audio_vae/config.json
| 1 | { |
| 2 | "_class_name": "AutoencoderKLLTX2Audio", |
| 3 | "_diffusers_version": "0.37.0.dev0", |
| 4 | "attn_resolutions": null, |
| 5 | "base_channels": 128, |
| 6 | "causality_axis": "height", |
| 7 | "ch_mult": [ |
| 8 | 1, |
| 9 | 2, |
| 10 | 4 |
| 11 | ], |
| 12 | "double_z": true, |
| 13 | "dropout": 0.0, |
| 14 | "in_channels": 2, |
| 15 | "is_causal": true, |
| 16 | "latent_channels": 8, |
| 17 | "mel_bins": 64, |
| 18 | "mel_hop_length": 160, |
| 19 | "mid_block_add_attention": false, |
| 20 | "norm_type": "pixel", |
| 21 | "num_res_blocks": 2, |
| 22 | "output_channels": 2, |
| 23 | "resolution": 256, |
| 24 | "sample_rate": 16000 |
| 25 | } |
| 26 | |