FluidInference/supertonic-3-coreml
873
1{2 "tts_version": "v1.7.3",3 "split": "opensource-multilingual",4 "ttl": {5 "latent_dim": 24,6 "chunk_compress_factor": 6,7 "batch_expander": {8 "n_batch_expand": 69 },10 "normalizer": {11 "scale": 0.2512 },13 "text_encoder": {14 "n_langs": 0,15 "lang_emb_dim": 0,16 "text_embedder": {17 "char_emb_dim": 25618 },19 "convnext": {20 "idim": 256,21 "ksz": 5,22 "intermediate_dim": 1024,23 "num_layers": 6,24 "dilation_lst": [25 1,26 1,27 2,28 2,29 4,30 431 ]32 },33 "attn_encoder": {34 "hidden_channels": 256,35 "filter_channels": 1024,36 "n_heads": 4,37 "n_layers": 4,38 "p_dropout": 0.039 },40 "proj_out": {41 "idim": 256,42 "odim": 25643 }44 },45 "flow_matching": {46 "sig_min": 1e-0847 },48 "style_encoder": {49 "proj_in": {50 "ldim": 24,51 "chunk_compress_factor": 6,52 "odim": 25653 },54 "convnext": {55 "idim": 256,56 "ksz": 5,57 "intermediate_dim": 1024,58 "num_layers": 6,59 "dilation_lst": [60 1,61 1,62 1,63 1,64 1,65 166 ]67 },68 "style_token_layer": {69 "input_dim": 256,70 "n_style": 50,71 "style_key_dim": 256,72 "style_value_dim": 256,73 "prototype_dim": 256,74 "n_units": 256,75 "n_heads": 276 }77 },78 "speech_prompted_text_encoder": {79 "text_dim": 256,80 "style_dim": 256,81 "n_units": 256,82 "n_heads": 283 },84 "uncond_masker": {85 "prob_both_uncond": 0.04,86 "prob_text_uncond": 0.01,87 "std": 0.1,88 "text_dim": 256,89 "n_style": 50,90 "style_key_dim": 256,91 "style_value_dim": 25692 },93 "vector_field": {94 "n_langs": 0,95 "lang_emb_dim": 0,96 "proj_in": {97 "ldim": 24,98 "chunk_compress_factor": 6,99 "odim": 512100 },101 "time_encoder": {102 "time_dim": 64,103 "hdim": 256104 },105 "main_blocks": {106 "n_blocks": 4,107 "time_cond_layer": {108 "idim": 512,109 "time_dim": 64110 },111 "style_cond_layer": {112 "idim": 512,113 "style_dim": 256114 },115 "text_cond_layer": {116 "idim": 512,117 "text_dim": 256,118 "n_heads": 8,119 "n_units": 512,120 "use_residual": true,121 "rotary_base": 10000,122 "rotary_scale": 10123 },124 "convnext_0": {125 "idim": 512,126 "ksz": 5,127 "intermediate_dim": 2048,128 "num_layers": 4,129 "dilation_lst": [130 1,131 2,132 4,133 8134 ]135 },136 "convnext_1": {137 "idim": 512,138 "ksz": 5,139 "intermediate_dim": 2048,140 "num_layers": 1,141 "dilation_lst": [142 1143 ]144 },145 "convnext_2": {146 "idim": 512,147 "ksz": 5,148 "intermediate_dim": 2048,149 "num_layers": 1,150 "dilation_lst": [151 1152 ]153 }154 },155 "last_convnext": {156 "idim": 512,157 "ksz": 5,158 "intermediate_dim": 2048,159 "num_layers": 4,160 "dilation_lst": [161 1,162 1,163 1,164 1165 ]166 },167 "proj_out": {168 "idim": 512,169 "chunk_compress_factor": 6,170 "ldim": 24171 }172 }173 },174 "ae": {175 "sample_rate": 44100,176 "n_delay": 0,177 "base_chunk_size": 512,178 "chunk_compress_factor": 1,179 "ldim": 24,180 "encoder": {181 "spec_processor": {182 "n_fft": 2048,183 "win_length": 2048,184 "hop_length": 512,185 "n_mels": 228,186 "sample_rate": 44100,187 "eps": 1e-05,188 "norm_mean": 0.0,189 "norm_std": 1.0190 },191 "ksz_init": 7,192 "ksz": 7,193 "num_layers": 10,194 "dilation_lst": [195 1,196 1,197 1,198 1,199 1,200 1,201 1,202 1,203 1,204 1205 ],206 "intermediate_dim": 2048,207 "idim": 1253,208 "hdim": 512,209 "odim": 24210 },211 "decoder": {212 "ksz_init": 7,213 "ksz": 7,214 "num_layers": 10,215 "dilation_lst": [216 1,217 2,218 4,219 1,220 2,221 4,222 1,223 1,224 1,225 1226 ],227 "intermediate_dim": 2048,228 "idim": 24,229 "hdim": 512,230 "head": {231 "idim": 512,232 "hdim": 2048,233 "odim": 512,234 "ksz": 3235 }236 }237 },238 "dp": {239 "latent_dim": 24,240 "chunk_compress_factor": 6,241 "normalizer": {242 "scale": 1.0243 },244 "sentence_encoder": {245 "char_emb_dim": 64,246 "text_embedder": {247 "char_emb_dim": 64248 },249 "convnext": {250 "idim": 64,251 "ksz": 5,252 "intermediate_dim": 256,253 "num_layers": 6,254 "dilation_lst": [255 1,256 1,257 1,258 1,259 1,260 1261 ]262 },263 "attn_encoder": {264 "hidden_channels": 64,265 "filter_channels": 256,266 "n_heads": 2,267 "n_layers": 2,268 "p_dropout": 0.0269 },270 "proj_out": {271 "idim": 64,272 "odim": 64273 }274 },275 "style_encoder": {276 "proj_in": {277 "ldim": 24,278 "chunk_compress_factor": 6,279 "odim": 64280 },281 "convnext": {282 "idim": 64,283 "ksz": 5,284 "intermediate_dim": 256,285 "num_layers": 4,286 "dilation_lst": [287 1,288 1,289 1,290 1291 ]292 },293 "style_token_layer": {294 "input_dim": 64,295 "n_style": 8,296 "style_key_dim": 0,297 "style_value_dim": 16,298 "prototype_dim": 64,299 "n_units": 64,300 "n_heads": 2301 }302 },303 "predictor": {304 "sentence_dim": 64,305 "n_style": 8,306 "style_dim": 16,307 "hdim": 128,308 "n_layer": 2309 }310 }311}312 