.vocoder.backbone.stages.0.1.gamma', 'vae.vocoder.backbone.stages.0.1.dwconv.weight', 'vae.vocoder.backbone.stages.0.1.dwconv.bias', 'vae.vocoder.backbone.stages.0.1.norm.weight', 'vae.vocoder.backbone.stages.0.1.norm.bias', 'vae.vocoder.backbone.stages.0.1.pwconv1.weight', 'vae.vocoder.backbone.stages.0.1.pwconv1.bias', 'vae.vocoder.backbone.stages.0.1.pwconv2.weight', 'vae.vocoder.backbone.stages.0.1.pwconv2.bias', 'vae.vocoder.backbone.stages.0.2.gamma', 'vae.vocoder.backbone.stages.0.2.dwconv.weight', 'vae.vocoder.backbone.stages.0.2.dwconv.bias', 'vae.vocoder.backbone.stages.0.2.norm.weight', 'vae.vocoder.backbone.stages.0.2.norm.bias', 'vae.vocoder.backbone.stages.0.2.pwconv1.weight', 'vae.vocoder.backbone.stages.0.2.pwconv1.bias', 'vae.vocoder.backbone.stages.0.2.pwconv2.weight', 'vae.vocoder.backbone.stages.0.2.pwconv2.bias', 'vae.vocoder.backbone.stages.1.0.gamma', 'vae.vocoder.backbone.stages.1.0.dwconv.weight', 'vae.vocoder.backbone.stages.1.0.dwconv.bias', 'vae.vocoder.backbone.stages.1.0.norm.weight', 'vae.vocoder.backbone.stages.1.0.norm.bias', 'vae.vocoder.backbone.stages.1.0.pwconv1.weight', 'vae.vocoder.backbone.stages.1.0.pwconv1.bias', 'vae.vocoder.backbone.stages.1.0.pwconv2.weight', 'vae.vocoder.backbone.stages.1.0.pwconv2.bias', 'vae.vocoder.backbone.stages.1.1.gamma', 'vae.vocoder.backbone.stages.1.1.dwconv.weight', 'vae.vocoder.backbone.stages.1.1.dwconv.bias', 'vae.vocoder.backbone.stages.1.1.norm.weight', 'vae.vocoder.backbone.stages.1.1.norm.bias', 'vae.vocoder.backbone.stages.1.1.pwconv1.weight', 'vae.vocoder.backbone.stages.1.1.pwconv1.bias', 'vae.vocoder.backbone.stages.1.1.pwconv2.weight', 'vae.vocoder.backbone.stages.1.1.pwconv2.bias', 'vae.vocoder.backbone.stages.1.2.gamma', 'vae.vocoder.backbone.stages.1.2.dwconv.weight', 'vae.vocoder.backbone.stages.1.2.dwconv.bias', 'vae.vocoder.backbone.stages.1.2.norm.weight', 'vae.vocoder.backbone.stages.1.2.norm.bias', 'vae.vocoder.backbone.stages.1.2.pwconv1.weight', 'vae.vocoder.backbone.stages.1.2.pwconv1.bias', 'vae.vocoder.backbone.stages.1.2.pwconv2.weight', 'vae.vocoder.backbone.stages.1.2.pwconv2.bias', 'vae.vocoder.backbone.stages.2.0.gamma', 'vae.vocoder.backbone.stages.2.0.dwconv.weight', 'vae.vocoder.backbone.stages.2.0.dwconv.bias', 'vae.vocoder.backbone.stages.2.0.norm.weight', 'vae.vocoder.backbone.stages.2.0.norm.bias', 'vae.vocoder.backbone.stages.2.0.pwconv1.weight', 'vae.vocoder.backbone.stages.2.0.pwconv1.bias', 'vae.vocoder.backbone.stages.2.0.pwconv2.weight', 'vae.vocoder.backbone.stages.2.0.pwconv2.bias', 'vae.vocoder.backbone.stages.2.1.gamma', 'vae.vocoder.backbone.stages.2.1.dwconv.weight', 'vae.vocoder.backbone.stages.2.1.dwconv.bias', 'vae.vocoder.backbone.stages.2.1.norm.weight', 'vae.vocoder.backbone.stages.2.1.norm.bias', 'vae.vocoder.backbone.stages.2.1.pwconv1.weight', 'vae.vocoder.backbone.stages.2.1.pwconv1.bias', 'vae.vocoder.backbone.stages.2.1.pwconv2.weight', 'vae.vocoder.backbone.stages.2.1.pwconv2.bias', 'vae.vocoder.backbone.stages.2.2.gamma', 'vae.vocoder.backbone.stages.2.2.dwconv.weight', 'vae.vocoder.backbone.stages.2.2.dwconv.bias', 'vae.vocoder.backbone.stages.2.2.norm.weight', 'vae.vocoder.backbone.stages.2.2.norm.bias', 'vae.vocoder.backbone.stages.2.2.pwconv1.weight', 'vae.vocoder.backbone.stages.2.2.pwconv1.bias', 'vae.vocoder.backbone.stages.2.2.pwconv2.weight', 'vae.vocoder.backbone.stages.2.2.pwconv2.bias', 'vae.vocoder.backbone.stages.2.3.gamma', 'vae.vocoder.backbone.stages.2.3.dwconv.weight', 'vae.vocoder.backbone.stages.2.3.dwconv.bias', 'vae.vocoder.backbone.stages.2.3.norm.weight', 'vae.vocoder.backbone.stages.2.3.norm.bias', 'vae.vocoder.backbone.stages.2.3.pwconv1.weight', 'vae.vocoder.backbone.stages.2.3.pwconv1.bias', 'vae.vocoder.backbone.stages.2.3.pwconv2.weight', 'vae.vocoder.backbone.stages.2.3.pwconv2.bias', 'vae.vocoder.backbone.stages.2.4.gamma', 'vae.vocoder.backbone.stages.2.4.dwconv.weight', 'vae.vocoder.backbone.stages.2.4.dwconv.bias', 'vae.vocoder.backbone.stages.2.4.norm.weight', 'vae.vocoder.backbone.stages.2.4.norm.bias', 'vae.vocoder.backbone.stages.2.4.pwconv1.weight', 'vae.vocoder.backbone.stages.2.4.pwconv1.bias', 'vae.vocoder.backbone.stages.2.4.pwconv2.weight', 'vae.vocoder.backbone.stages.2.4.pwconv2.bias', 'vae.vocoder.backbone.stages.2.5.gamma', 'vae.vocoder.backbone.stages.2.5.dwconv.weight', 'vae.vocoder.backbone.stages.2.5.dwconv.bias', 'vae.vocoder.backbone.stages.2.5.norm.weight', 'vae.vocoder.backbone.stages.2.5.norm.bias', 'vae.vocoder.backbone.stages.2.5.pwconv1.weight', 'vae.vocoder.backbone.stages.2.5.pwconv1.bias', 'vae.vocoder.backbone.stages.2.5.pwconv2.weight', 'vae.vocoder.backbone.stages.2.5.pwconv2.bias', 'vae.vocoder.backbone.stages.2.6.gamma', 'vae.vocoder.backbone.stages.2.6.dwconv.weight', 'vae.vocoder.backbone.stages.2.6.dwconv.bias', 'vae.vocoder.backbone.stages.2.6.norm.weight', 'vae.vocoder.backbone.stages.2.6.norm.bias', 'vae.vocoder.backbone.stages.2.6.pwconv1.weight', 'vae.vocoder.backbone.stages.2.6.pwconv1.bias', 'vae.vocoder.backbone.stages.2.6.pwconv2.weight', 'vae.vocoder.backbone.stages.2.6.pwconv2.bias', 'vae.vocoder.backbone.stages.2.7.gamma', 'vae.vocoder.backbone.stages.2.7.dwconv.weight', 'vae.vocoder.backbone.stages.2.7.dwconv.bias', 'vae.vocoder.backbone.stages.2.7.norm.weight', 'vae.vocoder.backbone.stages.2.7.norm.bias', 'vae.vocoder.backbone.stages.2.7.pwconv1.weight', 'vae.vocoder.backbone.stages.2.7.pwconv1.bias', 'vae.vocoder.backbone.stages.2.7.pwconv2.weight', 'vae.vocoder.backbone.stages.2.7.pwconv2.bias', 'vae.vocoder.backbone.stages.2.8.gamma', 'vae.vocoder.backbone.stages.2.8.dwconv.weight', 'vae.vocoder.backbone.stages.2.8.dwconv.bias', 'vae.vocoder.backbone.stages.2.8.norm.weight', 'vae.vocoder.backbone.stages.2.8.norm.bias', 'vae.vocoder.backbone.stages.2.8.pwconv1.weight', 'vae.vocoder.backbone.stages.2.8.pwconv1.bias', 'vae.vocoder.backbone.stages.2.8.pwconv2.weight', 'vae.vocoder.backbone.stages.2.8.pwconv2.bias', 'vae.vocoder.backbone.stages.3.0.gamma', 'vae.vocoder.backbone.stages.3.0.dwconv.weight', 'vae.vocoder.backbone.stages.3.0.dwconv.bias', 'vae.vocoder.backbone.stages.3.0.norm.weight', 'vae.vocoder.backbone.stages.3.0.norm.bias', 'vae.vocoder.backbone.stages.3.0.pwconv1.weight', 'vae.vocoder.backbone.stages.3.0.pwconv1.bias', 'vae.vocoder.backbone.stages.3.0.pwconv2.weight', 'vae.vocoder.backbone.stages.3.0.pwconv2.bias', 'vae.vocoder.backbone.stages.3.1.gamma', 'vae.vocoder.backbone.stages.3.1.dwconv.weight', 'vae.vocoder.backbone.stages.3.1.dwconv.bias', 'vae.vocoder.backbone.stages.3.1.norm.weight', 'vae.vocoder.backbone.stages.3.1.norm.bias', 'vae.vocoder.backbone.stages.3.1.pwconv1.weight', 'vae.vocoder.backbone.stages.3.1.pwconv1.bias', 'vae.vocoder.backbone.stages.3.1.pwconv2.weight', 'vae.vocoder.backbone.stages.3.1.pwconv2.bias', 'vae.vocoder.backbone.stages.3.2.gamma', 'vae.vocoder.backbone.stages.3.2.dwconv.weight', 'vae.vocoder.backbone.stages.3.2.dwconv.bias', 'vae.vocoder.backbone.stages.3.2.norm.weight', 'vae.vocoder.backbone.stages.3.2.norm.bias', 'vae.vocoder.backbone.stages.3.2.pwconv1.weight', 'vae.vocoder.backbone.stages.3.2.pwconv1.bias', 'vae.vocoder.backbone.stages.3.2.pwconv2.weight', 'vae.vocoder.backbone.stages.3.2.pwconv2.bias', 'vae.vocoder.backbone.norm.weight', 'vae.vocoder.backbone.norm.bias', 'vae.vocoder.head.conv_pre.bias', 'vae.vocoder.head.conv_pre.weight_g', 'vae.vocoder.head.conv_pre.weight_v', 'vae.vocoder.head.ups.0.bias', 'vae.vocoder.head.ups.0.weight_g', 'vae.vocoder.head.ups.0.weight_v', 'vae.vocoder.head.ups.1.bias', 'vae.vocoder.head.ups.1.weight_g', 'vae.vocoder.head.ups.1.weight_v', 'vae.vocoder.head.ups.2.bias', 'vae.vocoder.head.ups.2.weight_g', 'vae.vocoder.head.ups.2.weight_v', 'vae.vocoder.head.ups.3.bias', 'vae.vocoder.head.ups.3.weight_g', 'vae.vocoder.head.ups.3.weight_v', 'vae.vocoder.head.ups.4.bias', 'vae.vocoder.head.ups.4.weight_g', 'vae.vocoder.head.ups.4.weight_v', 'vae.vocoder.head.ups.5.bias', 'vae.vocoder.head.ups.5.weight_g', 'vae.vocoder.head.ups.5.weight_v', 'vae.vocoder.head.ups.6.bias', 'vae.vocoder.head.ups.6.weight_g', 'vae.vocoder.head.ups.6.weight_v', 'vae.vocoder.head.resblocks.0.convs1.0.bias', 'vae.vocoder.head.resblocks.0.convs1.0.weight_g', 'vae.vocoder.head.resblocks.0.convs1.0.weight_v', 'vae.vocoder.head.resblocks.0.convs1.1.bias', 'vae.vocoder.head.resblocks.0.convs1.1.weight_g', 'vae.vocoder.head.resblocks.0.convs1.1.weight_v', 'vae.vocoder.head.resblocks.0.convs1.2.bias', 'vae.vocoder.head.resblocks.0.convs1.2.weight_g', 'vae.vocoder.head.resblocks.0.convs1.2.weight_v', 'vae.vocoder.head.resblocks.0.convs2.0.bias', 'vae.vocoder.head.resblocks.0.convs2.0.weight_g', 'vae.vocoder.head.resblocks.0.convs2.0.weight_v', 'vae.vocoder.head.resblocks.0.convs2.1.bias', 'vae.vocoder.head.resblocks.0.convs2.1.weight_g', 'vae.vocoder.head.resblocks.0.convs2.1.weight_v', 'vae.vocoder.head.resblocks.0.convs2.2.bias', 'vae.vocoder.head.resblocks.0.convs2.2.weight_g', 'vae.vocoder.head.resblocks.0.convs2.2.weight_v', 'vae.vocoder.head.resblocks.1.convs1.0.bias', 'vae.vocoder.head.resblocks.1.convs1.0.weight_g', 'vae.vocoder.head.resblocks.1.convs1.0.weight_v', 'vae.vocoder.head.resblocks.1.convs1.1.bias', 'vae.vocoder.head.resblocks.1.convs1.1.weight_g', 'vae.vocoder.head.resblocks.1.convs1.1.weight_v', 'vae.vocoder.head.resblocks.1.convs1.2.bias', 'vae.vocoder.head.resblocks.1.convs1.2.weight_g', 'vae.vocoder.head.resblocks.1.convs1.2.weight_v', 'vae.vocoder.head.resblocks.1.convs2.0.bias', 'vae.vocoder.head.resblocks.1.convs2.0.weight_g', 'vae.vocoder.head.resblocks.1.convs2.0.weight_v', 'vae.vocoder.head.resblocks.1.convs2.1.bias', 'vae.vocoder.head.resblocks.1.convs2.1.weight_g', 'vae.vocoder.head.resblocks.1.convs2.1.weight_v', 'vae.vocoder.head.resblocks.1.convs2.2.bias', 'vae.vocoder.head.resblocks.1.convs2.2.weight_g', 'vae.vocoder.head.resblocks.1.convs2.2.weight_v', 'vae.vocoder.head.resblocks.2.convs1.0.bias', 'vae.vocoder.head.resblocks.2.convs1.0.weight_g', 'vae.vocoder.head.resblocks.2.convs1.0.weight_v', 'vae.vocoder.head.resblocks.2.convs1.1.bias', 'vae.vocoder.head.resblocks.2.convs1.1.weight_g', 'vae.vocoder.head.resblocks.2.convs1.1.weight_v', 'vae.vocoder.head.resblocks.2.convs1.2.bias', 'vae.vocoder.head.resblocks.2.convs1.2.weight_g', 'vae.vocoder.head.resblocks.2.convs1.2.weight_v', 'vae.vocoder.head.resblocks.2.convs2.0.bias', 'vae.vocoder.head.resblocks.2.convs2.0.weight_g', 'vae.vocoder.head.resblocks.2.convs2.0.weight_v', 'vae.vocoder.head.resblocks.2.convs2.1.bias', 'vae.vocoder.head.resblocks.2.convs2.1.weight_g', 'vae.vocoder.head.resblocks.2.convs2.1.weight_v', 'vae.vocoder.head.resblocks.2.convs2.2.bias', 'vae.vocoder.head.resblocks.2.convs2.2.weight_g', 'vae.vocoder.head.resblocks.2.convs2.2.weight_v', 'vae.vocoder.head.resblocks.3.convs1.0.bias', 'vae.vocoder.head.resblocks.3.convs1.0.weight_g', 'vae.vocoder.head.resblocks.3.convs1.0.weight_v', 'vae.vocoder.head.resblocks.3.convs1.1.bias', 'vae.vocoder.head.resblocks.3.convs1.1.weight_g', 'vae.vocoder.head.resblocks.3.convs1.1.weight_v', 'vae.vocoder.head.resblocks.3.convs1.2.bias', 'vae.vocoder.head.resblocks.3.convs1.2.weight_g', 'vae.vocoder.head.resblocks.3.convs1.2.weight_v', 'vae.vocoder.head.resblocks.3.convs2.0.bias', 'vae.vocoder.head.resblocks.3.convs2.0.weight_g', 'vae.vocoder.head.resblocks.3.convs2.0.weight_v', 'vae.vocoder.head.resblocks.3.convs2.1.bias', 'vae.vocoder.head.resblocks.3.convs2.1.weight_g', 'vae.vocoder.head.resblocks.3.convs2.1.weight_v', 'vae.vocoder.head.resblocks.3.convs2.2.bias', 'vae.vocoder.head.resblocks.3.convs2.2.weight_g', 'vae.vocoder.head.resblocks.3.convs2.2.weight_v', 'vae.vocoder.head.resblocks.4.convs1.0.bias', 'vae.vocoder.head.resblocks.4.convs1.0.weight_g', 'vae.vocoder.head.resblocks.4.convs1.0.weight_v', 'vae.vocoder.head.resblocks.4.convs1.1.bias', 'vae.vocoder.head.resblocks.4.convs1.1.weight_g', 'vae.vocoder.head.resblocks.4.convs1.1.weight_v', 'vae.vocoder.head.resblocks.4.convs1.2.bias', 'vae.vocoder.head.resblocks.4.convs1.2.weight_g', 'vae.vocoder.head.resblocks.4.convs1.2.weight_v', 'vae.vocoder.head.resblocks.4.convs2.0.bias', 'vae.vocoder.head.resblocks.4.convs2.0.weight_g', 'vae.vocoder.head.resblocks.4.convs2.0.weight_v', 'vae.vocoder.head.resblocks.4.convs2.1.bias', 'vae.vocoder.head.resblocks.4.convs2.1.weight_g', 'vae.vocoder.head.resblocks.4.convs2.1.weight_v', 'vae.vocoder.head.resblocks.4.convs2.2.bias', 'vae.vocoder.head.resblocks.4.convs2.2.weight_g', 'vae.vocoder.head.resblocks.4.convs2.2.weight_v', 'vae.vocoder.head.resblocks.5.convs1.0.bias', 'vae.vocoder.head.resblocks.5.convs1.0.weight_g', 'vae.vocoder.head.resblocks.5.convs1.0.weight_v', 'vae.vocoder.head.resblocks.5.convs1.1.bias', 'vae.vocoder.head.resblocks.5.convs1.1.weight_g', 'vae.vocoder.head.resblocks.5.convs1.1.weight_v', 'vae.vocoder.head.resblocks.5.convs1.2.bias', 'vae.vocoder.head.resblocks.5.convs1.2.weight_g', 'vae.vocoder.head.resblocks.5.convs1.2.weight_v', 'vae.vocoder.head.resblocks.5.convs2.0.bias', 'vae.vocoder.head.resblocks.5.convs2.0.weight_g', 'vae.vocoder.head.resblocks.5.convs2.0.weight_v', 'vae.vocoder.head.resblocks.5.convs2.1.bias', 'vae.vocoder.head.resblocks.5.convs2.1.weight_g', 'vae.vocoder.head.resblocks.5.convs2.1.weight_v', 'vae.vocoder.head.resblocks.5.convs2.2.bias', 'vae.vocoder.head.resblocks.5.convs2.2.weight_g', 'vae.vocoder.head.resblocks.5.convs2.2.weight_v', 'vae.vocoder.head.resblocks.6.convs1.0.bias', 'vae.vocoder.head.resblocks.6.convs1.0.weight_g', 'vae.vocoder.head.resblocks.6.convs1.0.weight_v', 'vae.vocoder.head.resblocks.6.convs1.1.bias', 'vae.vocoder.head.resblocks.6.convs1.1.weight_g', 'vae.vocoder.head.resblocks.6.convs1.1.weight_v', 'vae.vocoder.head.resblocks.6.convs1.2.bias', 'vae.vocoder.head.resblocks.6.convs1.2.weight_g', 'vae.vocoder.head.resblocks.6.convs1.2.weight_v', 'vae.vocoder.head.resblocks.6.convs2.0.bias', 'vae.vocoder.head.resblocks.6.convs2.0.weight_g', 'vae.vocoder.head.resblocks.6.convs2.0.weight_v', 'vae.vocoder.head.resblocks.6.convs2.1.bias', 'vae.vocoder.head.resblocks.6.convs2.1.weight_g', 'vae.vocoder.head.resblocks.6.convs2.1.weight_v', 'vae.vocoder.head.resblocks.6.convs2.2.bias', 'vae.vocoder.head.resblocks.6.convs2.2.weight_g', 'vae.vocoder.head.resblocks.6.convs2.2.weight_v', 'vae.vocoder.head.resblocks.7.convs1.0.bias', 'vae.vocoder.head.resblocks.7.convs1.0.weight_g', 'vae.vocoder.head.resblocks.7.convs1.0.weight_v', 'vae.vocoder.head.resblocks.7.convs1.1.bias', 'vae.vocoder.head.resblocks.7.convs1.1.weight_g', 'vae.vocoder.head.resblocks.7.convs1.1.weight_v', 'vae.vocoder.head.resblocks.7.convs1.2.bias', 'vae.vocoder.head.resblocks.7.convs1.2.weight_g', 'vae.vocoder.head.resblocks.7.convs1.2.weight_v', 'vae.vocoder.head.resblocks.7.convs2.0.bias', 'vae.vocoder.head.resblocks.7.convs2.0.weight_g', 'vae.vocoder.head.resblocks.7.convs2.0.weight_v', 'vae.vocoder.head.resblocks.7.convs2.1.bias', 'vae.vocoder.head.resblocks.7.convs2.1.weight_g', 'vae.vocoder.head.resblocks.7.convs2.1.weight_v', 'vae.vocoder.head.resblocks.7.convs2.2.bias', 'vae.vocoder.head.resblocks.7.convs2.2.weight_g', 'vae.vocoder.head.resblocks.7.convs2.2.weight_v', 'vae.vocoder.head.resblocks.8.convs1.0.bias', 'vae.vocoder.head.resblocks.8.convs1.0.weight_g', 'vae.vocoder.head.resblocks.8.convs1.0.weight_v', 'vae.vocoder.head.resblocks.8.convs1.1.bias', 'vae.vocoder.head.resblocks.8.convs1.1.weight_g', 'vae.vocoder.head.resblocks.8.convs1.1.weight_v', 'vae.vocoder.head.resblocks.8.convs1.2.bias', 'vae.vocoder.head.resblocks.8.convs1.2.weight_g', 'vae.vocoder.head.resblocks.8.convs1.2.weight_v', 'vae.vocoder.head.resblocks.8.convs2.0.bias', 'vae.vocoder.head.resblocks.8.convs2.0.weight_g', 'vae.vocoder.head.resblocks.8.convs2.0.weight_v', 'vae.vocoder.head.resblocks.8.convs2.1.bias', 'vae.vocoder.head.resblocks.8.convs2.1.weight_g', 'vae.vocoder.head.resblocks.8.convs2.1.weight_v', 'vae.vocoder.head.resblocks.8.convs2.2.bias', 'vae.vocoder.head.resblocks.8.convs2.2.weight_g', 'vae.vocoder.head.resblocks.8.convs2.2.weight_v', 'vae.vocoder.head.resblocks.9.convs1.0.bias', 'vae.vocoder.head.resblocks.9.convs1.0.weight_g', 'vae.vocoder.head.resblocks.9.convs1.0.weight_v', 'vae.vocoder.head.resblocks.9.convs1.1.bias', 'vae.vocoder.head.resblocks.9.convs1.1.weight_g', 'vae.vocoder.head.resblocks.9.convs1.1.weight_v', 'vae.vocoder.head.resblocks.9.convs1.2.bias', 'vae.vocoder.head.resblocks.9.convs1.2.weight_g', 'vae.vocoder.head.resblocks.9.convs1.2.weight_v', 'vae.vocoder.head.resblocks.9.convs2.0.bias', 'vae.vocoder.head.resblocks.9.convs2.0.weight_g', 'vae.vocoder.head.resblocks.9.convs2.0.weight_v', 'vae.vocoder.head.resblocks.9.convs2.1.bias', 'vae.vocoder.head.resblocks.9.convs2.1.weight_g', 'vae.vocoder.head.resblocks.9.convs2.1.weight_v', 'vae.vocoder.head.resblocks.9.convs2.2.bias', 'vae.vocoder.head.resblocks.9.convs2.2.weight_g', 'vae.vocoder.head.resblocks.9.convs2.2.weight_v', 'vae.vocoder.head.resblocks.10.convs1.0.bias', 'vae.vocoder.head.resblocks.10.convs1.0.weight_g', 'vae.vocoder.head.resblocks.10.convs1.0.weight_v', 'vae.vocoder.head.resblocks.10.convs1.1.bias', 'vae.vocoder.head.resblocks.10.convs1.1.weight_g', 'vae.vocoder.head.resblocks.10.convs1.1.weight_v', 'vae.vocoder.head.resblocks.10.convs1.2.bias', 'vae.vocoder.head.resblocks.10.convs1.2.weight_g', 'vae.vocoder.head.resblocks.10.convs1.2.weight_v', 'vae.vocoder.head.resblocks.10.convs2.0.bias', 'vae.vocoder.head.resblocks.10.convs2.0.weight_g', 'vae.vocoder.head.resblocks.10.convs2.0.weight_v', 'vae.vocoder.head.resblocks.10.convs2.1.bias', 'vae.vocoder.head.resblocks.10.convs2.1.weight_g', 'vae.vocoder.head.resblocks.10.convs2.1.weight_v', 'vae.vocoder.head.resblocks.10.convs2.2.bias', 'vae.vocoder.head.resblocks.10.convs2.2.weight_g', 'vae.vocoder.head.resblocks.10.convs2.2.weight_v', 'vae.vocoder.head.resblocks.11.convs1.0.bias', 'vae.vocoder.head.resblocks.11.convs1.0.weight_g', 'vae.vocoder.head.resblocks.11.convs1.0.weight_v', 'vae.vocoder.head.resblocks.11.convs1.1.bias', 'vae.vocoder.head.resblocks.11.convs1.1.weight_g', 'vae.vocoder.head.resblocks.11.convs1.1.weight_v', 'vae.vocoder.head.resblocks.11.convs1.2.bias', 'vae.vocoder.head.resblocks.11.convs1.2.weight_g', 'vae.vocoder.head.resblocks.11.convs1.2.weight_v', 'vae.vocoder.head.resblocks.11.convs2.0.bias', 'vae.vocoder.head.resblocks.11.convs2.0.weight_g', 'vae.vocoder.head.resblocks.11.convs2.0.weight_v', 'vae.vocoder.head.resblocks.11.convs2.1.bias', 'vae.vocoder.head.resblocks.11.convs2.1.weight_g', 'vae.vocoder.head.resblocks.11.convs2.1.weight_v', 'vae.vocoder.head.resblocks.11.convs2.2.bias', 'vae.vocoder.head.resblocks.11.convs2.2.weight_g', 'vae.vocoder.head.resblocks.11.convs2.2.weight_v', 'vae.vocoder.head.resblocks.12.convs1.0.bias', 'vae.vocoder.head.resblocks.12.convs1.0.weight_g', 'vae.vocoder.head.resblocks.12.convs1.0.weight_v', 'vae.vocoder.head.resblocks.12.convs1.1.bias', 'vae.vocoder.head.resblocks.12.convs1.1.weight_g', 'vae.vocoder.head.resblocks.12.convs1.1.weight_v', 'vae.vocoder.head.resblocks.12.convs1.2.bias', 'vae.vocoder.head.resblocks.12.convs1.2.weight_g', 'vae.vocoder.head.resblocks.12.convs1.2.weight_v', 'vae.vocoder.head.resblocks.12.convs2.0.bias', 'vae.vocoder.head.resblocks.12.convs2.0.weight_g', 'vae.vocoder.head.resblocks.12.convs2.0.weight_v', 'vae.vocoder.head.resblocks.12.convs2.1.bias', 'vae.vocoder.head.resblocks.12.convs2.1.weight_g', 'vae.vocoder.head.resblocks.12.convs2.1.weight_v', 'vae.vocoder.head.resblocks.12.convs2.2.bias', 'vae.vocoder.head.resblocks.12.convs2.2.weight_g', 'vae.vocoder.head.resblocks.12.convs2.2.weight_v', 'vae.vocoder.head.resblocks.13.convs1.0.bias', 'vae.vocoder.head.resblocks.13.convs1.0.weight_g', 'vae.vocoder.head.resblocks.13.convs1.0.weight_v', 'vae.vocoder.head.resblocks.13.convs1.1.bias', 'vae.vocoder.head.resblocks.13.convs1.1.weight_g', 'vae.vocoder.head.resblocks.13.convs1.1.weight_v', 'vae.vocoder.head.resblocks.13.convs1.2.bias', 'vae.vocoder.head.resblocks.13.convs1.2.weight_g', 'vae.vocoder.head.resblocks.13.convs1.2.weight_v', 'vae.vocoder.head.resblocks.13.convs2.0.bias', 'vae.vocoder.head.resblocks.13.convs2.0.weight_g', 'vae.vocoder.head.resblocks.13.convs2.0.weight_v', 'vae.vocoder.head.resblocks.13.convs2.1.bias', 'vae.vocoder.head.resblocks.13.convs2.1.weight_g', 'vae.vocoder.head.resblocks.13.convs2.1.weight_v', 'vae.vocoder.head.resblocks.13.convs2.2.bias', 'vae.vocoder.head.resblocks.13.convs2.2.weight_g', 'vae.vocoder.head.resblocks.13.convs2.2.weight_v', 'vae.vocoder.head.resblocks.14.convs1.0.bias', 'vae.vocoder.head.resblocks.14.convs1.0.weight_g', 'vae.vocoder.head.resblocks.14.convs1.0.weight_v', 'vae.vocoder.head.resblocks.14.convs1.1.bias', 'vae.vocoder.head.resblocks.14.convs1.1.weight_g', 'vae.vocoder.head.resblocks.14.convs1.1.weight_v', 'vae.vocoder.head.resblocks.14.convs1.2.bias', 