Model:
TextConditionedSeparator(
  (clap): ClapModel(
    (text_model): ClapTextModel(
      (embeddings): ClapTextEmbeddings(
        (word_embeddings): Embedding(50265, 768, padding_idx=1)
        (position_embeddings): Embedding(514, 768, padding_idx=1)
        (token_type_embeddings): Embedding(1, 768)
        (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
        (dropout): Dropout(p=0.1, inplace=False)
      )
      (encoder): ClapTextEncoder(
        (layer): ModuleList(
          (0-11): 12 x ClapTextLayer(
            (attention): ClapTextAttention(
              (self): ClapTextSelfAttention(
                (query): Linear(in_features=768, out_features=768, bias=True)
                (key): Linear(in_features=768, out_features=768, bias=True)
                (value): Linear(in_features=768, out_features=768, bias=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
              (output): ClapTextSelfOutput(
                (dense): Linear(in_features=768, out_features=768, bias=True)
                (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
                (dropout): Dropout(p=0.1, inplace=False)
              )
            )
            (intermediate): ClapTextIntermediate(
              (dense): Linear(in_features=768, out_features=3072, bias=True)
              (intermediate_act_fn): GELUActivation()
            )
            (output): ClapTextOutput(
              (dense): Linear(in_features=3072, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
        )
      )
      (pooler): ClapTextPooler(
        (dense): Linear(in_features=768, out_features=768, bias=True)
        (activation): Tanh()
      )
    )
    (text_projection): ClapProjectionLayer(
      (linear1): Linear(in_features=768, out_features=512, bias=True)
      (activation): ReLU()
      (linear2): Linear(in_features=512, out_features=512, bias=True)
    )
    (audio_model): ClapAudioModel(
      (audio_encoder): ClapAudioEncoder(
        (patch_embed): ClapAudioPatchEmbed(
          (proj): Conv2d(1, 96, kernel_size=(4, 4), stride=(4, 4))
          (norm): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
        )
        (layers): ModuleList(
          (0): ClapAudioStage(
            (blocks): ModuleList(
              (0-1): 2 x ClapAudioLayer(
                (layernorm_before): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
                (attention): ClapAudioAttention(
                  (self): ClapAudioSelfAttention(
                    (query): Linear(in_features=96, out_features=96, bias=True)
                    (key): Linear(in_features=96, out_features=96, bias=True)
                    (value): Linear(in_features=96, out_features=96, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                  (output): ClapAudioSelfOutput(
                    (dense): Linear(in_features=96, out_features=96, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                )
                (drop_path): Identity()
                (layernorm_after): LayerNorm((96,), eps=1e-05, elementwise_affine=True)
                (intermediate): ClapAudioIntermediate(
                  (dense): Linear(in_features=96, out_features=384, bias=True)
                  (intermediate_act_fn): GELUActivation()
                )
                (output): ClapAudioOutput(
                  (dense): Linear(in_features=384, out_features=96, bias=True)
                  (dropout): Dropout(p=0.1, inplace=False)
                )
              )
            )
            (downsample): ClapAudioPatchMerging(
              (reduction): Linear(in_features=384, out_features=192, bias=False)
              (norm): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
            )
          )
          (1): ClapAudioStage(
            (blocks): ModuleList(
              (0-1): 2 x ClapAudioLayer(
                (layernorm_before): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
                (attention): ClapAudioAttention(
                  (self): ClapAudioSelfAttention(
                    (query): Linear(in_features=192, out_features=192, bias=True)
                    (key): Linear(in_features=192, out_features=192, bias=True)
                    (value): Linear(in_features=192, out_features=192, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                  (output): ClapAudioSelfOutput(
                    (dense): Linear(in_features=192, out_features=192, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                )
                (drop_path): Identity()
                (layernorm_after): LayerNorm((192,), eps=1e-05, elementwise_affine=True)
                (intermediate): ClapAudioIntermediate(
                  (dense): Linear(in_features=192, out_features=768, bias=True)
                  (intermediate_act_fn): GELUActivation()
                )
                (output): ClapAudioOutput(
                  (dense): Linear(in_features=768, out_features=192, bias=True)
                  (dropout): Dropout(p=0.1, inplace=False)
                )
              )
            )
            (downsample): ClapAudioPatchMerging(
              (reduction): Linear(in_features=768, out_features=384, bias=False)