'vae.vocoder.head.resblocks.14.convs1.2.weight_g', 'vae.vocoder.head.resblocks.14.convs1.2.weight_v', 'vae.vocoder.head.resblocks.14.convs2.0.bias', 'vae.vocoder.head.resblocks.14.convs2.0.weight_g', 'vae.vocoder.head.resblocks.14.convs2.0.weight_v', 'vae.vocoder.head.resblocks.14.convs2.1.bias', 'vae.vocoder.head.resblocks.14.convs2.1.weight_g', 'vae.vocoder.head.resblocks.14.convs2.1.weight_v', 'vae.vocoder.head.resblocks.14.convs2.2.bias', 'vae.vocoder.head.resblocks.14.convs2.2.weight_g', 'vae.vocoder.head.resblocks.14.convs2.2.weight_v', 'vae.vocoder.head.resblocks.15.convs1.0.bias', 'vae.vocoder.head.resblocks.15.convs1.0.weight_g', 'vae.vocoder.head.resblocks.15.convs1.0.weight_v', 'vae.vocoder.head.resblocks.15.convs1.1.bias', 'vae.vocoder.head.resblocks.15.convs1.1.weight_g', 'vae.vocoder.head.resblocks.15.convs1.1.weight_v', 'vae.vocoder.head.resblocks.15.convs1.2.bias', 'vae.vocoder.head.resblocks.15.convs1.2.weight_g', 'vae.vocoder.head.resblocks.15.convs1.2.weight_v', 'vae.vocoder.head.resblocks.15.convs2.0.bias', 'vae.vocoder.head.resblocks.15.convs2.0.weight_g', 'vae.vocoder.head.resblocks.15.convs2.0.weight_v', 'vae.vocoder.head.resblocks.15.convs2.1.bias', 'vae.vocoder.head.resblocks.15.convs2.1.weight_g', 'vae.vocoder.head.resblocks.15.convs2.1.weight_v', 'vae.vocoder.head.resblocks.15.convs2.2.bias', 'vae.vocoder.head.resblocks.15.convs2.2.weight_g', 'vae.vocoder.head.resblocks.15.convs2.2.weight_v', 'vae.vocoder.head.resblocks.16.convs1.0.bias', 'vae.vocoder.head.resblocks.16.convs1.0.weight_g', 'vae.vocoder.head.resblocks.16.convs1.0.weight_v', 'vae.vocoder.head.resblocks.16.convs1.1.bias', 'vae.vocoder.head.resblocks.16.convs1.1.weight_g', 'vae.vocoder.head.resblocks.16.convs1.1.weight_v', 'vae.vocoder.head.resblocks.16.convs1.2.bias', 'vae.vocoder.head.resblocks.16.convs1.2.weight_g', 'vae.vocoder.head.resblocks.16.convs1.2.weight_v', 'vae.vocoder.head.resblocks.16.convs2.0.bias', 'vae.vocoder.head.resblocks.16.convs2.0.weight_g', 'vae.vocoder.head.resblocks.16.convs2.0.weight_v', 'vae.vocoder.head.resblocks.16.convs2.1.bias', 'vae.vocoder.head.resblocks.16.convs2.1.weight_g', 'vae.vocoder.head.resblocks.16.convs2.1.weight_v', 'vae.vocoder.head.resblocks.16.convs2.2.bias', 'vae.vocoder.head.resblocks.16.convs2.2.weight_g', 'vae.vocoder.head.resblocks.16.convs2.2.weight_v', 'vae.vocoder.head.resblocks.17.convs1.0.bias', 'vae.vocoder.head.resblocks.17.convs1.0.weight_g', 'vae.vocoder.head.resblocks.17.convs1.0.weight_v', 'vae.vocoder.head.resblocks.17.convs1.1.bias', 'vae.vocoder.head.resblocks.17.convs1.1.weight_g', 'vae.vocoder.head.resblocks.17.convs1.1.weight_v', 'vae.vocoder.head.resblocks.17.convs1.2.bias', 'vae.vocoder.head.resblocks.17.convs1.2.weight_g', 'vae.vocoder.head.resblocks.17.convs1.2.weight_v', 'vae.vocoder.head.resblocks.17.convs2.0.bias', 'vae.vocoder.head.resblocks.17.convs2.0.weight_g', 'vae.vocoder.head.resblocks.17.convs2.0.weight_v', 'vae.vocoder.head.resblocks.17.convs2.1.bias', 'vae.vocoder.head.resblocks.17.convs2.1.weight_g', 'vae.vocoder.head.resblocks.17.convs2.1.weight_v', 'vae.vocoder.head.resblocks.17.convs2.2.bias', 'vae.vocoder.head.resblocks.17.convs2.2.weight_g', 'vae.vocoder.head.resblocks.17.convs2.2.weight_v', 'vae.vocoder.head.resblocks.18.convs1.0.bias', 'vae.vocoder.head.resblocks.18.convs1.0.weight_g', 'vae.vocoder.head.resblocks.18.convs1.0.weight_v', 'vae.vocoder.head.resblocks.18.convs1.1.bias', 'vae.vocoder.head.resblocks.18.convs1.1.weight_g', 'vae.vocoder.head.resblocks.18.convs1.1.weight_v', 'vae.vocoder.head.resblocks.18.convs1.2.bias', 'vae.vocoder.head.resblocks.18.convs1.2.weight_g', 'vae.vocoder.head.resblocks.18.convs1.2.weight_v', 'vae.vocoder.head.resblocks.18.convs2.0.bias', 'vae.vocoder.head.resblocks.18.convs2.0.weight_g', 'vae.vocoder.head.resblocks.18.convs2.0.weight_v', 'vae.vocoder.head.resblocks.18.convs2.1.bias', 'vae.vocoder.head.resblocks.18.convs2.1.weight_g', 'vae.vocoder.head.resblocks.18.convs2.1.weight_v', 'vae.vocoder.head.resblocks.18.convs2.2.bias', 'vae.vocoder.head.resblocks.18.convs2.2.weight_g', 'vae.vocoder.head.resblocks.18.convs2.2.weight_v', 'vae.vocoder.head.resblocks.19.convs1.0.bias', 'vae.vocoder.head.resblocks.19.convs1.0.weight_g', 'vae.vocoder.head.resblocks.19.convs1.0.weight_v', 'vae.vocoder.head.resblocks.19.convs1.1.bias', 'vae.vocoder.head.resblocks.19.convs1.1.weight_g', 'vae.vocoder.head.resblocks.19.convs1.1.weight_v', 'vae.vocoder.head.resblocks.19.convs1.2.bias', 'vae.vocoder.head.resblocks.19.convs1.2.weight_g', 'vae.vocoder.head.resblocks.19.convs1.2.weight_v', 'vae.vocoder.head.resblocks.19.convs2.0.bias', 'vae.vocoder.head.resblocks.19.convs2.0.weight_g', 'vae.vocoder.head.resblocks.19.convs2.0.weight_v', 'vae.vocoder.head.resblocks.19.convs2.1.bias', 'vae.vocoder.head.resblocks.19.convs2.1.weight_g', 'vae.vocoder.head.resblocks.19.convs2.1.weight_v', 'vae.vocoder.head.resblocks.19.convs2.2.bias', 'vae.vocoder.head.resblocks.19.convs2.2.weight_g', 'vae.vocoder.head.resblocks.19.convs2.2.weight_v', 'vae.vocoder.head.resblocks.20.convs1.0.bias', 'vae.vocoder.head.resblocks.20.convs1.0.weight_g', 'vae.vocoder.head.resblocks.20.convs1.0.weight_v', 'vae.vocoder.head.resblocks.20.convs1.1.bias', 'vae.vocoder.head.resblocks.20.convs1.1.weight_g', 'vae.vocoder.head.resblocks.20.convs1.1.weight_v', 'vae.vocoder.head.resblocks.20.convs1.2.bias', 'vae.vocoder.head.resblocks.20.convs1.2.weight_g', 'vae.vocoder.head.resblocks.20.convs1.2.weight_v', 'vae.vocoder.head.resblocks.20.convs2.0.bias', 'vae.vocoder.head.resblocks.20.convs2.0.weight_g', 'vae.vocoder.head.resblocks.20.convs2.0.weight_v', 'vae.vocoder.head.resblocks.20.convs2.1.bias', 'vae.vocoder.head.resblocks.20.convs2.1.weight_g', 'vae.vocoder.head.resblocks.20.convs2.1.weight_v', 'vae.vocoder.head.resblocks.20.convs2.2.bias', 'vae.vocoder.head.resblocks.20.convs2.2.weight_g', 'vae.vocoder.head.resblocks.20.convs2.2.weight_v', 'vae.vocoder.head.resblocks.21.convs1.0.bias', 'vae.vocoder.head.resblocks.21.convs1.0.weight_g', 'vae.vocoder.head.resblocks.21.convs1.0.weight_v', 'vae.vocoder.head.resblocks.21.convs1.1.bias', 'vae.vocoder.head.resblocks.21.convs1.1.weight_g', 'vae.vocoder.head.resblocks.21.convs1.1.weight_v', 'vae.vocoder.head.resblocks.21.convs1.2.bias', 'vae.vocoder.head.resblocks.21.convs1.2.weight_g', 'vae.vocoder.head.resblocks.21.convs1.2.weight_v', 'vae.vocoder.head.resblocks.21.convs2.0.bias', 'vae.vocoder.head.resblocks.21.convs2.0.weight_g', 'vae.vocoder.head.resblocks.21.convs2.0.weight_v', 'vae.vocoder.head.resblocks.21.convs2.1.bias', 'vae.vocoder.head.resblocks.21.convs2.1.weight_g', 'vae.vocoder.head.resblocks.21.convs2.1.weight_v', 'vae.vocoder.head.resblocks.21.convs2.2.bias', 'vae.vocoder.head.resblocks.21.convs2.2.weight_g', 'vae.vocoder.head.resblocks.21.convs2.2.weight_v', 'vae.vocoder.head.resblocks.22.convs1.0.bias', 'vae.vocoder.head.resblocks.22.convs1.0.weight_g', 'vae.vocoder.head.resblocks.22.convs1.0.weight_v', 'vae.vocoder.head.resblocks.22.convs1.1.bias', 'vae.vocoder.head.resblocks.22.convs1.1.weight_g', 'vae.vocoder.head.resblocks.22.convs1.1.weight_v', 'vae.vocoder.head.resblocks.22.convs1.2.bias', 'vae.vocoder.head.resblocks.22.convs1.2.weight_g', 'vae.vocoder.head.resblocks.22.convs1.2.weight_v', 'vae.vocoder.head.resblocks.22.convs2.0.bias', 'vae.vocoder.head.resblocks.22.convs2.0.weight_g', 'vae.vocoder.head.resblocks.22.convs2.0.weight_v', 'vae.vocoder.head.resblocks.22.convs2.1.bias', 'vae.vocoder.head.resblocks.22.convs2.1.weight_g', 'vae.vocoder.head.resblocks.22.convs2.1.weight_v', 'vae.vocoder.head.resblocks.22.convs2.2.bias', 'vae.vocoder.head.resblocks.22.convs2.2.weight_g', 'vae.vocoder.head.resblocks.22.convs2.2.weight_v', 'vae.vocoder.head.resblocks.23.convs1.0.bias', 'vae.vocoder.head.resblocks.23.convs1.0.weight_g', 'vae.vocoder.head.resblocks.23.convs1.0.weight_v', 'vae.vocoder.head.resblocks.23.convs1.1.bias', 'vae.vocoder.head.resblocks.23.convs1.1.weight_g', 'vae.vocoder.head.resblocks.23.convs1.1.weight_v', 'vae.vocoder.head.resblocks.23.convs1.2.bias', 'vae.vocoder.head.resblocks.23.convs1.2.weight_g', 'vae.vocoder.head.resblocks.23.convs1.2.weight_v', 'vae.vocoder.head.resblocks.23.convs2.0.bias', 'vae.vocoder.head.resblocks.23.convs2.0.weight_g', 'vae.vocoder.head.resblocks.23.convs2.0.weight_v', 'vae.vocoder.head.resblocks.23.convs2.1.bias', 'vae.vocoder.head.resblocks.23.convs2.1.weight_g', 'vae.vocoder.head.resblocks.23.convs2.1.weight_v', 'vae.vocoder.head.resblocks.23.convs2.2.bias', 'vae.vocoder.head.resblocks.23.convs2.2.weight_g', 'vae.vocoder.head.resblocks.23.convs2.2.weight_v', 'vae.vocoder.head.resblocks.24.convs1.0.bias', 'vae.vocoder.head.resblocks.24.convs1.0.weight_g', 'vae.vocoder.head.resblocks.24.convs1.0.weight_v', 'vae.vocoder.head.resblocks.24.convs1.1.bias', 'vae.vocoder.head.resblocks.24.convs1.1.weight_g', 'vae.vocoder.head.resblocks.24.convs1.1.weight_v', 'vae.vocoder.head.resblocks.24.convs1.2.bias', 'vae.vocoder.head.resblocks.24.convs1.2.weight_g', 'vae.vocoder.head.resblocks.24.convs1.2.weight_v', 'vae.vocoder.head.resblocks.24.convs2.0.bias', 'vae.vocoder.head.resblocks.24.convs2.0.weight_g', 'vae.vocoder.head.resblocks.24.convs2.0.weight_v', 'vae.vocoder.head.resblocks.24.convs2.1.bias', 'vae.vocoder.head.resblocks.24.convs2.1.weight_g', 'vae.vocoder.head.resblocks.24.convs2.1.weight_v', 'vae.vocoder.head.resblocks.24.convs2.2.bias', 'vae.vocoder.head.resblocks.24.convs2.2.weight_g', 'vae.vocoder.head.resblocks.24.convs2.2.weight_v', 'vae.vocoder.head.resblocks.25.convs1.0.bias', 'vae.vocoder.head.resblocks.25.convs1.0.weight_g', 'vae.vocoder.head.resblocks.25.convs1.0.weight_v', 'vae.vocoder.head.resblocks.25.convs1.1.bias', 'vae.vocoder.head.resblocks.25.convs1.1.weight_g', 'vae.vocoder.head.resblocks.25.convs1.1.weight_v', 'vae.vocoder.head.resblocks.25.convs1.2.bias', 'vae.vocoder.head.resblocks.25.convs1.2.weight_g', 'vae.vocoder.head.resblocks.25.convs1.2.weight_v', 'vae.vocoder.head.resblocks.25.convs2.0.bias', 'vae.vocoder.head.resblocks.25.convs2.0.weight_g', 'vae.vocoder.head.resblocks.25.convs2.0.weight_v', 'vae.vocoder.head.resblocks.25.convs2.1.bias', 'vae.vocoder.head.resblocks.25.convs2.1.weight_g', 'vae.vocoder.head.resblocks.25.convs2.1.weight_v', 'vae.vocoder.head.resblocks.25.convs2.2.bias', 'vae.vocoder.head.resblocks.25.convs2.2.weight_g', 'vae.vocoder.head.resblocks.25.convs2.2.weight_v', 'vae.vocoder.head.resblocks.26.convs1.0.bias', 'vae.vocoder.head.resblocks.26.convs1.0.weight_g', 'vae.vocoder.head.resblocks.26.convs1.0.weight_v', 'vae.vocoder.head.resblocks.26.convs1.1.bias', 'vae.vocoder.head.resblocks.26.convs1.1.weight_g', 'vae.vocoder.head.resblocks.26.convs1.1.weight_v', 'vae.vocoder.head.resblocks.26.convs1.2.bias', 'vae.vocoder.head.resblocks.26.convs1.2.weight_g', 'vae.vocoder.head.resblocks.26.convs1.2.weight_v', 'vae.vocoder.head.resblocks.26.convs2.0.bias', 'vae.vocoder.head.resblocks.26.convs2.0.weight_g', 'vae.vocoder.head.resblocks.26.convs2.0.weight_v', 'vae.vocoder.head.resblocks.26.convs2.1.bias', 'vae.vocoder.head.resblocks.26.convs2.1.weight_g', 'vae.vocoder.head.resblocks.26.convs2.1.weight_v', 'vae.vocoder.head.resblocks.26.convs2.2.bias', 'vae.vocoder.head.resblocks.26.convs2.2.weight_g', 'vae.vocoder.head.resblocks.26.convs2.2.weight_v', 'vae.vocoder.head.resblocks.27.convs1.0.bias', 'vae.vocoder.head.resblocks.27.convs1.0.weight_g', 'vae.vocoder.head.resblocks.27.convs1.0.weight_v', 'vae.vocoder.head.resblocks.27.convs1.1.bias', 'vae.vocoder.head.resblocks.27.convs1.1.weight_g', 'vae.vocoder.head.resblocks.27.convs1.1.weight_v', 'vae.vocoder.head.resblocks.27.convs1.2.bias', 'vae.vocoder.head.resblocks.27.convs1.2.weight_g', 'vae.vocoder.head.resblocks.27.convs1.2.weight_v', 'vae.vocoder.head.resblocks.27.convs2.0.bias', 'vae.vocoder.head.resblocks.27.convs2.0.weight_g', 'vae.vocoder.head.resblocks.27.convs2.0.weight_v', 'vae.vocoder.head.resblocks.27.convs2.1.bias', 'vae.vocoder.head.resblocks.27.convs2.1.weight_g', 'vae.vocoder.head.resblocks.27.convs2.1.weight_v', 'vae.vocoder.head.resblocks.27.convs2.2.bias', 'vae.vocoder.head.resblocks.27.convs2.2.weight_g', 'vae.vocoder.head.resblocks.27.convs2.2.weight_v', 'vae.vocoder.head.conv_post.bias', 'vae.vocoder.head.conv_post.weight_g', 'vae.vocoder.head.conv_post.weight_v', 'transformers.transformer_blocks.0.scale_shift_table', 'transformers.transformer_blocks.0.attn.to_q.weight', 'transformers.transformer_blocks.0.attn.to_q.bias', 'transformers.transformer_blocks.0.attn.to_k.weight', 'transformers.transformer_blocks.0.attn.to_k.bias', 'transformers.transformer_blocks.0.attn.to_v.weight', 'transformers.transformer_blocks.0.attn.to_v.bias', 'transformers.transformer_blocks.0.attn.to_out.0.weight', 'transformers.transformer_blocks.0.attn.to_out.0.bias', 'transformers.transformer_blocks.0.cross_attn.to_q.weight', 'transformers.transformer_blocks.0.cross_attn.to_q.bias', 'transformers.transformer_blocks.0.cross_attn.to_k.weight', 'transformers.transformer_blocks.0.cross_attn.to_k.bias', 'transformers.transformer_blocks.0.cross_attn.to_v.weight', 'transformers.transformer_blocks.0.cross_attn.to_v.bias', 'transformers.transformer_blocks.0.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.0.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.0.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.0.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.0.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.0.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.0.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.0.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.0.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.0.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.0.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.0.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.0.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.0.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.0.ff.point_conv.conv.weight', 'transformers.transformer_blocks.1.scale_shift_table', 'transformers.transformer_blocks.1.attn.to_q.weight', 'transformers.transformer_blocks.1.attn.to_q.bias', 'transformers.transformer_blocks.1.attn.to_k.weight', 'transformers.transformer_blocks.1.attn.to_k.bias', 'transformers.transformer_blocks.1.attn.to_v.weight', 'transformers.transformer_blocks.1.attn.to_v.bias', 'transformers.transformer_blocks.1.attn.to_out.0.weight', 'transformers.transformer_blocks.1.attn.to_out.0.bias', 'transformers.transformer_blocks.1.cross_attn.to_q.weight', 'transformers.transformer_blocks.1.cross_attn.to_q.bias', 'transformers.transformer_blocks.1.cross_attn.to_k.weight', 'transformers.transformer_blocks.1.cross_attn.to_k.bias', 'transformers.transformer_blocks.1.cross_attn.to_v.weight', 'transformers.transformer_blocks.1.cross_attn.to_v.bias', 'transformers.transformer_blocks.1.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.1.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.1.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.1.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.1.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.1.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.1.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.1.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.1.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.1.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.1.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.1.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.1.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.1.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.1.ff.point_conv.conv.weight', 'transformers.transformer_blocks.2.scale_shift_table', 'transformers.transformer_blocks.2.attn.to_q.weight', 'transformers.transformer_blocks.2.attn.to_q.bias', 'transformers.transformer_blocks.2.attn.to_k.weight', 'transformers.transformer_blocks.2.attn.to_k.bias', 'transformers.transformer_blocks.2.attn.to_v.weight', 'transformers.transformer_blocks.2.attn.to_v.bias', 'transformers.transformer_blocks.2.attn.to_out.0.weight', 'transformers.transformer_blocks.2.attn.to_out.0.bias', 'transformers.transformer_blocks.2.cross_attn.to_q.weight', 'transformers.transformer_blocks.2.cross_attn.to_q.bias', 'transformers.transformer_blocks.2.cross_attn.to_k.weight', 'transformers.transformer_blocks.2.cross_attn.to_k.bias', 'transformers.transformer_blocks.2.cross_attn.to_v.weight', 'transformers.transformer_blocks.2.cross_attn.to_v.bias', 'transformers.transformer_blocks.2.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.2.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.2.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.2.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.2.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.2.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.2.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.2.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.2.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.2.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.2.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.2.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.2.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.2.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.2.ff.point_conv.conv.weight', 'transformers.transformer_blocks.3.scale_shift_table', 'transformers.transformer_blocks.3.attn.to_q.weight', 'transformers.transformer_blocks.3.attn.to_q.bias', 'transformers.transformer_blocks.3.attn.to_k.weight', 'transformers.transformer_blocks.3.attn.to_k.bias', 'transformers.transformer_blocks.3.attn.to_v.weight', 'transformers.transformer_blocks.3.attn.to_v.bias', 'transformers.transformer_blocks.3.attn.to_out.0.weight', 'transformers.transformer_blocks.3.attn.to_out.0.bias', 'transformers.transformer_blocks.3.cross_attn.to_q.weight', 'transformers.transformer_blocks.3.cross_attn.to_q.bias', 'transformers.transformer_blocks.3.cross_attn.to_k.weight', 'transformers.transformer_blocks.3.cross_attn.to_k.bias', 'transformers.transformer_blocks.3.cross_attn.to_v.weight', 'transformers.transformer_blocks.3.cross_attn.to_v.bias', 'transformers.transformer_blocks.3.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.3.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.3.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.3.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.3.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.3.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.3.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.3.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.3.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.3.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.3.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.3.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.3.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.3.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.3.ff.point_conv.conv.weight', 'transformers.transformer_blocks.4.scale_shift_table', 'transformers.transformer_blocks.4.attn.to_q.weight', 'transformers.transformer_blocks.4.attn.to_q.bias', 'transformers.transformer_blocks.4.attn.to_k.weight', 'transformers.transformer_blocks.4.attn.to_k.bias', 'transformers.transformer_blocks.4.attn.to_v.weight', 'transformers.transformer_blocks.4.attn.to_v.bias', 'transformers.transformer_blocks.4.attn.to_out.0.weight', 'transformers.transformer_blocks.4.attn.to_out.0.bias', 'transformers.transformer_blocks.4.cross_attn.to_q.weight', 'transformers.transformer_blocks.4.cross_attn.to_q.bias', 'transformers.transformer_blocks.4.cross_attn.to_k.weight', 'transformers.transformer_blocks.4.cross_attn.to_k.bias', 'transformers.transformer_blocks.4.cross_attn.to_v.weight', 'transformers.transformer_blocks.4.cross_attn.to_v.bias', 'transformers.transformer_blocks.4.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.4.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.4.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.4.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.4.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.4.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.4.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.4.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.4.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.4.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.4.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.4.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.4.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.4.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.4.ff.point_conv.conv.weight', 'transformers.transformer_blocks.5.scale_shift_table', 'transformers.transformer_blocks.5.attn.to_q.weight', 'transformers.transformer_blocks.5.attn.to_q.bias', 'transformers.transformer_blocks.5.attn.to_k.weight', 'transformers.transformer_blocks.5.attn.to_k.bias', 'transformers.transformer_blocks.5.attn.to_v.weight', 'transformers.transformer_blocks.5.attn.to_v.bias', 'transformers.transformer_blocks.5.attn.to_out.0.weight', 'transformers.transformer_blocks.5.attn.to_out.0.bias', 'transformers.transformer_blocks.5.cross_attn.to_q.weight', 'transformers.transformer_blocks.5.cross_attn.to_q.bias', 'transformers.transformer_blocks.5.cross_attn.to_k.weight', 'transformers.transformer_blocks.5.cross_attn.to_k.bias', 'transformers.transformer_blocks.5.cross_attn.to_v.weight', 'transformers.transformer_blocks.5.cross_attn.to_v.bias', 'transformers.transformer_blocks.5.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.5.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.5.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.5.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.5.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.5.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.5.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.5.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.5.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.5.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.5.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.5.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.5.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.5.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.5.ff.point_conv.conv.weight', 'transformers.transformer_blocks.6.scale_shift_table', 'transformers.transformer_blocks.6.attn.to_q.weight', 'transformers.transformer_blocks.6.attn.to_q.bias', 'transformers.transformer_blocks.6.attn.to_k.weight', 'transformers.transformer_blocks.6.attn.to_k.bias', 'transformers.transformer_blocks.6.attn.to_v.weight', 'transformers.transformer_blocks.6.attn.to_v.bias', 'transformers.transformer_blocks.6.attn.to_out.0.weight', 'transformers.transformer_blocks.6.attn.to_out.0.bias', 'transformers.transformer_blocks.6.cross_attn.to_q.weight', 'transformers.transformer_blocks.6.cross_attn.to_q.bias', 'transformers.transformer_blocks.6.cross_attn.to_k.weight', 'transformers.transformer_blocks.6.cross_attn.to_k.bias', 'transformers.transformer_blocks.6.cross_attn.to_v.weight', 'transformers.transformer_blocks.6.cross_attn.to_v.bias', 'transformers.transformer_blocks.6.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.6.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.6.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.6.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.6.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.6.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.6.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.6.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.6.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.6.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.6.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.6.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.6.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.6.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.6.ff.point_conv.conv.weight', 'transformers.transformer_blocks.7.scale_shift_table', 'transformers.transformer_blocks.7.attn.to_q.weight', 'transformers.transformer_blocks.7.attn.to_q.bias', 'transformers.transformer_blocks.7.attn.to_k.weight', 'transformers.transformer_blocks.7.attn.to_k.bias', 'transformers.transformer_blocks.7.attn.to_v.weight', 'transformers.transformer_blocks.7.attn.to_v.bias', 'transformers.transformer_blocks.7.attn.to_out.0.weight', 'transformers.transformer_blocks.7.attn.to_out.0.bias', 'transformers.transformer_blocks.7.cross_attn.to_q.weight', 'transformers.transformer_blocks.7.cross_attn.to_q.bias', 'transformers.transformer_blocks.7.cross_attn.to_k.weight', 'transformers.transformer_blocks.7.cross_attn.to_k.bias', 'transformers.transformer_blocks.7.cross_attn.to_v.weight', 'transformers.transformer_blocks.7.cross_attn.to_v.bias', 'transformers.transformer_blocks.7.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.7.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.7.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.7.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.7.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.7.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.7.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.7.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.7.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.7.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.7.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.7.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.7.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.7.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.7.ff.point_conv.conv.weight', 'transformers.transformer_blocks.8.scale_shift_table', 'transformers.transformer_blocks.8.attn.to_q.weight', 'transformers.transformer_blocks.8.attn.to_q.bias', 'transformers.transformer_blocks.8.attn.to_k.weight', 'transformers.transformer_blocks.8.attn.to_k.bias', 'transformers.transformer_blocks.8.attn.to_v.weight', 'transformers.transformer_blocks.8.attn.to_v.bias', 'transformers.transformer_blocks.8.attn.to_out.0.weight', 'transformers.transformer_blocks.8.attn.to_out.0.bias', 'transformers.transformer_blocks.8.cross_attn.to_q.weight', 'transformers.transformer_blocks.8.cross_attn.to_q.bias', 'transformers.transformer_blocks.8.cross_attn.to_k.weight', 'transformers.transformer_blocks.8.cross_attn.to_k.bias', 'transformers.transformer_blocks.8.cross_attn.to_v.weight', 'transformers.transformer_blocks.8.cross_attn.to_v.bias', 'transformers.transformer_blocks.8.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.8.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.8.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.8.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.8.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.8.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.8.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.8.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.8.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.8.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.8.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.8.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.8.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.8.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.8.ff.point_conv.conv.weight', 'transformers.transformer_blocks.9.scale_shift_table', 'transformers.transformer_blocks.9.attn.to_q.weight', 'transformers.transformer_blocks.9.attn.to_q.bias', 'transformers.transformer_blocks.9.attn.to_k.weight', 'transformers.transformer_blocks.9.attn.to_k.bias', 'transformers.transformer_blocks.9.attn.to_v.weight', 'transformers.transformer_blocks.9.attn.to_v.bias', 'transformers.transformer_blocks.9.attn.to_out.0.weight', 'transformers.transformer_blocks.9.attn.to_out.0.bias', 'transformers.transformer_blocks.9.cross_attn.to_q.weight', 'transformers.transformer_blocks.9.cross_attn.to_q.bias', 'transformers.transformer_blocks.9.cross_attn.to_k.weight', 'transformers.transformer_blocks.9.cross_attn.to_k.bias', 'transformers.transformer_blocks.9.cross_attn.to_v.weight', 'transformers.transformer_blocks.9.cross_attn.to_v.bias', 'transformers.transformer_blocks.9.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.9.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.9.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.9.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.9.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.9.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.9.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.9.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.9.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.9.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.9.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.9.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.9.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.9.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.9.ff.point_conv.conv.weight', 'transformers.transformer_blocks.10.scale_shift_table', 'transformers.transformer_blocks.10.attn.to_q.weight', 'transformers.transformer_blocks.10.attn.to_q.bias', 'transformers.transformer_blocks.10.attn.to_k.weight', 'transformers.transformer_blocks.10.attn.to_k.bias', 'transformers.transformer_blocks.10.attn.to_v.weight', 'transformers.transformer_blocks.10.attn.to_v.bias', 'transformers.transformer_blocks.10.attn.to_out.0.weight', 'transformers.transformer_blocks.10.attn.to_out.0.bias', 'transformers.transformer_blocks.10.cross_attn.to_q.weight', 'transformers.transformer_blocks.10.cross_attn.to_q.bias', 'transformers.transformer_blocks.10.cross_attn.to_k.weight', 'transformers.transformer_blocks.10.cross_attn.to_k.bias', 'transformers.transformer_blocks.10.cross_attn.to_v.weight', 'transformers.transformer_blocks.10.cross_attn.to_v.bias', 'transformers.transformer_blocks.10.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.10.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.10.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.10.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.10.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.10.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.10.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.10.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.10.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.10.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.10.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.10.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.10.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.10.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.10.ff.point_conv.conv.weight', 'transformers.transformer_blocks.11.scale_shift_table', 'transformers.transformer_blocks.11.attn.to_q.weight', 'transformers.transformer_blocks.11.attn.to_q.bias', 'transformers.transformer_blocks.11.attn.to_k.weight', 'transformers.transformer_blocks.11.attn.to_k.bias', 'transformers.transformer_blocks.11.attn.to_v.weight', 'transformers.transformer_blocks.11.attn.to_v.bias', 'transformers.transformer_blocks.11.attn.to_out.0.weight', 'transformers.transformer_blocks.11.attn.to_out.0.bias', 'transformers.transformer_blocks.11.cross_attn.to_q.weight', 'transformers.transformer_blocks.11.cross_attn.to_q.bias', 'transformers.transformer_blocks.11.cross_attn.to_k.weight', 'transformers.transformer_blocks.11.cross_attn.to_k.bias', 'transformers.transformer_blocks.11.cross_attn.to_v.weight', 'transformers.transformer_blocks.11.cross_attn.to_v.bias', 'transformers.transformer_blocks.11.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.11.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.11.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.11.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.11.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.11.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.11.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.11.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.11.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.11.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.11.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.11.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.11.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.11.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.11.ff.point_conv.conv.weight', 'transformers.transformer_blocks.12.scale_shift_table', 'transformers.transformer_blocks.12.attn.to_q.weight', 'transformers.transformer_blocks.12.attn.to_q.bias', 'transformers.transformer_blocks.12.attn.to_k.weight', 'transformers.transformer_blocks.12.attn.to_k.bias', 'transformers.transformer_blocks.12.attn.to_v.weight', 'transformers.transformer_blocks.12.attn.to_v.bias', 'transformers.transformer_blocks.12.attn.to_out.0.weight', 'transformers.transformer_blocks.12.attn.to_out.0.bias', 'transformers.transformer_blocks.12.cross_attn.to_q.weight', 'transformers.transformer_blocks.12.cross_attn.to_q.bias', 'transformers.transformer_blocks.12.cross_attn.to_k.weight', 'transformers.transformer_blocks.12.cross_attn.to_k.bias', 'transformers.transformer_blocks.12.cross_attn.to_v.weight', 'transformers.transformer_blocks.12.cross_attn.to_v.bias', 'transformers.transformer_blocks.12.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.12.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.12.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.12.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.12.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.12.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.12.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.12.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.12.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.12.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.12.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.12.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.12.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.12.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.12.ff.point_conv.conv.weight', 'transformers.transformer_blocks.13.scale_shift_table', 'transformers.transformer_blocks.13.attn.to_q.weight', 'transformers.transformer_blocks.13.attn.to_q.bias', 'transformers.transformer_blocks.13.attn.to_k.weight', 'transformers.transformer_blocks.13.attn.to_k.bias', 'transformers.transformer_blocks.13.attn.to_v.weight', 'transformers.transformer_blocks.13.attn.to_v.bias', 'transformers.transformer_blocks.13.attn.to_out.0.weight', 'transformers.transformer_blocks.13.attn.to_out.0.bias', 'transformers.transformer_blocks.13.cross_attn.to_q.weight', 'transformers.transformer_blocks.13.cross_attn.to_q.bias', 'transformers.transformer_blocks.13.cross_attn.to_k.weight', 'transformers.transformer_blocks.13.cross_attn.to_k.bias', 'transformers.transformer_blocks.13.cross_attn.to_v.weight', 'transformers.transformer_blocks.13.cross_attn.to_v.bias', 'transformers.transformer_blocks.13.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.13.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.13.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.13.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.13.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.13.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.13.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.13.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.13.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.13.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.13.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.13.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.13.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.13.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.13.ff.point_conv.conv.weight', 'transformers.transformer_blocks.14.scale_shift_table', 'transformers.transformer_blocks.14.attn.to_q.weight', 'transformers.transformer_blocks.14.attn.to_q.bias', 'transformers.transformer_blocks.14.attn.to_k.weight', 'transformers.transformer_blocks.14.attn.to_k.bias', 'transformers.transformer_blocks.14.attn.to_v.weight', 'transformers.transformer_blocks.14.attn.to_v.bias', 'transformers.transformer_blocks.14.attn.to_out.0.weight', 'transformers.transformer_blocks.14.attn.to_out.0.bias', 'transformers.transformer_blocks.14.cross_attn.to_q.weight', 'transformers.transformer_blocks.14.cross_attn.to_q.bias', 'transformers.transformer_blocks.14.cross_attn.to_k.weight', 'transformers.transformer_blocks.14.cross_attn.to_k.bias', 'transformers.transformer_blocks.14.cross_attn.to_v.weight', 'transformers.transformer_blocks.14.cross_attn.to_v.bias', 'transformers.transformer_blocks.14.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.14.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.14.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.14.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.14.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.14.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.14.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.14.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.14.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.14.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.14.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.14.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.14.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.14.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.14.ff.point_conv.conv.weight', 'transformers.transformer_blocks.15.scale_shift_table', 'transformers.transformer_blocks.15.attn.to_q.weight', 'transformers.transformer_blocks.15.attn.to_q.bias', 'transformers.transformer_blocks.15.attn.to_k.weight', 'transformers.transformer_blocks.15.attn.to_k.bias', 'transformers.transformer_blocks.15.attn.to_v.weight', 'transformers.transformer_blocks.15.attn.to_v.bias', 'transformers.transformer_blocks.15.attn.to_out.0.weight', 'transformers.transformer_blocks.15.attn.to_out.0.bias', 'transformers.transformer_blocks.15.cross_attn.to_q.weight', 'transformers.transformer_blocks.15.cross_attn.to_q.bias', 'transformers.transformer_blocks.15.cross_attn.to_k.weight', 'transformers.transformer_blocks.15.cross_attn.to_k.bias', 'transformers.transformer_blocks.15.cross_attn.to_v.weight', 'transformers.transformer_blocks.15.cross_attn.to_v.bias', 'transformers.transformer_blocks.15.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.15.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.15.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.15.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.15.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.15.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.15.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.15.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.15.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.15.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.15.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.15.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.15.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.15.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.15.ff.point_conv.conv.weight', 'transformers.transformer_blocks.16.scale_shift_table', 'transformers.transformer_blocks.16.attn.to_q.weight', 'transformers.transformer_blocks.16.attn.to_q.bias', 'transformers.transformer_blocks.16.attn.to_k.weight', 'transformers.transformer_blocks.16.attn.to_k.bias', 'transformers.transformer_blocks.16.attn.to_v.weight', 'transformers.transformer_blocks.16.attn.to_v.bias', 'transformers.transformer_blocks.16.attn.to_out.0.weight', 'transformers.transformer_blocks.16.attn.to_out.0.bias', 'transformers.transformer_blocks.16.cross_attn.to_q.weight', 'transformers.transformer_blocks.16.cross_attn.to_q.bias', 'transformers.transformer_blocks.16.cross_attn.to_k.weight', 'transformers.transformer_blocks.16.cross_attn.to_k.bias', 'transformers.transformer_blocks.16.cross_attn.to_v.weight', 'transformers.transformer_blocks.16.cross_attn.to_v.bias', 'transformers.transformer_blocks.16.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.16.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.16.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.16.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.16.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.16.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.16.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.16.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.16.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.16.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.16.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.16.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.16.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.16.