              (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
            )
          )
          (2): ClapAudioStage(
            (blocks): ModuleList(
              (0-5): 6 x ClapAudioLayer(
                (layernorm_before): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
                (attention): ClapAudioAttention(
                  (self): ClapAudioSelfAttention(
                    (query): Linear(in_features=384, out_features=384, bias=True)
                    (key): Linear(in_features=384, out_features=384, bias=True)
                    (value): Linear(in_features=384, out_features=384, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                  (output): ClapAudioSelfOutput(
                    (dense): Linear(in_features=384, out_features=384, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                )
                (drop_path): Identity()
                (layernorm_after): LayerNorm((384,), eps=1e-05, elementwise_affine=True)
                (intermediate): ClapAudioIntermediate(
                  (dense): Linear(in_features=384, out_features=1536, bias=True)
                  (intermediate_act_fn): GELUActivation()
                )
                (output): ClapAudioOutput(
                  (dense): Linear(in_features=1536, out_features=384, bias=True)
                  (dropout): Dropout(p=0.1, inplace=False)
                )
              )
            )
            (downsample): ClapAudioPatchMerging(
              (reduction): Linear(in_features=1536, out_features=768, bias=False)
              (norm): LayerNorm((1536,), eps=1e-05, elementwise_affine=True)
            )
          )
          (3): ClapAudioStage(
            (blocks): ModuleList(
              (0-1): 2 x ClapAudioLayer(
                (layernorm_before): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
                (attention): ClapAudioAttention(
                  (self): ClapAudioSelfAttention(
                    (query): Linear(in_features=768, out_features=768, bias=True)
                    (key): Linear(in_features=768, out_features=768, bias=True)
                    (value): Linear(in_features=768, out_features=768, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                  (output): ClapAudioSelfOutput(
                    (dense): Linear(in_features=768, out_features=768, bias=True)
                    (dropout): Dropout(p=0.0, inplace=False)
                  )
                )
                (drop_path): Identity()
                (layernorm_after): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
                (intermediate): ClapAudioIntermediate(
                  (dense): Linear(in_features=768, out_features=3072, bias=True)
                  (intermediate_act_fn): GELUActivation()
                )
                (output): ClapAudioOutput(
                  (dense): Linear(in_features=3072, out_features=768, bias=True)
                  (dropout): Dropout(p=0.1, inplace=False)
                )
              )
            )
          )
        )
        (batch_norm): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        (norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
        (avgpool): AdaptiveAvgPool1d(output_size=1)
      )
    )
    (audio_projection): ClapProjectionLayer(
      (linear1): Linear(in_features=768, out_features=512, bias=True)
      (activation): ReLU()
      (linear2): Linear(in_features=512, out_features=512, bias=True)
    )
  )
  (z_encoder): PatchConv1d(
    (conv): Conv1d(1, 256, kernel_size=(16,), stride=(8,))
  )
  (text_proj): Linear(in_features=512, out_features=256, bias=True)
  (z_proj): Linear(in_features=256, out_features=256, bias=True)
  (cross): CrossAttention(
    (attn): MultiheadAttention(
      (out_proj): NonDynamicallyQuantizableLinear(in_features=256, out_features=256, bias=True)
    )
    (ln1): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
    (ff): MLP(
      (fc1): Linear(in_features=256, out_features=1024, bias=True)
      (fc2): Linear(in_features=1024, out_features=256, bias=True)
      (act): GELU(approximate='none')
    )
    (ln2): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
  )
  (transformer): TransformerEncoder(
    (layers): ModuleList(
      (0-5): 6 x TransformerEncoderLayer(
        (self_attn): MultiheadAttention(
          (out_proj): NonDynamicallyQuantizableLinear(in_features=256, out_features=256, bias=True)
        )
        (linear1): Linear(in_features=256, out_features=1024, bias=True)
        (dropout): Dropout(p=0.1, inplace=False)
        (linear2): Linear(in_features=1024, out_features=256, bias=True)
        (norm1): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
        (norm2): LayerNorm((256,), eps=1e-05, elementwise_affine=True)
        (dropout1): Dropout(p=0.1, inplace=False)
        (dropout2): Dropout(p=0.1, inplace=False)
      )
    )
  )
  (spec_decoder): Sequential(
    (0): Linear(in_features=256, out_features=256, bias=True)
    (1): GELU(approximate='none')
    (2): Linear(in_features=256, out_features=2049, bias=True)
  )
)
output waveform shape: torch.Size([2, 1, 48000])
output spectrogram shape: torch.Size([2, 12001, 2049])