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.16.ff.point_conv.conv.weight', 'transformers.transformer_blocks.17.scale_shift_table', 'transformers.transformer_blocks.17.attn.to_q.weight', 'transformers.transformer_blocks.17.attn.to_q.bias', 'transformers.transformer_blocks.17.attn.to_k.weight', 'transformers.transformer_blocks.17.attn.to_k.bias', 'transformers.transformer_blocks.17.attn.to_v.weight', 'transformers.transformer_blocks.17.attn.to_v.bias', 'transformers.transformer_blocks.17.attn.to_out.0.weight', 'transformers.transformer_blocks.17.attn.to_out.0.bias', 'transformers.transformer_blocks.17.cross_attn.to_q.weight', 'transformers.transformer_blocks.17.cross_attn.to_q.bias', 'transformers.transformer_blocks.17.cross_attn.to_k.weight', 'transformers.transformer_blocks.17.cross_attn.to_k.bias', 'transformers.transformer_blocks.17.cross_attn.to_v.weight', 'transformers.transformer_blocks.17.cross_attn.to_v.bias', 'transformers.transformer_blocks.17.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.17.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.17.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.17.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.17.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.17.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.17.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.17.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.17.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.17.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.17.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.17.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.17.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.17.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.17.ff.point_conv.conv.weight', 'transformers.transformer_blocks.18.scale_shift_table', 'transformers.transformer_blocks.18.attn.to_q.weight', 'transformers.transformer_blocks.18.attn.to_q.bias', 'transformers.transformer_blocks.18.attn.to_k.weight', 'transformers.transformer_blocks.18.attn.to_k.bias', 'transformers.transformer_blocks.18.attn.to_v.weight', 'transformers.transformer_blocks.18.attn.to_v.bias', 'transformers.transformer_blocks.18.attn.to_out.0.weight', 'transformers.transformer_blocks.18.attn.to_out.0.bias', 'transformers.transformer_blocks.18.cross_attn.to_q.weight', 'transformers.transformer_blocks.18.cross_attn.to_q.bias', 'transformers.transformer_blocks.18.cross_attn.to_k.weight', 'transformers.transformer_blocks.18.cross_attn.to_k.bias', 'transformers.transformer_blocks.18.cross_attn.to_v.weight', 'transformers.transformer_blocks.18.cross_attn.to_v.bias', 'transformers.transformer_blocks.18.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.18.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.18.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.18.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.18.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.18.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.18.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.18.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.18.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.18.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.18.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.18.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.18.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.18.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.18.ff.point_conv.conv.weight', 'transformers.transformer_blocks.19.scale_shift_table', 'transformers.transformer_blocks.19.attn.to_q.weight', 'transformers.transformer_blocks.19.attn.to_q.bias', 'transformers.transformer_blocks.19.attn.to_k.weight', 'transformers.transformer_blocks.19.attn.to_k.bias', 'transformers.transformer_blocks.19.attn.to_v.weight', 'transformers.transformer_blocks.19.attn.to_v.bias', 'transformers.transformer_blocks.19.attn.to_out.0.weight', 'transformers.transformer_blocks.19.attn.to_out.0.bias', 'transformers.transformer_blocks.19.cross_attn.to_q.weight', 'transformers.transformer_blocks.19.cross_attn.to_q.bias', 'transformers.transformer_blocks.19.cross_attn.to_k.weight', 'transformers.transformer_blocks.19.cross_attn.to_k.bias', 'transformers.transformer_blocks.19.cross_attn.to_v.weight', 'transformers.transformer_blocks.19.cross_attn.to_v.bias', 'transformers.transformer_blocks.19.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.19.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.19.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.19.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.19.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.19.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.19.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.19.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.19.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.19.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.19.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.19.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.19.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.19.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.19.ff.point_conv.conv.weight', 'transformers.transformer_blocks.20.scale_shift_table', 'transformers.transformer_blocks.20.attn.to_q.weight', 'transformers.transformer_blocks.20.attn.to_q.bias', 'transformers.transformer_blocks.20.attn.to_k.weight', 'transformers.transformer_blocks.20.attn.to_k.bias', 'transformers.transformer_blocks.20.attn.to_v.weight', 'transformers.transformer_blocks.20.attn.to_v.bias', 'transformers.transformer_blocks.20.attn.to_out.0.weight', 'transformers.transformer_blocks.20.attn.to_out.0.bias', 'transformers.transformer_blocks.20.cross_attn.to_q.weight', 'transformers.transformer_blocks.20.cross_attn.to_q.bias', 'transformers.transformer_blocks.20.cross_attn.to_k.weight', 'transformers.transformer_blocks.20.cross_attn.to_k.bias', 'transformers.transformer_blocks.20.cross_attn.to_v.weight', 'transformers.transformer_blocks.20.cross_attn.to_v.bias', 'transformers.transformer_blocks.20.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.20.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.20.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.20.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.20.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.20.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.20.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.20.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.20.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.20.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.20.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.20.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.20.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.20.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.20.ff.point_conv.conv.weight', 'transformers.transformer_blocks.21.scale_shift_table', 'transformers.transformer_blocks.21.attn.to_q.weight', 'transformers.transformer_blocks.21.attn.to_q.bias', 'transformers.transformer_blocks.21.attn.to_k.weight', 'transformers.transformer_blocks.21.attn.to_k.bias', 'transformers.transformer_blocks.21.attn.to_v.weight', 'transformers.transformer_blocks.21.attn.to_v.bias', 'transformers.transformer_blocks.21.attn.to_out.0.weight', 'transformers.transformer_blocks.21.attn.to_out.0.bias', 'transformers.transformer_blocks.21.cross_attn.to_q.weight', 'transformers.transformer_blocks.21.cross_attn.to_q.bias', 'transformers.transformer_blocks.21.cross_attn.to_k.weight', 'transformers.transformer_blocks.21.cross_attn.to_k.bias', 'transformers.transformer_blocks.21.cross_attn.to_v.weight', 'transformers.transformer_blocks.21.cross_attn.to_v.bias', 'transformers.transformer_blocks.21.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.21.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.21.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.21.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.21.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.21.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.21.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.21.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.21.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.21.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.21.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.21.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.21.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.21.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.21.ff.point_conv.conv.weight', 'transformers.transformer_blocks.22.scale_shift_table', 'transformers.transformer_blocks.22.attn.to_q.weight', 'transformers.transformer_blocks.22.attn.to_q.bias', 'transformers.transformer_blocks.22.attn.to_k.weight', 'transformers.transformer_blocks.22.attn.to_k.bias', 'transformers.transformer_blocks.22.attn.to_v.weight', 'transformers.transformer_blocks.22.attn.to_v.bias', 'transformers.transformer_blocks.22.attn.to_out.0.weight', 'transformers.transformer_blocks.22.attn.to_out.0.bias', 'transformers.transformer_blocks.22.cross_attn.to_q.weight', 'transformers.transformer_blocks.22.cross_attn.to_q.bias', 'transformers.transformer_blocks.22.cross_attn.to_k.weight', 'transformers.transformer_blocks.22.cross_attn.to_k.bias', 'transformers.transformer_blocks.22.cross_attn.to_v.weight', 'transformers.transformer_blocks.22.cross_attn.to_v.bias', 'transformers.transformer_blocks.22.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.22.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.22.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.22.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.22.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.22.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.22.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.22.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.22.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.22.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.22.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.22.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.22.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.22.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.22.ff.point_conv.conv.weight', 'transformers.transformer_blocks.23.scale_shift_table', 'transformers.transformer_blocks.23.attn.to_q.weight', 'transformers.transformer_blocks.23.attn.to_q.bias', 'transformers.transformer_blocks.23.attn.to_k.weight', 'transformers.transformer_blocks.23.attn.to_k.bias', 'transformers.transformer_blocks.23.attn.to_v.weight', 'transformers.transformer_blocks.23.attn.to_v.bias', 'transformers.transformer_blocks.23.attn.to_out.0.weight', 'transformers.transformer_blocks.23.attn.to_out.0.bias', 'transformers.transformer_blocks.23.cross_attn.to_q.weight', 'transformers.transformer_blocks.23.cross_attn.to_q.bias', 'transformers.transformer_blocks.23.cross_attn.to_k.weight', 'transformers.transformer_blocks.23.cross_attn.to_k.bias', 'transformers.transformer_blocks.23.cross_attn.to_v.weight', 'transformers.transformer_blocks.23.cross_attn.to_v.bias', 'transformers.transformer_blocks.23.cross_attn.add_k_proj.weight', 'transformers.transformer_blocks.23.cross_attn.add_k_proj.bias', 'transformers.transformer_blocks.23.cross_attn.add_v_proj.weight', 'transformers.transformer_blocks.23.cross_attn.add_v_proj.bias', 'transformers.transformer_blocks.23.cross_attn.add_q_proj.weight', 'transformers.transformer_blocks.23.cross_attn.add_q_proj.bias', 'transformers.transformer_blocks.23.cross_attn.to_out.0.weight', 'transformers.transformer_blocks.23.cross_attn.to_out.0.bias', 'transformers.transformer_blocks.23.cross_attn.to_add_out.weight', 'transformers.transformer_blocks.23.cross_attn.to_add_out.bias', 'transformers.transformer_blocks.23.ff.inverted_conv.conv.weight', 'transformers.transformer_blocks.23.ff.inverted_conv.conv.bias', 'transformers.transformer_blocks.23.ff.depth_conv.conv.weight', 'transformers.transformer_blocks.23.ff.depth_conv.conv.bias', 'transformers.transformer_blocks.23.ff.point_conv.conv.weight', 'transformers.timestep_embedder.linear_1.weight', 'transformers.timestep_embedder.linear_1.bias', 'transformers.timestep_embedder.linear_2.weight', 'transformers.timestep_embedder.linear_2.bias', 'transformers.t_block.1.weight', 'transformers.t_block.1.bias', 'transformers.speaker_embedder.weight', 'transformers.speaker_embedder.bias', 'transformers.genre_embedder.weight', 'transformers.genre_embedder.bias', 'transformers.lyric_embs.weight', 'transformers.lyric_encoder.embed.out.0.weight', 'transformers.lyric_encoder.embed.out.0.bias', 'transformers.lyric_encoder.embed.out.1.weight', 'transformers.lyric_encoder.embed.out.1.bias', 'transformers.lyric_encoder.after_norm.weight', 'transformers.lyric_encoder.after_norm.bias', 'transformers.lyric_encoder.encoders.0.self_attn.pos_bias_u', 'transformers.lyric_encoder.encoders.0.self_attn.pos_bias_v', 'transformers.lyric_encoder.encoders.0.self_attn.linear_q.weight', 'transformers.lyric_encoder.encoders.0.self_attn.linear_q.bias', 'transformers.lyric_encoder.encoders.0.self_attn.linear_k.weight', 'transformers.lyric_encoder.encoders.0.self_attn.linear_k.bias', 'transformers.lyric_encoder.encoders.0.self_attn.linear_v.weight', 'transformers.lyric_encoder.encoders.0.self_attn.linear_v.bias', 'transformers.lyric_encoder.encoders.0.self_attn.linear_out.weight', 'transformers.lyric_encoder.encoders.0.self_attn.linear_out.bias', 'transformers.lyric_encoder.encoders.0.self_attn.linear_pos.weight', 'transformers.lyric_encoder.encoders.0.feed_forward.w_1.weight', 'transformers.lyric_encoder.encoders.0.feed_forward.w_1.bias', 'transformers.lyric_encoder.encoders.0.feed_forward.w_2.weight', 'transformers.lyric_encoder.encoders.0.feed_forward.w_2.bias', 'transformers.lyric_encoder.encoders.0.norm_ff.weight', 'transformers.lyric_encoder.encoders.0.norm_ff.bias', 'transformers.lyric_encoder.encoders.0.norm_mha.weight', 'transformers.lyric_encoder.encoders.0.norm_mha.bias', 'transformers.lyric_encoder.encoders.1.self_attn.pos_bias_u', 'transformers.lyric_encoder.encoders.1.self_attn.pos_bias_v', 'transformers.lyric_encoder.encoders.1.self_attn.linear_q.weight', 'transformers.lyric_encoder.encoders.1.self_attn.linear_q.bias', 'transformers.lyric_encoder.encoders.1.self_attn.linear_k.weight', 'transformers.lyric_encoder.encoders.1.self_attn.linear_k.bias', 'transformers.lyric_encoder.encoders.1.self_attn.linear_v.weight', 'transformers.lyric_encoder.encoders.1.self_attn.linear_v.bias', 'transformers.lyric_encoder.encoders.1.self_attn.linear_out.weight', 'transformers.lyric_encoder.encoders.1.self_attn.linear_out.bias', 'transformers.lyric_encoder.encoders.1.self_attn.linear_pos.weight', 'transformers.lyric_encoder.encoders.1.feed_forward.w_1.weight', 'transformers.lyric_encoder.encoders.1.feed_forward.w_1.bias', 'transformers.lyric_encoder.encoders.1.feed_forward.w_2.weight', 'transformers.lyric_encoder.encoders.1.feed_forward.w_2.bias', 'transformers.lyric_encoder.encoders.1.norm_ff.weight', 'transformers.lyric_encoder.encoders.1.norm_ff.bias', 'transformers.lyric_encoder.encoders.1.norm_mha.weight', 'transformers.lyric_encoder.encoders.1.norm_mha.bias', 'transformers.lyric_encoder.encoders.2.self_attn.pos_bias_u', 'transformers.lyric_encoder.encoders.2.self_attn.pos_bias_v', 'transformers.lyric_encoder.encoders.2.self_attn.linear_q.weight', 'transformers.lyric_encoder.encoders.2.self_attn.linear_q.bias', 'transformers.lyric_encoder.encoders.2.self_attn.linear_k.weight', 'transformers.lyric_encoder.encoders.2.self_attn.linear_k.bias', 'transformers.lyric_encoder.encoders.2.self_attn.linear_v.weight', 'transformers.lyric_encoder.encoders.2.self_attn.linear_v.bias', 'transformers.lyric_encoder.encoders.2.self_attn.linear_out.weight', 'transformers.lyric_encoder.encoders.2.self_attn.linear_out.bias', 'transformers.lyric_encoder.encoders.2.self_attn.linear_pos.weight', 'transformers.lyric_encoder.encoders.2.feed_forward.w_1.weight', 'transformers.lyric_encoder.encoders.2.feed_forward.w_1.bias', 'transformers.lyric_encoder.encoders.2.feed_forward.w_2.weight', 'transformers.lyric_encoder.encoders.2.feed_forward.w_2.bias', 'transformers.lyric_encoder.encoders.2.norm_ff.weight', 'transformers.lyric_encoder.encoders.2.norm_ff.bias', 'transformers.lyric_encoder.encoders.2.norm_mha.weight', 'transformers.lyric_encoder.encoders.2.norm_mha.bias', 'transformers.lyric_encoder.encoders.3.self_attn.pos_bias_u', 'transformers.lyric_encoder.encoders.3.self_attn.pos_bias_v', 'transformers.lyric_encoder.encoders.3.self_attn.linear_q.weight', 'transformers.lyric_encoder.encoders.3.self_attn.linear_q.bias', 'transformers.lyric_encoder.encoders.3.self_attn.linear_k.weight', 'transformers.lyric_encoder.encoders.3.self_attn.linear_k.bias', 'transformers.lyric_encoder.encoders.3.self_attn.linear_v.weight', 'transformers.lyric_encoder.encoders.3.self_attn.linear_v.bias', 'transformers.lyric_encoder.encoders.3.self_attn.linear_out.weight', 'transformers.lyric_encoder.encoders.3.self_attn.linear_out.bias', 'transformers.lyric_encoder.encoders.3.self_attn.linear_pos.weight', 'transformers.lyric_encoder.encoders.3.feed_forward.w_1.weight', 'transformers.lyric_encoder.encoders.3.feed_forward.w_1.bias', 'transformers.lyric_encoder.encoders.3.feed_forward.w_2.weight', 'transformers.lyric_encoder.encoders.3.feed_forward.w_2.bias', 'transformers.lyric_encoder.encoders.3.norm_ff.weight', 'transformers.lyric_encoder.encoders.3.norm_ff.bias', 'transformers.lyric_encoder.encoders.3.norm_mha.weight', 'transformers.lyric_encoder.encoders.3.norm_mha.bias', 'transformers.lyric_encoder.encoders.4.self_attn.pos_bias_u', 'transformers.lyric_encoder.encoders.4.self_attn.pos_bias_v', 'transformers.lyric_encoder.encoders.4.self_attn.linear_q.weight', 'transformers.lyric_encoder.encoders.4.self_attn.linear_q.bias', 'transformers.lyric_encoder.encoders.4.self_attn.linear_k.weight', 'transformers.lyric_encoder.encoders.4.self_attn.linear_k.bias', 'transformers.lyric_encoder.encoders.4.self_attn.linear_v.weight', 'transformers.lyric_encoder.encoders.4.self_attn.linear_v.bias', 'transformers.lyric_encoder.encoders.4.self_attn.linear_out.weight', 'transformers.lyric_encoder.encoders.4.self_attn.linear_out.bias', 'transformers.lyric_encoder.encoders.4.self_attn.linear_pos.weight', 'transformers.lyric_encoder.encoders.4.feed_forward.w_1.weight', 'transformers.lyric_encoder.encoders.4.feed_forward.w_1.bias', 'transformers.lyric_encoder.encoders.4.feed_forward.w_2.weight', 'transformers.lyric_encoder.encoders.4.feed_forward.w_2.bias', 'transformers.lyric_encoder.encoders.4.norm_ff.weight', 'transformers.lyric_encoder.encoders.4.norm_ff.bias', 'transformers.lyric_encoder.encoders.4.norm_mha.weight', 'transformers.lyric_encoder.encoders.4.norm_mha.bias', 'transformers.lyric_encoder.encoders.5.self_attn.pos_bias_u', 'transformers.lyric_encoder.encoders.5.self_attn.pos_bias_v', 'transformers.lyric_encoder.encoders.5.self_attn.linear_q.weight', 'transformers.lyric_encoder.encoders.5.self_attn.linear_q.bias', 'transformers.lyric_encoder.encoders.5.self_attn.linear_k.weight', 'transformers.lyric_encoder.encoders.5.self_attn.linear_k.bias', 'transformers.lyric_encoder.encoders.5.self_attn.linear_v.weight', 'transformers.lyric_encoder.encoders.5.self_attn.linear_v.bias', 'transformers.lyric_encoder.encoders.5.self_attn.linear_out.weight', 'transformers.lyric_encoder.encoders.5.self_attn.linear_out.bias', 'transformers.lyric_encoder.encoders.5.self_attn.linear_pos.weight', 'transformers.lyric_encoder.encoders.5.feed_forward.w_1.weight', 'transformers.lyric_encoder.encoders.5.feed_forward.w_1.bias', 'transformers.lyric_encoder.encoders.5.feed_forward.w_2.weight', 'transformers.lyric_encoder.encoders.5.feed_forward.w_2.bias', 'transformers.lyric_encoder.encoders.5.norm_ff.weight', 'transformers.lyric_encoder.encoders.5.norm_ff.bias', 'transformers.lyric_encoder.encoders.5.norm_mha.weight', 'transformers.lyric_encoder.encoders.5.norm_mha.bias', 'transformers.lyric_proj.weight', 'transformers.lyric_proj.bias', 'transformers.projectors.0.0.weight', 'transformers.projectors.0.0.bias', 'transformers.projectors.0.2.weight', 'transformers.projectors.0.2.bias', 'transformers.projectors.0.4.weight', 'transformers.projectors.0.4.bias', 'transformers.projectors.1.0.weight', 'transformers.projectors.1.0.bias', 'transformers.projectors.1.2.weight', 'transformers.projectors.1.2.bias', 'transformers.projectors.1.4.weight', 'transformers.projectors.1.4.bias', 'transformers.proj_in.early_conv_layers.0.weight', 'transformers.proj_in.early_conv_layers.0.bias', 'transformers.proj_in.early_conv_layers.1.weight', 'transformers.proj_in.early_conv_layers.1.bias', 'transformers.proj_in.early_conv_layers.2.weight', 'transformers.proj_in.early_conv_layers.2.bias', 'transformers.final_layer.scale_shift_table', 'transformers.final_layer.linear.weight', 'transformers.final_layer.linear.bias', 'lyric_processor.lyric_text_model.encoder.embed_tokens.weight'])

In [7]:

transformers_checkpoint = {k.replace("transformers.", ""):v for k, v in checkpoint.items() if k.startswith("transformers.")}

In [1]:

from models.fusic_transformer import FusicTransformer2DModel

/root/miniconda3/envs/sag_train/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

In [3]:

import json

with open("models/config.json") as f:
    config = json.load(f)

In [8]:

fusic_model = FusicTransformer2DModel(**config)

In [9]:

fusic_model.load_state_dict(transformers_checkpoint, strict=True)

Out [9]:

<All keys matched successfully>

In [13]:

fusic_model = fusic_model.to(torch.bfloat16)

In [14]:

fusic_model.save_pretrained("checkpoints/fusic_transformer")

In [15]:

from diffusers import AutoencoderDC

DEFAULT_PRETRAINED_PATH = "./checkpoints/music_dcae_f8c8"
dcae = AutoencoderDC.from_pretrained(DEFAULT_PRETRAINED_PATH)

In [18]:

dcae = dcae.to(torch.bfloat16)

In [19]:

dcae.save_pretrained("checkpoints/music_dcae_f8c8")

In [1]:

from music_dcae.music_vocoder import ADaMoSHiFiGANV1

/root/miniconda3/envs/sag_train/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

In [2]:

vocoder = ADaMoSHiFiGANV1(checkpoint_path="/root/sag_train/checkpoints/music_vocoder.pt")

/root/miniconda3/envs/sag_train/lib/python3.10/site-packages/torch/nn/utils/weight_norm.py:143: FutureWarning: `torch.nn.utils.weight_norm` is deprecated in favor of `torch.nn.utils.parametrizations.weight_norm`.
  WeightNorm.apply(module, name, dim)
/vepfs-d-data/q-ace/repo/gongjunmin/ACEFlow3/music_dcae/music_vocoder.py:540: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
  ckpt_state = torch.load(checkpoint_path, map_location="cpu")

In [5]:

import torch

vocoder = vocoder.to(torch.bfloat16)

In [6]:

vocoder.save_pretrained("checkpoints/music_vocoder")

In [1]:

from music_dcae.music_dcae_pipeline import MusicDCAE

/root/miniconda3/envs/sag_train/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from .autonotebook import tqdm as notebook_tqdm

In [2]:

music_dcae = MusicDCAE.from_pretrained("./checkpoints/music_dcae_vocoder")

Some weights of the model checkpoint were not used when initializing MusicDCAE: 
 ['dcae.decoder.conv_in.bias, dcae.decoder.conv_in.weight, dcae.decoder.conv_out.bias, dcae.decoder.conv_out.weight, dcae.decoder.norm_out.bias, dcae.decoder.norm_out.weight, dcae.decoder.up_blocks.0.0.conv.bias, dcae.decoder.up_blocks.0.0.conv.weight, dcae.decoder.up_blocks.0.1.conv1.bias, dcae.decoder.up_blocks.0.1.conv1.weight, dcae.decoder.up_blocks.0.1.conv2.weight, dcae.decoder.up_blocks.0.1.norm.bias, dcae.decoder.up_blocks.0.1.norm.weight, dcae.decoder.up_blocks.0.2.conv1.bias, dcae.decoder.up_blocks.0.2.conv1.weight, dcae.decoder.up_blocks.0.2.conv2.weight, dcae.decoder.up_blocks.0.2.norm.bias, dcae.decoder.up_blocks.0.2.norm.weight, dcae.decoder.up_blocks.0.3.conv1.bias, dcae.decoder.up_blocks.0.3.conv1.weight, dcae.decoder.up_blocks.0.3.conv2.weight, dcae.decoder.up_blocks.0.3.norm.bias, dcae.decoder.up_blocks.0.3.norm.weight, dcae.decoder.up_blocks.1.0.conv.bias, dcae.decoder.up_blocks.1.0.conv.weight, dcae.decoder.up_blocks.1.1.conv1.bias, dcae.decoder.up_blocks.1.1.conv1.weight, dcae.decoder.up_blocks.1.1.conv2.weight, dcae.decoder.up_blocks.1.1.norm.bias, dcae.decoder.up_blocks.1.1.norm.weight, dcae.decoder.up_blocks.1.2.conv1.bias, dcae.decoder.up_blocks.1.2.conv1.weight, dcae.decoder.up_blocks.1.2.conv2.weight, dcae.decoder.up_blocks.1.2.norm.bias, dcae.decoder.up_blocks.1.2.norm.weight, dcae.decoder.up_blocks.1.3.conv1.bias, dcae.decoder.up_blocks.1.3.conv1.weight, dcae.decoder.up_blocks.1.3.conv2.weight, dcae.decoder.up_blocks.1.3.norm.bias, dcae.decoder.up_blocks.1.3.norm.weight, dcae.decoder.up_blocks.2.0.conv.bias, dcae.decoder.up_blocks.2.0.conv.weight, dcae.decoder.up_blocks.2.1.conv1.bias, dcae.decoder.up_blocks.2.1.conv1.weight, dcae.decoder.up_blocks.2.1.conv2.weight, dcae.decoder.up_blocks.2.1.norm.bias, dcae.decoder.up_blocks.2.1.norm.weight, dcae.decoder.up_blocks.2.2.conv1.bias, dcae.decoder.up_blocks.2.2.conv1.weight, dcae.decoder.up_blocks.2.2.conv2.weight, dcae.decoder.up_blocks.2.2.norm.bias, dcae.decoder.up_blocks.2.2.norm.weight, dcae.decoder.up_blocks.2.3.conv1.bias, dcae.decoder.up_blocks.2.3.conv1.weight, dcae.decoder.up_blocks.2.3.conv2.weight, dcae.decoder.up_blocks.2.3.norm.bias, dcae.decoder.up_blocks.2.3.norm.weight, dcae.decoder.up_blocks.3.0.attn.norm_out.bias, dcae.decoder.up_blocks.3.0.attn.norm_out.weight, dcae.decoder.up_blocks.3.0.attn.to_k.weight, dcae.decoder.up_blocks.3.0.attn.to_out.weight, dcae.decoder.up_blocks.3.0.attn.to_q.weight, dcae.decoder.up_blocks.3.0.attn.to_qkv_multiscale.0.proj_in.weight, dcae.decoder.up_blocks.3.0.attn.to_qkv_multiscale.0.proj_out.weight, dcae.decoder.up_blocks.3.0.attn.to_v.weight, dcae.decoder.up_blocks.3.0.conv_out.conv_depth.bias, dcae.decoder.up_blocks.3.0.conv_out.conv_depth.weight, dcae.decoder.up_blocks.3.0.conv_out.conv_inverted.bias, dcae.decoder.up_blocks.3.0.conv_out.conv_inverted.weight, dcae.decoder.up_blocks.3.0.conv_out.conv_point.weight, dcae.decoder.up_blocks.3.0.conv_out.norm.bias, dcae.decoder.up_blocks.3.0.conv_out.norm.weight, dcae.decoder.up_blocks.3.1.attn.norm_out.bias, dcae.decoder.up_blocks.3.1.attn.norm_out.weight, dcae.decoder.up_blocks.3.1.attn.to_k.weight, dcae.decoder.up_blocks.3.1.attn.to_out.weight, dcae.decoder.up_blocks.3.1.attn.to_q.weight, dcae.decoder.up_blocks.3.1.attn.to_qkv_multiscale.0.proj_in.weight, dcae.decoder.up_blocks.3.1.attn.to_qkv_multiscale.0.proj_out.weight, dcae.decoder.up_blocks.3.1.attn.to_v.weight, dcae.decoder.up_blocks.3.1.conv_out.conv_depth.bias, dcae.decoder.up_blocks.3.1.conv_out.conv_depth.weight, dcae.decoder.up_blocks.3.1.conv_out.conv_inverted.bias, dcae.decoder.up_blocks.3.1.conv_out.conv_inverted.weight, dcae.decoder.up_blocks.3.1.conv_out.conv_point.weight, dcae.decoder.up_blocks.3.1.conv_out.norm.bias, dcae.decoder.up_blocks.3.1.conv_out.norm.weight, dcae.decoder.up_blocks.3.2.attn.norm_out.bias, dcae.decoder.up_blocks.3.2.attn.norm_out.weight, dcae.decoder.up_blocks.3.2.attn.to_k.weight, dcae.decoder.up_blocks.3.2.attn.to_out.weight, dcae.decoder.up_blocks.3.2.attn.to_q.weight, dcae.decoder.up_blocks.3.2.attn.to_qkv_multiscale.0.proj_in.weight, dcae.decoder.up_blocks.3.2.attn.to_qkv_multiscale.0.proj_out.weight, dcae.decoder.up_blocks.3.2.attn.to_v.weight, dcae.decoder.up_blocks.3.2.conv_out.conv_depth.bias, dcae.decoder.up_blocks.3.2.conv_out.conv_depth.weight, dcae.decoder.up_blocks.3.2.conv_out.conv_inverted.bias, dcae.decoder.up_blocks.3.2.conv_out.conv_inverted.weight, dcae.decoder.up_blocks.3.2.conv_out.conv_point.weight, dcae.decoder.up_blocks.3.2.conv_out.norm.bias, dcae.decoder.up_blocks.3.2.conv_out.norm.weight, dcae.encoder.conv_in.bias, dcae.encoder.conv_in.weight, dcae.encoder.conv_out.bias, dcae.encoder.conv_out.weight, dcae.encoder.down_blocks.0.0.conv1.bias, dcae.encoder.down_blocks.0.0.conv1.weight, dcae.encoder.down_blocks.0.0.conv2.weight, dcae.encoder.down_blocks.0.0.norm.bias, dcae.encoder.down_blocks.0.0.norm.weight, dcae.encoder.down_blocks.0.1.conv1.bias, dcae.encoder.down_blocks.0.1.conv1.weight, dcae.encoder.down_blocks.0.1.conv2.weight, dcae.encoder.down_blocks.0.1.norm.bias, dcae.encoder.down_blocks.0.1.norm.weight, dcae.encoder.down_blocks.0.2.conv.bias, dcae.encoder.down_blocks.0.2.conv.weight, dcae.encoder.down_blocks.1.0.conv1.bias, dcae.encoder.down_blocks.1.0.conv1.weight, dcae.encoder.down_blocks.1.0.conv2.weight, dcae.encoder.down_blocks.1.0.norm.bias, dcae.encoder.down_blocks.1.0.norm.weight, dcae.encoder.down_blocks.1.1.conv1.bias, dcae.encoder.down_blocks.1.1.conv1.weight, dcae.encoder.down_blocks.1.1.conv2.weight, dcae.encoder.down_blocks.1.1.norm.bias, dcae.encoder.down_blocks.1.1.norm.weight, dcae.encoder.down_blocks.1.2.conv.bias, dcae.encoder.down_blocks.1.2.conv.weight, dcae.encoder.down_blocks.2.0.conv1.bias, dcae.encoder.down_blocks.2.0.conv1.weight, dcae.encoder.down_blocks.2.0.conv2.weight, dcae.encoder.down_blocks.2.0.norm.bias, dcae.encoder.down_blocks.2.0.norm.weight, dcae.encoder.down_blocks.2.1.conv1.bias, dcae.encoder.down_blocks.2.1.conv1.weight, dcae.encoder.down_blocks.2.1.conv2.weight, dcae.encoder.down_blocks.2.1.norm.bias, dcae.encoder.down_blocks.2.1.norm.weight, dcae.encoder.down_blocks.2.2.conv1.bias, dcae.encoder.down_blocks.2.2.conv1.weight, dcae.encoder.down_blocks.2.2.conv2.weight, dcae.encoder.down_blocks.2.2.norm.bias, dcae.encoder.down_blocks.2.2.norm.weight, dcae.encoder.down_blocks.2.3.conv.bias, dcae.encoder.down_blocks.2.3.conv.weight, dcae.encoder.down_blocks.3.0.attn.norm_out.bias, dcae.encoder.down_blocks.3.0.attn.norm_out.weight, dcae.encoder.down_blocks.3.0.attn.to_k.weight, dcae.encoder.down_blocks.3.0.attn.to_out.weight, dcae.encoder.down_blocks.3.0.attn.to_q.weight, dcae.encoder.down_blocks.3.0.attn.to_qkv_multiscale.0.proj_in.weight, dcae.encoder.down_blocks.3.0.attn.to_qkv_multiscale.0.proj_out.weight, dcae.encoder.down_blocks.3.0.attn.to_v.weight, dcae.encoder.down_blocks.3.0.conv_out.conv_depth.bias, dcae.encoder.down_blocks.3.0.conv_out.conv_depth.weight, dcae.encoder.down_blocks.3.0.conv_out.conv_inverted.bias, dcae.encoder.down_blocks.3.0.conv_out.conv_inverted.weight, dcae.encoder.down_blocks.3.0.conv_out.conv_point.weight, dcae.encoder.down_blocks.3.0.conv_out.norm.bias, dcae.encoder.down_blocks.3.0.conv_out.norm.weight, dcae.encoder.down_blocks.3.1.attn.norm_out.bias, dcae.encoder.down_blocks.3.1.attn.norm_out.weight, dcae.encoder.down_blocks.3.1.attn.to_k.weight, dcae.encoder.down_blocks.3.1.attn.to_out.weight, dcae.encoder.down_blocks.3.1.attn.to_q.weight, dcae.encoder.down_blocks.3.1.attn.to_qkv_multiscale.0.proj_in.weight, dcae.encoder.down_blocks.3.1.attn.to_qkv_multiscale.0.proj_out.weight, dcae.encoder.down_blocks.3.1.attn.to_v.weight, dcae.encoder.down_blocks.3.1.conv_out.conv_depth.bias, dcae.encoder.down_blocks.3.1.conv_out.conv_depth.weight, dcae.encoder.down_blocks.3.1.conv_out.conv_inverted.bias, dcae.encoder.down_blocks.3.1.conv_out.conv_inverted.weight, dcae.encoder.down_blocks.3.1.conv_out.conv_point.weight, dcae.encoder.down_blocks.3.1.conv_out.norm.bias, dcae.encoder.down_blocks.3.1.conv_out.norm.weight, dcae.encoder.down_blocks.3.2.attn.norm_out.bias, dcae.encoder.down_blocks.3.2.attn.norm_out.weight, dcae.encoder.down_blocks.3.2.attn.to_k.weight, dcae.encoder.down_blocks.3.2.attn.to_out.weight, dcae.encoder.down_blocks.3.2.attn.to_q.weight, dcae.encoder.down_blocks.3.2.attn.to_qkv_multiscale.0.proj_in.weight, dcae.encoder.down_blocks.3.2.attn.to_qkv_multiscale.0.proj_out.weight, dcae.encoder.down_blocks.3.2.attn.to_v.weight, dcae.encoder.down_blocks.3.2.conv_out.conv_depth.bias, dcae.encoder.down_blocks.3.2.conv_out.conv_depth.weight, dcae.encoder.down_blocks.3.2.conv_out.conv_inverted.bias, dcae.encoder.down_blocks.3.2.conv_out.conv_inverted.weight, dcae.encoder.down_blocks.3.2.conv_out.conv_point.weight, dcae.encoder.down_blocks.3.2.conv_out.norm.bias, dcae.encoder.down_blocks.3.2.conv_out.norm.weight, vocoder.backbone.channel_layers.0.0.bias, vocoder.backbone.channel_layers.0.0.weight, vocoder.backbone.channel_layers.0.1.bias, vocoder.backbone.channel_layers.0.1.weight, vocoder.backbone.channel_layers.1.0.bias, vocoder.backbone.channel_layers.1.0.weight, vocoder.backbone.channel_layers.1.1.bias, vocoder.backbone.channel_layers.1.1.weight, vocoder.backbone.channel_layers.2.0.bias, vocoder.backbone.channel_layers.2.0.weight, vocoder.backbone.channel_layers.2.1.bias, vocoder.backbone.channel_layers.2.1.weight, vocoder.backbone.channel_layers.3.0.bias, vocoder.backbone.channel_layers.3.0.weight, vocoder.backbone.channel_layers.3.1.bias, vocoder.backbone.channel_layers.3.1.weight, vocoder.backbone.norm.bias, vocoder.backbone.norm.weight, vocoder.backbone.stages.0.0.dwconv.bias, vocoder.backbone.stages.0.0.dwconv.weight, vocoder.backbone.stages.0.0.gamma, vocoder.backbone.stages.0.0.norm.bias, vocoder.backbone.stages.0.0.norm.weight, vocoder.backbone.stages.0.0.pwconv1.bias, vocoder.backbone.stages.0.0.pwconv1.weight, vocoder.backbone.stages.0.0.pwconv2.bias, vocoder.backbone.stages.0.0.pwconv2.weight, vocoder.backbone.stages.0.1.dwconv.bias, vocoder.backbone.stages.0.1.dwconv.weight, vocoder.backbone.stages.0.1.gamma, vocoder.backbone.stages.0.1.norm.bias, vocoder.backbone.stages.0.1.norm.weight, vocoder.backbone.stages.0.1.pwconv1.bias, vocoder.backbone.stages.0.1.pwconv1.weight, vocoder.backbone.stages.0.1.pwconv2.bias, vocoder.backbone.stages.0.1.pwconv2.weight, vocoder.backbone.stages.0.2.dwconv.bias, vocoder.backbone.stages.0.2.dwconv.weight, vocoder.backbone.stages.0.2.gamma, vocoder.backbone.stages.0.2.norm.bias, vocoder.backbone.stages.0.2.norm.weight, vocoder.backbone.stages.0.2.pwconv1.bias, vocoder.backbone.stages.0.2.pwconv1.weight, vocoder.backbone.stages.0.2.pwconv2.bias, vocoder.backbone.stages.0.2.pwconv2.weight, vocoder.backbone.stages.1.0.dwconv.bias, vocoder.backbone.stages.1.0.dwconv.weight, vocoder.backbone.stages.1.0.gamma, vocoder.backbone.stages.1.0.norm.bias, vocoder.backbone.stages.1.0.norm.weight, vocoder.backbone.stages.1.0.pwconv1.bias, vocoder.backbone.stages.1.0.pwconv1.weight, vocoder.backbone.stages.1.0.pwconv2.bias, vocoder.backbone.stages.1.0.pwconv2.weight, vocoder.backbone.stages.1.1.dwconv.bias, vocoder.backbone.stages.1.1.dwconv.weight, vocoder.backbone.stages.1.1.gamma, vocoder.backbone.stages.1.1.norm.bias, vocoder.backbone.stages.1.1.norm.weight, vocoder.backbone.stages.1.1.pwconv1.bias, vocoder.backbone.stages.1.1.pwconv1.weight, vocoder.backbone.stages.1.1.pwconv2.bias, vocoder.backbone.stages.1.1.pwconv2.weight, vocoder.backbone.stages.1.2.dwconv.bias, vocoder.backbone.stages.1.2.dwconv.weight, vocoder.backbone.stages.1.2.gamma, vocoder.backbone.stages.1.2.norm.bias, vocoder.backbone.stages.1.2.norm.weight, vocoder.backbone.stages.1.2.pwconv1.bias, vocoder.backbone.stages.1.2.pwconv1.weight, vocoder.backbone.stages.1.2.pwconv2.bias, vocoder.backbone.stages.1.2.pwconv2.weight, vocoder.backbone.stages.2.0.dwconv.bias, vocoder.backbone.stages.2.0.dwconv.weight, vocoder.backbone.stages.2.0.gamma, vocoder.backbone.stages.2.0.norm.bias, vocoder.backbone.stages.2.0.norm.weight, vocoder.backbone.stages.2.0.pwconv1.bias, vocoder.backbone.stages.2.0.pwconv1.weight, vocoder.backbone.stages.2.0.pwconv2.bias, vocoder.backbone.stages.2.0.pwconv2.weight, vocoder.backbone.stages.2.1.dwconv.bias, vocoder.backbone.stages.2.1.dwconv.weight, vocoder.backbone.stages.2.1.gamma, vocoder.backbone.stages.2.1.norm.bias, vocoder.backbone.stages.2.1.norm.weight, vocoder.backbone.stages.2.1.pwconv1.bias, vocoder.backbone.stages.2.1.pwconv1.weight, vocoder.backbone.stages.2.1.pwconv2.bias, vocoder.backbone.stages.2.1.pwconv2.weight, vocoder.backbone.stages.2.2.dwconv.bias, vocoder.backbone.stages.2.2.dwconv.weight, vocoder.backbone.stages.2.2.gamma, vocoder.backbone.stages.2.2.norm.bias, vocoder.backbone.stages.2.2.norm.weight, vocoder.backbone.stages.2.2.pwconv1.bias, vocoder.backbone.stages.2.2.pwconv1.weight, vocoder.backbone.stages.2.2.pwconv2.bias, vocoder.backbone.stages.2.2.pwconv2.weight, vocoder.backbone.stages.2.3.dwconv.bias, vocoder.backbone.stages.2.3.dwconv.weight, vocoder.backbone.stages.2.3.gamma, vocoder.backbone.stages.2.3.norm.bias, vocoder.backbone.stages.2.3.norm.weight, vocoder.backbone.stages.2.3.pwconv1.bias, vocoder.backbone.stages.2.3.pwconv1.weight, vocoder.backbone.stages.2.3.pwconv2.bias, vocoder.backbone.stages.2.3.pwconv2.weight, vocoder.backbone.stages.2.4.dwconv.bias, vocoder.backbone.stages.2.4.dwconv.weight, vocoder.backbone.stages.2.4.gamma, vocoder.backbone.stages.2.4.norm.bias, vocoder.backbone.stages.2.4.norm.weight, vocoder.backbone.stages.2.4.pwconv1.bias, vocoder.backbone.stages.2.4.pwconv1.weight, vocoder.backbone.stages.2.4.pwconv2.bias, vocoder.backbone.stages.2.4.pwconv2.weight, vocoder.backbone.stages.2.5.dwconv.bias, vocoder.backbone.stages.2.5.dwconv.weight, vocoder.backbone.stages.2.5.gamma, vocoder.backbone.stages.2.5.norm.bias, vocoder.backbone.stages.2.5.norm.weight, vocoder.backbone.stages.2.5.pwconv1.bias, vocoder.backbone.stages.2.5.pwconv1.weight, vocoder.backbone.stages.2.5.pwconv2.bias, vocoder.backbone.stages.2.5.pwconv2.weight, vocoder.backbone.stages.2.6.dwconv.bias, vocoder.backbone.stages.2.6.dwconv.weight, vocoder.backbone.stages.2.6.gamma, vocoder.backbone.stages.2.6.norm.bias, vocoder.backbone.stages.2.6.norm.weight, vocoder.backbone.stages.2.6.pwconv1.bias, vocoder.backbone.stages.2.6.pwconv1.weight, vocoder.backbone.stages.2.6.pwconv2.bias, vocoder.backbone.stages.2.6.pwconv2.weight, vocoder.backbone.stages.2.7.dwconv.bias, vocoder.backbone.stages.2.7.dwconv.weight, vocoder.backbone.stages.2.7.gamma, vocoder.backbone.stages.2.7.norm.bias, vocoder.backbone.stages.2.7.norm.weight, vocoder.backbone.stages.2.7.pwconv1.bias, vocoder.backbone.stages.2.7.pwconv1.weight, vocoder.backbone.stages.2.7.pwconv2.bias, vocoder.backbone.stages.2.7.pwconv2.weight, vocoder.backbone.stages.2.8.dwconv.bias, vocoder.backbone.stages.2.8.dwconv.weight, vocoder.backbone.stages.2.8.gamma, vocoder.backbone.stages.2.8.norm.bias, vocoder.backbone.stages.2.8.norm.weight, vocoder.backbone.stages.2.8.pwconv1.bias, vocoder.backbone.stages.2.8.pwconv1.weight, vocoder.backbone.stages.2.8.pwconv2.bias, vocoder.backbone.stages.2.8.pwconv2.weight, vocoder.backbone.stages.3.0.dwconv.bias, vocoder.backbone.stages.3.0.dwconv.weight, vocoder.backbone.stages.3.0.gamma, vocoder.backbone.stages.3.0.norm.bias, vocoder.backbone.stages.3.0.norm.weight, vocoder.backbone.stages.3.0.pwconv1.bias, vocoder.backbone.stages.3.0.pwconv1.weight, vocoder.backbone.stages.3.0.pwconv2.bias, vocoder.backbone.stages.3.0.pwconv2.weight, vocoder.backbone.stages.3.1.dwconv.bias, vocoder.backbone.stages.3.1.dwconv.weight, vocoder.backbone.stages.3.1.gamma, vocoder.backbone.stages.3.1.norm.bias, vocoder.backbone.stages.3.1.norm.weight, vocoder.backbone.stages.3.1.pwconv1.bias, vocoder.backbone.stages.3.1.pwconv1.weight, vocoder.backbone.stages.3.1.pwconv2.bias, vocoder.backbone.stages.3.1.pwconv2.weight, vocoder.backbone.stages.3.2.dwconv.bias, vocoder.backbone.stages.3.2.dwconv.weight, vocoder.backbone.stages.3.2.gamma, vocoder.backbone.stages.3.2.norm.bias, vocoder.backbone.stages.3.2.norm.weight, vocoder.backbone.stages.3.2.pwconv1.bias, vocoder.backbone.stages.3.2.pwconv1.weight, vocoder.backbone.stages.3.2.pwconv2.bias, vocoder.backbone.stages.3.2.pwconv2.weight, vocoder.head.conv_post.bias, vocoder.head.conv_post.weight_g, vocoder.head.conv_post.weight_v, vocoder.head.conv_pre.bias, vocoder.head.conv_pre.weight_g, vocoder.head.conv_pre.weight_v, vocoder.head.resblocks.0.convs1.0.bias, vocoder.head.resblocks.0.convs1.0.weight_g, vocoder.head.resblocks.0.convs1.0.weight_v, vocoder.head.resblocks.0.convs1.1.bias, vocoder.head.resblocks.0.convs1.1.weight_g, vocoder.head.resblocks.0.convs1.1.weight_v, vocoder.head.resblocks.0.convs1.2.bias, vocoder.head.resblocks.0.convs1.2.weight_g, vocoder.head.resblocks.0.convs1.2.weight_v, vocoder.head.resblocks.0.convs2.0.bias, vocoder.head.resblocks.0.convs2.0.weight_g, vocoder.head.resblocks.0.convs2.0.weight_v, vocoder.head.resblocks.0.convs2.1.bias, vocoder.head.resblocks.0.convs2.1.weight_g, vocoder.head.resblocks.0.convs2.1.weight_v, vocoder.head.resblocks.0.convs2.2.bias, vocoder.head.resblocks.0.convs2.2.weight_g, vocoder.head.resblocks.0.convs2.2.weight_v, vocoder.head.resblocks.1.convs1.0.bias, vocoder.head.resblocks.1.convs1.0.weight_g, vocoder.head.resblocks.1.convs1.0.weight_v, vocoder.head.resblocks.1.convs1.1.bias, vocoder.head.resblocks.1.convs1.1.weight_g, vocoder.head.resblocks.1.convs1.1.weight_v, vocoder.head.resblocks.1.convs1.2.bias, vocoder.head.resblocks.1.convs1.2.weight_g, vocoder.head.resblocks.1.convs1.2.weight_v, vocoder.head.resblocks.1.convs2.0.bias, vocoder.head.resblocks.1.convs2.0.weight_g, vocoder.head.resblocks.1.convs2.0.weight_v, vocoder.head.resblocks.1.convs2.1.bias, vocoder.head.resblocks.1.convs2.1.weight_g, vocoder.head.resblocks.1.convs2.1.weight_v, vocoder.head.resblocks.1.convs2.2.bias, vocoder.head.resblocks.1.convs2.2.weight_g, vocoder.head.resblocks.1.convs2.2.weight_v, vocoder.head.resblocks.10.convs1.0.bias, vocoder.head.resblocks.10.convs1.0.weight_g, vocoder.head.resblocks.10.convs1.0.weight_v, vocoder.head.resblocks.10.convs1.1.bias, vocoder.head.resblocks.10.convs1.1.weight_g, vocoder.head.resblocks.10.convs1.1.weight_v, vocoder.head.resblocks.10.convs1.2.bias, vocoder.head.resblocks.10.convs1.2.weight_g, vocoder.head.resblocks.10.convs1.2.weight_v, vocoder.head.resblocks.10.convs2.0.bias, vocoder.head.resblocks.10.convs2.0.weight_g, vocoder.head.resblocks.10.convs2.0.weight_v, vocoder.head.resblocks.10.convs2.1.bias, vocoder.head.resblocks.10.convs2.1.weight_g, vocoder.head.resblocks.10.convs2.1.weight_v, vocoder.head.resblocks.10.convs2.2.bias, vocoder.head.resblocks.10.convs2.2.weight_g, vocoder.head.resblocks.10.convs2.2.weight_v, vocoder.head.resblocks.11.convs1.0.bias, vocoder.head.resblocks.11.convs1.0.weight_g, vocoder.head.resblocks.11.convs1.0.weight_v, vocoder.head.resblocks.11.convs1.1.bias, vocoder.head.resblocks.11.convs1.1.weight_g, vocoder.head.resblocks.11.convs1.1.weight_v, vocoder.head.resblocks.11.convs1.2.bias, vocoder.head.resblocks.11.convs1.2.weight_g, vocoder.head.resblocks.11.convs1.2.weight_v, vocoder.head.resblocks.11.convs2.0.bias, vocoder.head.resblocks.11.convs2.0.weight_g, vocoder.head.resblocks.11.convs2.0.weight_v, vocoder.head.resblocks.11.convs2.1.bias, vocoder.head.resblocks.11.convs2.1.weight_g, vocoder.head.resblocks.11.convs2.1.weight_v, vocoder.head.resblocks.11.convs2.2.bias, vocoder.head.resblocks.11.convs2.2.weight_g, vocoder.head.resblocks.11.convs2.2.weight_v, vocoder.head.resblocks.12.convs1.0.bias, vocoder.head.resblocks.12.convs1.0.weight_g, vocoder.head.resblocks.12.convs1.0.weight_v, vocoder.head.resblocks.12.convs1.1.bias, vocoder.head.resblocks.12.convs1.1.weight_g, vocoder.head.resblocks.12.convs1.1.weight_v, vocoder.head.resblocks.12.convs1.2.bias, vocoder.head.resblocks.12.convs1.2.weight_g, vocoder.head.resblocks.12.convs1.2.weight_v, vocoder.head.resblocks.12.convs2.0.bias, vocoder.head.resblocks.12.convs2.0.weight_g, vocoder.head.resblocks.12.convs2.0.weight_v, vocoder.head.resblocks.12.convs2.1.bias, vocoder.head.resblocks.12.convs2.1.weight_g, vocoder.head.resblocks.12.convs2.1.weight_v, vocoder.head.resblocks.12.convs2.2.bias, vocoder.head.resblocks.12.convs2.2.weight_g, vocoder.head.resblocks.12.convs2.2.weight_v, vocoder.head.resblocks.13.convs1.0.bias, vocoder.head.resblocks.13.convs1.0.weight_g, vocoder.head.resblocks.13.convs1.0.weight_v, vocoder.head.resblocks.13.convs1.1.bias, vocoder.head.resblocks.13.convs1.1.weight_g, vocoder.head.resblocks.13.convs1.1.weight_v, vocoder.head.resblocks.13.convs1.2.bias, vocoder.head.resblocks.13.convs1.2.weight_g, vocoder.head.resblocks.13.convs1.2.weight_v, vocoder.head.resblocks.13.convs2.0.bias, vocoder.head.resblocks.13.convs2.0.weight_g, vocoder.head.resblocks.13.convs2.0.weight_v, vocoder.head.resblocks.13.convs2.1.bias, vocoder.head.resblocks.13.convs2.1.weight_g, vocoder.head.resblocks.13.convs2.1.weight_v, vocoder.head.resblocks.13.convs2.2.bias, vocoder.head.resblocks.13.convs2.2.weight_g, vocoder.head.resblocks.13.convs2.2.weight_v, vocoder.head.resblocks.14.convs1.0.bias, vocoder.head.resblocks.14.convs1.0.weight_g, vocoder.head.resblocks.14.convs1.0.weight_v, vocoder.head.resblocks.14.convs1.1.bias, vocoder.head.resblocks.14.convs1.1.weight_g, vocoder.head.resblocks.14.convs1.1.weight_v, vocoder.head.resblocks.14.convs1.2.bias, vocoder.head.resblocks.14.convs1.2.weight_g, vocoder.head.resblocks.14.convs1.2.weight_v, vocoder.head.resblocks.14.convs2.0.bias, vocoder.head.resblocks.14.convs2.0.weight_g, vocoder.head.resblocks.14.convs2.0.weight_v, vocoder.head.resblocks.14.convs2.1.bias, vocoder.head.resblocks.14.convs2.1.weight_g, vocoder.head.resblocks.14.convs2.1.weight_v, vocoder.head.resblocks.14.convs2.2.bias, vocoder.head.resblocks.14.convs2.2.weight_g, vocoder.head.resblocks.14.convs2.2.weight_v, vocoder.head.resblocks.15.convs1.0.bias, vocoder.head.resblocks.15.convs1.0.weight_g, vocoder.head.resblocks.15.convs1.0.weight_v, vocoder.head.resblocks.15.convs1.1.bias, vocoder.head.resblocks.15.convs1.1.weight_g, vocoder.head.resblocks.15.convs1.1.weight_v, vocoder.head.resblocks.15.convs1.2.bias, vocoder.head.resblocks.15.convs1.2.weight_g, vocoder.head.resblocks.15.convs1.2.weight_v, vocoder.head.resblocks.15.convs2.0.bias, vocoder.head.resblocks.15.convs2.0.weight_g, vocoder.head.resblocks.15.convs2.0.weight_v, vocoder.head.resblocks.15.convs2.1.bias, vocoder.head.resblocks.15.convs2.1.weight_g, vocoder.head.resblocks.15.convs2.1.weight_v, vocoder.head.resblocks.15.convs2.2.bias, vocoder.head.resblocks.15.convs2.2.weight_g, vocoder.head.resblocks.15.convs2.2.weight_v, vocoder.head.resblocks.16.convs1.0.bias, vocoder.head.resblocks.16.convs1.0.weight_g, vocoder.head.resblocks.16.convs1.0.weight_v, vocoder.head.resblocks.16.convs1.1.bias, vocoder.head.resblocks.16.convs1.1.weight_g, vocoder.head.resblocks.16.convs1.1.weight_v, vocoder.head.resblocks.16.convs1.2.bias, vocoder.head.resblocks.16.convs1.2.weight_g, vocoder.head.resblocks.16.convs1.2.weight_v, vocoder.head.resblocks.16.convs2.0.bias, vocoder.head.resblocks.16.convs2.0.weight_g, vocoder.head.resblocks.16.convs2.0.weight_v, vocoder.head.resblocks.16.convs2.1.bias, vocoder.head.resblocks.16.convs2.1.weight_g, vocoder.head.resblocks.16.convs2.1.weight_v, vocoder.head.resblocks.16.convs2.2.bias, vocoder.head.resblocks.16.convs2.2.weight_g, vocoder.head.resblocks.16.convs2.2.weight_v, vocoder.head.resblocks.17.convs1.0.bias, vocoder.head.resblocks.17.convs1.0.weight_g, vocoder.head.resblocks.17.convs1.0.weight_v, vocoder.head.resblocks.17.convs1.1.bias, vocoder.head.resblocks.17.convs1.1.weight_g, vocoder.head.resblocks.17.convs1.1.weight_v, vocoder.head.resblocks.17.convs1.2.bias, vocoder.head.resblocks.17.convs1.2.weight_g, vocoder.head.resblocks.17.convs1.2.weight_v, vocoder.head.resblocks.17.convs2.0.bias, vocoder.head.resblocks.17.convs2.0.weight_g, vocoder.head.resblocks.17.convs2.0.weight_v, vocoder.head.resblocks.17.convs2.1.bias, vocoder.head.resblocks.17.convs2.1.weight_g, vocoder.head.resblocks.17.convs2.1.weight_v, vocoder.head.resblocks.17.convs2.2.bias, vocoder.head.resblocks.17.convs2.2.weight_g, vocoder.head.resblocks.17.convs2.2.weight_v, vocoder.head.resblocks.18.convs1.0.bias, vocoder.head.resblocks.18.convs1.0.weight_g, vocoder.head.resblocks.18.convs1.0.weight_v, vocoder.head.resblocks.18.convs1.1.bias, vocoder.head.resblocks.18.convs1.1.weight_g, vocoder.head.resblocks.18.convs1.1.weight_v, vocoder.head.resblocks.18.convs1.2.bias, vocoder.head.resblocks.18.convs1.2.weight_g, vocoder.head.resblocks.18.convs1.2.weight_v, vocoder.head.resblocks.18.convs2.0.bias, vocoder.head.resblocks.18.convs2.0.weight_g, vocoder.head.resblocks.18.convs2.0.weight_v, vocoder.head.resblocks.18.convs2.1.bias, vocoder.head.resblocks.18.convs2.1.weight_g, vocoder.head.resblocks.18.convs2.1.weight_v, vocoder.head.resblocks.18.convs2.2.bias, vocoder.head.resblocks.18.convs2.2.weight_g, vocoder.head.resblocks.18.convs2.2.weight_v, vocoder.head.resblocks.19.convs1.0.bias, vocoder.head.resblocks.19.convs1.0.weight_g, vocoder.head.resblocks.19.convs1.0.weight_v, vocoder.head.resblocks.19.convs1.1.bias, vocoder.head.resblocks.19.convs1.1.weight_g, vocoder.head.resblocks.19.convs1.1.weight_v, vocoder.head.resblocks.19.convs1.2.bias, vocoder.head.resblocks.19.convs1.2.weight_g, vocoder.head.resblocks.19.convs1.2.weight_v, vocoder.head.resblocks.19.convs2.0.bias, vocoder.head.resblocks.19.convs2.0.weight_g, vocoder.head.resblocks.19.convs2.0.weight_v, vocoder.head.resblocks.19.convs2.1.bias, vocoder.head.resblocks.19.convs2.1.weight_g, vocoder.head.resblocks.19.convs2.1.weight_v, vocoder.head.resblocks.19.convs2.2.bias, vocoder.head.resblocks.19.convs2.2.weight_g, vocoder.head.resblocks.19.convs2.2.weight_v, vocoder.head.resblocks.2.convs1.0.bias, vocoder.head.resblocks.2.convs1.0.weight_g, vocoder.head.resblocks.2.convs1.0.weight_v, vocoder.head.resblocks.2.convs1.1.bias, vocoder.head.resblocks.2.convs1.1.weight_g, vocoder.head.resblocks.2.convs1.1.weight_v, vocoder.head.resblocks.2.convs1.2.bias, vocoder.head.resblocks.2.convs1.2.weight_g, vocoder.head.resblocks.2.convs1.2.weight_v, vocoder.head.resblocks.2.convs2.0.bias, vocoder.head.resblocks.2.convs2.0.weight_g, vocoder.head.resblocks.2.convs2.0.weight_v, vocoder.head.resblocks.2.convs2.1.bias, vocoder.head.resblocks.2.convs2.1.weight_g, vocoder.head.resblocks.2.convs2.1.weight_v, vocoder.head.resblocks.2.convs2.2.bias, vocoder.head.resblocks.2.convs2.2.weight_g, vocoder.head.resblocks.2.convs2.2.weight_v, vocoder.head.resblocks.20.convs1.0.bias, vocoder.head.resblocks.20.convs1.0.weight_g, vocoder.head.resblocks.20.convs1.0.weight_v, vocoder.head.resblocks.20.convs1.1.bias, vocoder.head.resblocks.20.convs1.1.weight_g, vocoder.head.resblocks.20.convs1.1.weight_v, vocoder.head.resblocks.20.convs1.2.bias, vocoder.head.resblocks.20.convs1.2.weight_g, vocoder.head.resblocks.20.convs1.2.weight_v, vocoder.head.resblocks.20.convs2.0.bias, vocoder.head.resblocks.20.convs2.0.weight_g, vocoder.head.resblocks.20.convs2.0.weight_v, vocoder.head.resblocks.20.convs2.1.bias, vocoder.head.resblocks.20.convs2.1.weight_g, vocoder.head.resblocks.20.convs2.1.weight_v, vocoder.head.resblocks.20.convs2.2.bias, vocoder.head.resblocks.20.convs2.2.weight_g, vocoder.head.resblocks.20.convs2.2.weight_v, vocoder.head.resblocks.21.convs1.0.bias, vocoder.head.resblocks.21.convs1.0.weight_g, vocoder.head.resblocks.21.convs1.0.weight_v, vocoder.head.resblocks.21.convs1.1.bias, vocoder.head.resblocks.21.convs1.1.weight_g, vocoder.head.resblocks.21.convs1.1.weight_v, vocoder.head.resblocks.21.convs1.2.bias, vocoder.head.resblocks.21.convs1.2.weight_g, vocoder.head.resblocks.21.convs1.2.weight_v, vocoder.head.resblocks.21.convs2.0.bias, vocoder.head.resblocks.21.convs2.0.weight_g, vocoder.head.resblocks.21.convs2.0.weight_v, vocoder.head.resblocks.21.convs2.1.bias, vocoder.head.resblocks.21.convs2.1.weight_g, vocoder.head.resblocks.21.convs2.1.weight_v, vocoder.head.resblocks.21.convs2.2.bias, vocoder.head.resblocks.21.convs2.2.weight_g, vocoder.head.resblocks.21.convs2.2.weight_v, vocoder.head.resblocks.22.convs1.0.bias, vocoder.head.resblocks.22.convs1.0.weight_g, vocoder.head.resblocks.22.convs1.0.weight_v, vocoder.head.resblocks.22.convs1.1.bias, vocoder.head.resblocks.22.convs1.1.weight_g, vocoder.head.resblocks.22.convs1.1.weight_v, vocoder.head.resblocks.22.convs1.2.bias, vocoder.head.resblocks.22.convs1.2.weight_g, vocoder.head.resblocks.22.convs1.2.weight_v, vocoder.head.resblocks.22.convs2.0.bias, vocoder.head.resblocks.22.convs2.0.weight_g, vocoder.head.resblocks.22.convs2.0.weight_v, vocoder.head.resblocks.22.convs2.1.bias, vocoder.head.resblocks.22.convs2.1.weight_g, vocoder.head.resblocks.22.convs2.1.weight_v, vocoder.head.resblocks.22.convs2.2.bias, vocoder.head.resblocks.22.convs2.2.weight_g, vocoder.head.resblocks.22.convs2.2.weight_v, vocoder.head.resblocks.23.convs1.0.bias, vocoder.head.resblocks.23.convs1.0.weight_g, vocoder.head.resblocks.23.convs1.0.weight_v, vocoder.head.resblocks.23.convs1.1.bias, vocoder.head.resblocks.23.convs1.1.weight_g, vocoder.head.resblocks.23.convs1.1.weight_v, vocoder.head.resblocks.23.convs1.2.bias, vocoder.head.resblocks.23.convs1.2.weight_g, vocoder.head.resblocks.23.convs1.2.weight_v, vocoder.head.resblocks.23.convs2.0.bias, vocoder.head.resblocks.23.convs2.0.weight_g, vocoder.head.resblocks.23.convs2.0.weight_v, vocoder.head.resblocks.23.convs2.1.bias, vocoder.head.resblocks.23.convs2.1.weight_g, vocoder.head.resblocks.23.convs2.1.weight_v, vocoder.head.resblocks.23.convs2.2.bias, vocoder.head.resblocks.23.convs2.2.weight_g, vocoder.head.resblocks.23.convs2.2.weight_v, vocoder.head.resblocks.24.convs1.0.bias, vocoder.head.resblocks.24.convs1.0.weight_g, vocoder.head.resblocks.24.convs1.0.weight_v, vocoder.head.resblocks.24.convs1.1.bias, vocoder.head.resblocks.24.convs1.1.weight_g, vocoder.head.resblocks.24.convs1.1.weight_v, vocoder.head.resblocks.24.convs1.2.bias, vocoder.head.resblocks.24.convs1.2.weight_g, vocoder.head.resblocks.24.convs1.2.weight_v, vocoder.head.resblocks.24.convs2.0.bias, vocoder.head.resblocks.24.convs2.0.weight_g, vocoder.head.resblocks.24.convs2.0.weight_v, vocoder.head.resblocks.24.convs2.1.bias, vocoder.head.resblocks.24.convs2.1.weight_g, vocoder.head.resblocks.24.convs2.1.weight_v, vocoder.head.resblocks.24.convs2.2.bias, vocoder.head.resblocks.24.convs2.2.weight_g, vocoder.head.resblocks.24.convs2.2.weight_v, vocoder.head.resblocks.25.convs1.0.bias, vocoder.head.resblocks.25.convs1.0.weight_g, vocoder.head.resblocks.25.convs1.0.weight_v, vocoder.head.resblocks.25.convs1.1.bias, vocoder.head.resblocks.25.convs1.1.weight_g, vocoder.head.resblocks.25.convs1.1.weight_v, vocoder.head.resblocks.25.convs1.2.bias, vocoder.head.resblocks.25.convs1.2.weight_g, vocoder.head.resblocks.25.convs1.2.weight_v, vocoder.head.resblocks.25.convs2.0.bias, vocoder.head.resblocks.25.convs2.0.weight_g, vocoder.head.resblocks.25.convs2.0.weight_v, vocoder.head.resblocks.25.convs2.1.bias, vocoder.head.resblocks.25.convs2.1.weight_g, vocoder.head.resblocks.25.convs2.1.weight_v, vocoder.head.resblocks.25.convs2.2.bias, vocoder.head.resblocks.25.convs2.2.weight_g, vocoder.head.resblocks.25.convs2.2.weight_v, vocoder.head.resblocks.26.convs1.0.bias, vocoder.head.resblocks.26.convs1.0.weight_g, vocoder.head.resblocks.26.convs1.0.weight_v, vocoder.head.resblocks.26.convs1.1.bias, vocoder.head.resblocks.26.convs1.1.weight_g, vocoder.head.resblocks.26.convs1.1.weight_v, vocoder.head.resblocks.26.convs1.2.bias, vocoder.head.resblocks.26.convs1.2.weight_g, vocoder.head.resblocks.26.convs1.2.weight_v, vocoder.head.resblocks.26.convs2.0.bias, vocoder.head.resblocks.26.convs2.0.weight_g, vocoder.head.resblocks.26.convs2.0.weight_v, vocoder.head.resblocks.26.convs2.1.bias, vocoder.head.resblocks.26.convs2.1.weight_g, vocoder.head.resblocks.26.convs2.1.weight_v, vocoder.head.resblocks.26.convs2.2.bias, vocoder.head.resblocks.26.convs2.2.weight_g, vocoder.head.resblocks.26.convs2.2.weight_v, vocoder.head.resblocks.27.convs1.0.bias, vocoder.head.resblocks.27.convs1.0.weight_g, vocoder.head.resblocks.27.convs1.0.weight_v, vocoder.head.resblocks.27.convs1.1.bias, vocoder.head.resblocks.27.convs1.1.weight_g, vocoder.head.resblocks.27.convs1.1.weight_v, vocoder.head.resblocks.27.convs1.2.bias, vocoder.head.resblocks.27.convs1.2.weight_g, vocoder.head.resblocks.27.convs1.2.weight_v, vocoder.head.resblocks.27.convs2.0.bias, vocoder.head.resblocks.27.convs2.0.weight_g, vocoder.head.resblocks.27.convs2.0.weight_v, vocoder.head.resblocks.27.convs2.1.bias, vocoder.head.resblocks.27.convs2.1.weight_g, vocoder.head.resblocks.27.convs2.1.weight_v, vocoder.head.resblocks.27.convs2.2.bias, vocoder.head.resblocks.27.convs2.2.weight_g, vocoder.head.resblocks.27.convs2.2.weight_v, vocoder.head.resblocks.3.convs1.0.bias, vocoder.head.resblocks.3.convs1.0.weight_g, vocoder.head.resblocks.3.convs1.0.weight_v, vocoder.head.resblocks.3.convs1.1.bias, vocoder.head.resblocks.3.convs1.1.weight_g, vocoder.head.resblocks.3.convs1.1.weight_v, vocoder.head.resblocks.3.convs1.2.bias, vocoder.head.resblocks.3.convs1.2.weight_g, vocoder.head.resblocks.3.convs1.2.weight_v, vocoder.head.resblocks.3.convs2.0.bias, vocoder.head.resblocks.3.convs2.0.weight_g, vocoder.head.resblocks.3.convs2.0.weight_v, vocoder.head.resblocks.3.convs2.1.bias, vocoder.head.resblocks.3.convs2.1.weight_g, vocoder.head.resblocks.3.convs2.1.weight_v, vocoder.head.resblocks.3.convs2.2.bias, vocoder.head.resblocks.3.convs2.2.weight_g, vocoder.head.resblocks.3.convs2.2.weight_v, vocoder.head.resblocks.4.convs1.0.bias, vocoder.head.resblocks.4.convs1.0.weight_g, vocoder.head.resblocks.4.convs1.0.weight_v, vocoder.head.resblocks.4.convs1.1.bias, vocoder.head.resblocks.4.convs1.1.weight_g, vocoder.head.resblocks.4.convs1.1.weight_v, vocoder.head.resblocks.4.convs1.2.bias, vocoder.head.resblocks.4.convs1.2.weight_g, vocoder.head.resblocks.4.convs1.2.weight_v, vocoder.head.resblocks.4.convs2.0.bias, vocoder.head.resblocks.4.convs2.0.weight_g, vocoder.head.resblocks.4.convs2.0.weight_v, vocoder.head.resblocks.4.convs2.1.bias, vocoder.head.resblocks.4.convs2.1.weight_g, vocoder.head.resblocks.4.convs2.1.weight_v, vocoder.head.resblocks.4.convs2.2.bias, vocoder.head.resblocks.4.convs2.2.weight_g, vocoder.head.resblocks.4.convs2.2.weight_v, vocoder.head.resblocks.5.convs1.0.bias, vocoder.head.resblocks.5.convs1.0.weight_g, vocoder.head.resblocks.5.convs1.0.weight_v, vocoder.head.resblocks.5.convs1.1.bias, vocoder.head.resblocks.5.convs1.1.weight_g, vocoder.head.resblocks.5.convs1.1.weight_v, vocoder.head.resblocks.5.convs1.2.bias, vocoder.head.resblocks.5.convs1.2.weight_g, vocoder.head.resblocks.5.convs1.2.weight_v, vocoder.head.resblocks.5.convs2.0.bias, vocoder.head.resblocks.5.convs2.0.weight_g, vocoder.head.resblocks.5.convs2.0.weight_v, vocoder.head.resblocks.5.convs2.1.bias, vocoder.head.resblocks.5.convs2.1.weight_g, vocoder.head.resblocks.5.convs2.1.weight_v, vocoder.head.resblocks.5.convs2.2.bias, vocoder.head.resblocks.5.convs2.2.weight_g, vocoder.head.resblocks.5.convs2.2.weight_v, vocoder.head.resblocks.6.convs1.0.bias, vocoder.head.resblocks.6.convs1.0.weight_g, vocoder.head.resblocks.6.convs1.0.weight_v, vocoder.head.resblocks.6.convs1.1.bias, vocoder.head.resblocks.6.convs1.1.weight_g, vocoder.head.resblocks.6.convs1.1.weight_v, vocoder.head.resblocks.6.convs1.2.bias, vocoder.head.resblocks.6.convs1.2.weight_g, vocoder.head.resblocks.6.convs1.2.weight_v, vocoder.head.resblocks.6.convs2.0.bias, vocoder.head.resblocks.6.convs2.0.weight_g, vocoder.head.resblocks.6.convs2.0.weight_v, vocoder.head.resblocks.6.convs2.1.bias, vocoder.head.resblocks.6.convs2.1.weight_g, vocoder.head.resblocks.6.convs2.1.weight_v, vocoder.head.resblocks.6.convs2.2.bias, vocoder.head.resblocks.6.convs2.2.weight_g, vocoder.head.resblocks.6.convs2.2.weight_v, vocoder.head.resblocks.7.convs1.0.bias, vocoder.head.resblocks.7.convs1.0.weight_g, vocoder.head.resblocks.7.convs1.0.weight_v, vocoder.head.resblocks.7.convs1.1.bias, vocoder.head.resblocks.7.convs1.1.weight_g, vocoder.head.resblocks.7.convs1.1.weight_v, vocoder.head.resblocks.7.convs1.2.bias, vocoder.head.resblocks.7.convs1.2.weight_g, vocoder.head.resblocks.7.convs1.2.weight_v, vocoder.head.resblocks.7.convs2.0.bias, vocoder.head.resblocks.7.convs2.0.weight_g, vocoder.head.resblocks.7.convs2.0.weight_v, vocoder.head.resblocks.7.convs2.1.bias, vocoder.head.resblocks.7.convs2.1.weight_g, vocoder.head.resblocks.7.convs2.1.weight_v, vocoder.head.resblocks.7.convs2.2.bias, vocoder.head.resblocks.7.convs2.2.weight_g, vocoder.head.resblocks.7.convs2.2.weight_v, vocoder.head.resblocks.8.convs1.0.bias, vocoder.head.resblocks.8.convs1.0.weight_g, vocoder.head.resblocks.8.convs1.0.weight_v, vocoder.head.resblocks.8.convs1.1.bias, vocoder.head.resblocks.8.convs1.1.weight_g, vocoder.head.resblocks.8.convs1.1.weight_v, vocoder.head.resblocks.8.convs1.2.bias, vocoder.head.resblocks.8.convs1.2.weight_g, vocoder.head.resblocks.8.convs1.2.weight_v, vocoder.head.resblocks.8.convs2.0.bias, vocoder.head.resblocks.8.convs2.0.weight_g, vocoder.head.resblocks.8.convs2.0.weight_v, vocoder.head.resblocks.8.convs2.1.bias, vocoder.head.resblocks.8.convs2.1.weight_g, vocoder.head.resblocks.8.convs2.1.weight_v, vocoder.head.resblocks.8.convs2.2.bias, vocoder.head.resblocks.8.convs2.2.weight_g, vocoder.head.resblocks.8.convs2.2.weight_v, vocoder.head.resblocks.9.convs1.0.bias, vocoder.head.resblocks.9.convs1.0.weight_g, vocoder.head.resblocks.9.convs1.0.weight_v, vocoder.head.resblocks.9.convs1.1.bias, vocoder.head.resblocks.9.convs1.1.weight_g, vocoder.head.resblocks.9.convs1.1.weight_v, vocoder.head.resblocks.9.convs1.2.bias, vocoder.head.resblocks.9.convs1.2.weight_g, vocoder.head.resblocks.9.convs1.2.weight_v, vocoder.head.resblocks.9.convs2.0.bias, vocoder.head.resblocks.9.convs2.0.weight_g, vocoder.head.resblocks.9.convs2.0.weight_v, vocoder.head.resblocks.9.convs2.1.bias, vocoder.head.resblocks.9.convs2.1.weight_g, vocoder.head.resblocks.9.convs2.1.weight_v, vocoder.head.resblocks.9.convs2.2.bias, vocoder.head.resblocks.9.convs2.2.weight_g, vocoder.head.resblocks.9.convs2.2.weight_v, vocoder.head.ups.0.bias, vocoder.head.ups.0.weight_g, vocoder.head.ups.0.weight_v, vocoder.head.ups.1.bias, vocoder.head.ups.1.weight_g, vocoder.head.ups.1.weight_v, vocoder.head.ups.2.bias, vocoder.head.ups.2.weight_g, vocoder.head.ups.2.weight_v, vocoder.head.ups.3.bias, vocoder.head.ups.3.weight_g, vocoder.head.ups.3.weight_v, vocoder.head.ups.4.bias, vocoder.head.ups.4.weight_g, vocoder.head.ups.4.weight_v, vocoder.head.ups.5.bias, vocoder.head.ups.5.weight_g, vocoder.head.ups.5.weight_v, vocoder.head.ups.6.bias, vocoder.head.ups.6.weight_g, vocoder.head.ups.6.weight_v, vocoder.mel_transform.mel_scale.fb, vocoder.mel_transform.spectrogram.window']

In [4]:

music_dcae

Out [4]:

MusicDCAE(
  (dcae): AutoencoderDC(
    (encoder): Encoder(
      (conv_in): Conv2d(2, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (down_blocks): ModuleList(
        (0): Sequential(
          (0): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (1): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (2): DCDownBlock2d(
            (conv): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
          )
        )
        (1): Sequential(
          (0): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (1): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (2): DCDownBlock2d(
            (conv): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
          )
        )
        (2): Sequential(
          (0): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (1): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (2): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (3): DCDownBlock2d(
            (conv): Conv2d(512, 1024, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
          )
        )
        (3): Sequential(
          (0): EfficientViTBlock(
            (attn): SanaMultiscaleLinearAttention(
              (to_q): Linear(in_features=1024, out_features=1024, bias=False)
              (to_k): Linear(in_features=1024, out_features=1024, bias=False)
              (to_v): Linear(in_features=1024, out_features=1024, bias=False)
              (to_qkv_multiscale): ModuleList(
                (0): SanaMultiscaleAttentionProjection(
                  (proj_in): Conv2d(3072, 3072, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=3072, bias=False)
                  (proj_out): Conv2d(3072, 3072, kernel_size=(1, 1), stride=(1, 1), groups=96, bias=False)
                )
              )
              (nonlinearity): ReLU()
              (to_out): Linear(in_features=2048, out_features=1024, bias=False)
              (norm_out): RMSNorm()
            )
            (conv_out): GLUMBConv(
              (nonlinearity): SiLU()
              (conv_inverted): Conv2d(1024, 8192, kernel_size=(1, 1), stride=(1, 1))
              (conv_depth): Conv2d(8192, 8192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8192)
              (conv_point): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
              (norm): RMSNorm()
            )
          )
          (1): EfficientViTBlock(
            (attn): SanaMultiscaleLinearAttention(
              (to_q): Linear(in_features=1024, out_features=1024, bias=False)
              (to_k): Linear(in_features=1024, out_features=1024, bias=False)
              (to_v): Linear(in_features=1024, out_features=1024, bias=False)
              (to_qkv_multiscale): ModuleList(
                (0): SanaMultiscaleAttentionProjection(
                  (proj_in): Conv2d(3072, 3072, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=3072, bias=False)
                  (proj_out): Conv2d(3072, 3072, kernel_size=(1, 1), stride=(1, 1), groups=96, bias=False)
                )
              )
              (nonlinearity): ReLU()
              (to_out): Linear(in_features=2048, out_features=1024, bias=False)
              (norm_out): RMSNorm()
            )
            (conv_out): GLUMBConv(
              (nonlinearity): SiLU()
              (conv_inverted): Conv2d(1024, 8192, kernel_size=(1, 1), stride=(1, 1))
              (conv_depth): Conv2d(8192, 8192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8192)
              (conv_point): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
              (norm): RMSNorm()
            )
          )
          (2): EfficientViTBlock(
            (attn): SanaMultiscaleLinearAttention(
              (to_q): Linear(in_features=1024, out_features=1024, bias=False)
              (to_k): Linear(in_features=1024, out_features=1024, bias=False)
              (to_v): Linear(in_features=1024, out_features=1024, bias=False)
              (to_qkv_multiscale): ModuleList(
                (0): SanaMultiscaleAttentionProjection(
                  (proj_in): Conv2d(3072, 3072, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=3072, bias=False)
                  (proj_out): Conv2d(3072, 3072, kernel_size=(1, 1), stride=(1, 1), groups=96, bias=False)
                )
              )
              (nonlinearity): ReLU()
              (to_out): Linear(in_features=2048, out_features=1024, bias=False)
              (norm_out): RMSNorm()
            )
            (conv_out): GLUMBConv(
              (nonlinearity): SiLU()
              (conv_inverted): Conv2d(1024, 8192, kernel_size=(1, 1), stride=(1, 1))
              (conv_depth): Conv2d(8192, 8192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8192)
              (conv_point): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
              (norm): RMSNorm()
            )
          )
        )
      )
      (conv_out): Conv2d(1024, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    )
    (decoder): Decoder(
      (conv_in): Conv2d(8, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (up_blocks): ModuleList(
        (0): Sequential(
          (0): DCUpBlock2d(
            (conv): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          )
          (1): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (2): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (3): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
        )
        (1): Sequential(
          (0): DCUpBlock2d(
            (conv): Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          )
          (1): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (2): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (3): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
        )
        (2): Sequential(
          (0): DCUpBlock2d(
            (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
          )
          (1): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (2): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
          (3): ResBlock(
            (nonlinearity): SiLU()
            (conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
            (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
            (norm): RMSNorm()
          )
        )
        (3): Sequential(
          (0): EfficientViTBlock(
            (attn): SanaMultiscaleLinearAttention(
              (to_q): Linear(in_features=1024, out_features=1024, bias=False)
              (to_k): Linear(in_features=1024, out_features=1024, bias=False)
              (to_v): Linear(in_features=1024, out_features=1024, bias=False)
              (to_qkv_multiscale): ModuleList(
                (0): SanaMultiscaleAttentionProjection(
                  (proj_in): Conv2d(3072, 3072, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=3072, bias=False)
                  (proj_out): Conv2d(3072, 3072, kernel_size=(1, 1), stride=(1, 1), groups=96, bias=False)
                )
              )
              (nonlinearity): ReLU()
              (to_out): Linear(in_features=2048, out_features=1024, bias=False)
              (norm_out): RMSNorm()
            )
            (conv_out): GLUMBConv(
              (nonlinearity): SiLU()
              (conv_inverted): Conv2d(1024, 8192, kernel_size=(1, 1), stride=(1, 1))
              (conv_depth): Conv2d(8192, 8192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8192)
              (conv_point): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
              (norm): RMSNorm()
            )
          )
          (1): EfficientViTBlock(
            (attn): SanaMultiscaleLinearAttention(
              (to_q): Linear(in_features=1024, out_features=1024, bias=False)
              (to_k): Linear(in_features=1024, out_features=1024, bias=False)
              (to_v): Linear(in_features=1024, out_features=1024, bias=False)
              (to_qkv_multiscale): ModuleList(
                (0): SanaMultiscaleAttentionProjection(
                  (proj_in): Conv2d(3072, 3072, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=3072, bias=False)
                  (proj_out): Conv2d(3072, 3072, kernel_size=(1, 1), stride=(1, 1), groups=96, bias=False)
                )
              )
              (nonlinearity): ReLU()
              (to_out): Linear(in_features=2048, out_features=1024, bias=False)
              (norm_out): RMSNorm()
            )
            (conv_out): GLUMBConv(
              (nonlinearity): SiLU()
              (conv_inverted): Conv2d(1024, 8192, kernel_size=(1, 1), stride=(1, 1))
              (conv_depth): Conv2d(8192, 8192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8192)
              (conv_point): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
              (norm): RMSNorm()
            )
          )
          (2): EfficientViTBlock(
            (attn): SanaMultiscaleLinearAttention(
              (to_q): Linear(in_features=1024, out_features=1024, bias=False)
              (to_k): Linear(in_features=1024, out_features=1024, bias=False)
              (to_v): Linear(in_features=1024, out_features=1024, bias=False)
              (to_qkv_multiscale): ModuleList(
                (0): SanaMultiscaleAttentionProjection(
                  (proj_in): Conv2d(3072, 3072, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2), groups=3072, bias=False)
                  (proj_out): Conv2d(3072, 3072, kernel_size=(1, 1), stride=(1, 1), groups=96, bias=False)
                )
              )
              (nonlinearity): ReLU()
              (to_out): Linear(in_features=2048, out_features=1024, bias=False)
              (norm_out): RMSNorm()
            )
            (conv_out): GLUMBConv(
              (nonlinearity): SiLU()
              (conv_inverted): Conv2d(1024, 8192, kernel_size=(1, 1), stride=(1, 1))
              (conv_depth): Conv2d(8192, 8192, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=8192)
              (conv_point): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
              (norm): RMSNorm()
            )
          )
        )
      )
      (norm_out): RMSNorm()
      (conv_act): ReLU()
      (conv_out): Conv2d(128, 2, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    )
  )
  (vocoder): ADaMoSHiFiGANV1(
    (backbone): ConvNeXtEncoder(
      (channel_layers): ModuleList(
        (0): Sequential(
          (0): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(3,), padding_mode=replicate)
          (1): LayerNorm()
        )
        (1): Sequential(
          (0): LayerNorm()
          (1): Conv1d(128, 256, kernel_size=(1,), stride=(1,))
        )
        (2): Sequential(
          (0): LayerNorm()
          (1): Conv1d(256, 384, kernel_size=(1,), stride=(1,))
        )
        (3): Sequential(
          (0): LayerNorm()
          (1): Conv1d(384, 512, kernel_size=(1,), stride=(1,))
        )
      )
      (stages): ModuleList(
        (0): Sequential(
          (0): ConvNeXtBlock(
            (dwconv): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(3,), groups=128)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=128, out_features=512, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=512, out_features=128, bias=True)
            (drop_path): Identity()
          )
          (1): ConvNeXtBlock(
            (dwconv): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(3,), groups=128)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=128, out_features=512, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=512, out_features=128, bias=True)
            (drop_path): Identity()
          )
          (2): ConvNeXtBlock(
            (dwconv): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(3,), groups=128)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=128, out_features=512, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=512, out_features=128, bias=True)
            (drop_path): Identity()
          )
        )
        (1): Sequential(
          (0): ConvNeXtBlock(
            (dwconv): Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(3,), groups=256)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=256, out_features=1024, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1024, out_features=256, bias=True)
            (drop_path): Identity()
          )
          (1): ConvNeXtBlock(
            (dwconv): Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(3,), groups=256)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=256, out_features=1024, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1024, out_features=256, bias=True)
            (drop_path): Identity()
          )
          (2): ConvNeXtBlock(
            (dwconv): Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(3,), groups=256)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=256, out_features=1024, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1024, out_features=256, bias=True)
            (drop_path): Identity()
          )
        )
        (2): Sequential(
          (0): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (1): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (2): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (3): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (4): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (5): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (6): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (7): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
          (8): ConvNeXtBlock(
            (dwconv): Conv1d(384, 384, kernel_size=(7,), stride=(1,), padding=(3,), groups=384)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=384, out_features=1536, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=1536, out_features=384, bias=True)
            (drop_path): Identity()
          )
        )
        (3): Sequential(
          (0): ConvNeXtBlock(
            (dwconv): Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(3,), groups=512)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=512, out_features=2048, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=2048, out_features=512, bias=True)
            (drop_path): Identity()
          )
          (1): ConvNeXtBlock(
            (dwconv): Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(3,), groups=512)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=512, out_features=2048, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=2048, out_features=512, bias=True)
            (drop_path): Identity()
          )
          (2): ConvNeXtBlock(
            (dwconv): Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(3,), groups=512)
            (norm): LayerNorm()
            (pwconv1): Linear(in_features=512, out_features=2048, bias=True)
            (act): GELU(approximate='none')
            (pwconv2): Linear(in_features=2048, out_features=512, bias=True)
            (drop_path): Identity()
          )
        )
      )
      (norm): LayerNorm()
    )
    (head): HiFiGANGenerator(
      (conv_pre): Conv1d(512, 1024, kernel_size=(13,), stride=(1,), padding=(6,))
      (noise_convs): ModuleList()
      (ups): ModuleList(
        (0): ConvTranspose1d(1024, 512, kernel_size=(8,), stride=(4,), padding=(2,))
        (1): ConvTranspose1d(512, 256, kernel_size=(8,), stride=(4,), padding=(2,))
        (2): ConvTranspose1d(256, 128, kernel_size=(4,), stride=(2,), padding=(1,))
        (3): ConvTranspose1d(128, 64, kernel_size=(4,), stride=(2,), padding=(1,))
        (4): ConvTranspose1d(64, 32, kernel_size=(4,), stride=(2,), padding=(1,))
        (5): ConvTranspose1d(32, 16, kernel_size=(4,), stride=(2,), padding=(1,))
        (6): ConvTranspose1d(16, 8, kernel_size=(4,), stride=(2,), padding=(1,))
      )
      (resblocks): ModuleList(
        (0): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(512, 512, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(512, 512, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(512, 512, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(512, 512, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (1): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(512, 512, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (2): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(512, 512, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(512, 512, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(512, 512, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(512, 512, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (3): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(512, 512, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(512, 512, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(512, 512, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(512, 512, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
        (4): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(256, 256, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(256, 256, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(256, 256, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(256, 256, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (5): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(256, 256, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (6): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(256, 256, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(256, 256, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(256, 256, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(256, 256, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (7): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(256, 256, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(256, 256, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(256, 256, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(256, 256, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
        (8): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(128, 128, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(128, 128, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(128, 128, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(128, 128, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (9): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(128, 128, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (10): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(128, 128, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(128, 128, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(128, 128, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(128, 128, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (11): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(128, 128, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(128, 128, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(128, 128, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(128, 128, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
        (12): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(64, 64, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(64, 64, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(64, 64, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(64, 64, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (13): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(64, 64, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (14): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(64, 64, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(64, 64, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(64, 64, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(64, 64, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (15): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(64, 64, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(64, 64, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(64, 64, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(64, 64, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
        (16): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(32, 32, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(32, 32, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(32, 32, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(32, 32, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (17): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(32, 32, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(32, 32, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(32, 32, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(32, 32, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (18): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(32, 32, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(32, 32, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(32, 32, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(32, 32, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (19): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(32, 32, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(32, 32, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(32, 32, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(32, 32, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
        (20): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(16, 16, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(16, 16, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(16, 16, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(16, 16, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (21): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(16, 16, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(16, 16, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(16, 16, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(16, 16, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (22): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(16, 16, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(16, 16, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(16, 16, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(16, 16, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (23): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(16, 16, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(16, 16, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(16, 16, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(16, 16, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
        (24): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(8, 8, kernel_size=(3,), stride=(1,), padding=(1,))
            (1): Conv1d(8, 8, kernel_size=(3,), stride=(1,), padding=(3,), dilation=(3,))
            (2): Conv1d(8, 8, kernel_size=(3,), stride=(1,), padding=(5,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(8, 8, kernel_size=(3,), stride=(1,), padding=(1,))
          )
        )
        (25): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(8, 8, kernel_size=(7,), stride=(1,), padding=(3,))
            (1): Conv1d(8, 8, kernel_size=(7,), stride=(1,), padding=(9,), dilation=(3,))
            (2): Conv1d(8, 8, kernel_size=(7,), stride=(1,), padding=(15,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(8, 8, kernel_size=(7,), stride=(1,), padding=(3,))
          )
        )
        (26): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(8, 8, kernel_size=(11,), stride=(1,), padding=(5,))
            (1): Conv1d(8, 8, kernel_size=(11,), stride=(1,), padding=(15,), dilation=(3,))
            (2): Conv1d(8, 8, kernel_size=(11,), stride=(1,), padding=(25,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(8, 8, kernel_size=(11,), stride=(1,), padding=(5,))
          )
        )
        (27): ResBlock1(
          (convs1): ModuleList(
            (0): Conv1d(8, 8, kernel_size=(13,), stride=(1,), padding=(6,))
            (1): Conv1d(8, 8, kernel_size=(13,), stride=(1,), padding=(18,), dilation=(3,))
            (2): Conv1d(8, 8, kernel_size=(13,), stride=(1,), padding=(30,), dilation=(5,))
          )
          (convs2): ModuleList(
            (0-2): 3 x Conv1d(8, 8, kernel_size=(13,), stride=(1,), padding=(6,))
          )
        )
      )
      (activation_post): SiLU(inplace=True)
      (conv_post): Conv1d(8, 1, kernel_size=(13,), stride=(1,), padding=(6,))
    )
    (mel_transform): LogMelSpectrogram(
      (spectrogram): LinearSpectrogram()
      (mel_scale): MelScale()
    )
  )
  (resampler): Resample()
)

In [2]:

music_dcae = MusicDCAE()

/root/miniconda3/envs/sag_train/lib/python3.10/site-packages/torch/nn/utils/weight_norm.py:143: FutureWarning: `torch.nn.utils.weight_norm` is deprecated in favor of `torch.nn.utils.parametrizations.weight_norm`.
  WeightNorm.apply(module, name, dim)

In [3]:

music_dcae.dtype

Out [3]:

torch.float32

In [4]:

import torch

music_dcae = music_dcae.to(torch.bfloat16)

In [5]:

music_dcae.save_pretrained("checkpoints/music_dcae_vocoder")

In [ ]: