optimum-internal-testing/tiny-random-SpeechT5ForTextToSpeech-openvino
010k
1<?xml version="1.0"?>2<net name="Model3" version="11">3 <layers>4 <layer id="4" name="inputs_embeds" type="Parameter" version="opset1">5 <data shape="?,?,80" element_type="f32" />6 <output>7 <port id="0" precision="FP32" names="inputs_embeds">8 <dim>-1</dim>9 <dim>-1</dim>10 <dim>80</dim>11 </port>12 </output>13 </layer>14 <layer id="3" name="speaker_embeddings" type="Parameter" version="opset1">15 <data shape="1,512" element_type="f32" />16 <output>17 <port id="0" precision="FP32" names="speaker_embeddings">18 <dim>1</dim>19 <dim>512</dim>20 </port>21 </output>22 </layer>23 <layer id="2" name="encoder_hidden_states" type="Parameter" version="opset1">24 <data shape="?,?,24" element_type="f32" />25 <output>26 <port id="0" precision="FP32" names="encoder_hidden_states">27 <dim>-1</dim>28 <dim>-1</dim>29 <dim>24</dim>30 </port>31 </output>32 </layer>33 <layer id="1" name="encoder_attention_mask" type="Parameter" version="opset1">34 <data shape="?,?" element_type="i64" />35 <output>36 <port id="0" precision="I64" names="encoder_attention_mask">37 <dim>-1</dim>38 <dim>-1</dim>39 </port>40 </output>41 </layer>42 <layer id="0" name="beam_idx" type="Parameter" version="opset1">43 <data shape="?" element_type="i32" />44 <output>45 <port id="0" precision="I32" names="beam_idx">46 <dim>-1</dim>47 </port>48 </output>49 </layer>50 <layer id="5" name="43" type="Const" version="opset1">51 <data element_type="i64" shape="" offset="0" size="8" />52 <output>53 <port id="0" precision="I64" names="43" />54 </output>55 </layer>56 <layer id="6" name="42" type="Const" version="opset1">57 <data element_type="i64" shape="" offset="0" size="8" />58 <output>59 <port id="0" precision="I64" names="42" />60 </output>61 </layer>62 <layer id="7" name="aten::select/Gather" type="Gather" version="opset8">63 <data batch_dims="0" />64 <input>65 <port id="0" precision="FP32">66 <dim>-1</dim>67 <dim>-1</dim>68 <dim>24</dim>69 </port>70 <port id="1" precision="I64" />71 <port id="2" precision="I64" />72 </input>73 <output>74 <port id="3" precision="FP32" names="44">75 <dim>-1</dim>76 <dim>24</dim>77 </port>78 </output>79 </layer>80 <layer id="8" name="self.speecht5.decoder.wrapped_decoder.layers.1.encoder_attn.k_proj.weight" type="Const" version="opset1">81 <data element_type="f32" shape="24, 24" offset="8" size="2304" />82 <output>83 <port id="0" precision="FP32" names="self.speecht5.decoder.wrapped_decoder.layers.1.encoder_attn.k_proj.weight">84 <dim>24</dim>85 <dim>24</dim>86 </port>87 </output>88 </layer>89 <layer id="9" name="__module.speecht5.decoder.wrapped_decoder.layers.1.encoder_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1">90 <data transpose_a="false" transpose_b="true" />91 <input>92 <port id="0" precision="FP32">93 <dim>-1</dim>94 <dim>24</dim>95 </port>96 <port id="1" precision="FP32">97 <dim>24</dim>98 <dim>24</dim>99 </port>100 </input>101 <output>102 <port id="2" precision="FP32" names="441,tensor.19">103 <dim>-1</dim>104 <dim>24</dim>105 </port>106 </output>107 </layer>108 <layer id="10" name="ReadValue_18630" type="ReadValue" version="opset6">109 <data variable_id="441" variable_type="dynamic" variable_shape="..." />110 <input>111 <port id="0" precision="FP32">112 <dim>-1</dim>113 <dim>24</dim>114 </port>115 </input>116 <output>117 <port id="1" precision="FP32">118 <dim>-1</dim>119 <dim>24</dim>120 </port>121 </output>122 </layer>123 <layer id="12" name="self.speecht5.decoder.wrapped_decoder.layers.0.encoder_attn.v_proj.weight" type="Const" version="opset1">124 <data element_type="f32" shape="24, 24" offset="2312" size="2304" />125 <output>126 <port id="0" precision="FP32" names="self.speecht5.decoder.wrapped_decoder.layers.0.encoder_attn.v_proj.weight">127 <dim>24</dim>128 <dim>24</dim>129 </port>130 </output>131 </layer>132 <layer id="13" name="__module.speecht5.decoder.wrapped_decoder.layers.0.encoder_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1">133 <data transpose_a="false" transpose_b="true" />134 <input>135 <port id="0" precision="FP32">136 <dim>-1</dim>137 <dim>24</dim>138 </port>139 <port id="1" precision="FP32">140 <dim>24</dim>141 <dim>24</dim>142 </port>143 </input>144 <output>145 <port id="2" precision="FP32" names="289,tensor.9">146 <dim>-1</dim>147 <dim>24</dim>148 </port>149 </output>150 </layer>151 <layer id="14" name="ReadValue_18628" type="ReadValue" version="opset6">152 <data variable_id="289" variable_type="dynamic" variable_shape="..." />153 <input>154 <port id="0" precision="FP32">155 <dim>-1</dim>156 <dim>24</dim>157 </port>158 </input>159 <output>160 <port id="1" precision="FP32">161 <dim>-1</dim>162 <dim>24</dim>163 </port>164 </output>165 </layer>166 <layer id="16" name="self.speecht5.decoder.wrapped_decoder.layers.1.encoder_attn.v_proj.weight" type="Const" version="opset1">167 <data element_type="f32" shape="24, 24" offset="4616" size="2304" />168 <output>169 <port id="0" precision="FP32" names="self.speecht5.decoder.wrapped_decoder.layers.1.encoder_attn.v_proj.weight">170 <dim>24</dim>171 <dim>24</dim>172 </port>173 </output>174 </layer>175 <layer id="17" name="__module.speecht5.decoder.wrapped_decoder.layers.1.encoder_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1">176 <data transpose_a="false" transpose_b="true" />177 <input>178 <port id="0" precision="FP32">179 <dim>-1</dim>180 <dim>24</dim>181 </port>182 <port id="1" precision="FP32">183 <dim>24</dim>184 <dim>24</dim>185 </port>186 </input>187 <output>188 <port id="2" precision="FP32" names="448,tensor.21">189 <dim>-1</dim>190 <dim>24</dim>191 </port>192 </output>193 </layer>194 <layer id="18" name="ReadValue_18626" type="ReadValue" version="opset6">195 <data variable_id="448" variable_type="dynamic" variable_shape="..." />196 <input>197 <port id="0" precision="FP32">198 <dim>-1</dim>199 <dim>24</dim>200 </port>201 </input>202 <output>203 <port id="1" precision="FP32">204 <dim>-1</dim>205 <dim>24</dim>206 </port>207 </output>208 </layer>209 <layer id="20" name="self.speecht5.decoder.wrapped_decoder.layers.0.encoder_attn.k_proj.weight" type="Const" version="opset1">210 <data element_type="f32" shape="24, 24" offset="6920" size="2304" />211 <output>212 <port id="0" precision="FP32" names="self.speecht5.decoder.wrapped_decoder.layers.0.encoder_attn.k_proj.weight">213 <dim>24</dim>214 <dim>24</dim>215 </port>216 </output>217 </layer>218 <layer id="21" name="__module.speecht5.decoder.wrapped_decoder.layers.0.encoder_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1">219 <data transpose_a="false" transpose_b="true" />220 <input>221 <port id="0" precision="FP32">222 <dim>-1</dim>223 <dim>24</dim>224 </port>225 <port id="1" precision="FP32">226 <dim>24</dim>227 <dim>24</dim>228 </port>229 </input>230 <output>231 <port id="2" precision="FP32" names="282,tensor.7">232 <dim>-1</dim>233 <dim>24</dim>234 </port>235 </output>236 </layer>237 <layer id="22" name="ReadValue_18624" type="ReadValue" version="opset6">238 <data variable_id="282" variable_type="dynamic" variable_shape="..." />239 <input>240 <port id="0" precision="FP32">241 <dim>-1</dim>242 <dim>24</dim>243 </port>244 </input>245 <output>246 <port id="1" precision="FP32">247 <dim>-1</dim>248 <dim>24</dim>249 </port>250 </output>251 </layer>252 <layer id="24" name="Constant_18386" type="Const" version="opset1">253 <data element_type="f32" shape="" offset="9224" size="4" />254 <output>255 <port id="0" precision="FP32" />256 </output>257 </layer>258 <layer id="25" name="ShapeOf_18372" type="ShapeOf" version="opset3">259 <data output_type="i64" />260 <input>261 <port id="0" precision="FP32">262 <dim>-1</dim>263 <dim>-1</dim>264 <dim>80</dim>265 </port>266 </input>267 <output>268 <port id="1" precision="I64">269 <dim>3</dim>270 </port>271 </output>272 </layer>273 <layer id="26" name="Constant_18374" type="Const" version="opset1">274 <data element_type="i64" shape="1" offset="0" size="8" />275 <output>276 <port id="0" precision="I64">277 <dim>1</dim>278 </port>279 </output>280 </layer>281 <layer id="27" name="Constant_18376" type="Const" version="opset1">282 <data element_type="i64" shape="" offset="0" size="8" />283 <output>284 <port id="0" precision="I64" />285 </output>286 </layer>287 <layer id="28" name="Gather_18377" type="Gather" version="opset8">288 <data batch_dims="0" />289 <input>290 <port id="0" precision="I64">291 <dim>3</dim>292 </port>293 <port id="1" precision="I64">294 <dim>1</dim>295 </port>296 <port id="2" precision="I64" />297 </input>298 <output>299 <port id="3" precision="I64">300 <dim>1</dim>301 </port>302 </output>303 </layer>304 <layer id="29" name="Constant_18379" type="Const" version="opset1">305 <data element_type="i64" shape="1" offset="9228" size="8" />306 <output>307 <port id="0" precision="I64">308 <dim>1</dim>309 </port>310 </output>311 </layer>312 <layer id="30" name="Constant_18381" type="Const" version="opset1">313 <data element_type="i64" shape="1" offset="0" size="8" />314 <output>315 <port id="0" precision="I64">316 <dim>1</dim>317 </port>318 </output>319 </layer>320 <layer id="31" name="Constant_18383" type="Const" version="opset1">321 <data element_type="i64" shape="1" offset="9236" size="8" />322 <output>323 <port id="0" precision="I64">324 <dim>1</dim>325 </port>326 </output>327 </layer>328 <layer id="32" name="Concat_18384" type="Concat" version="opset1">329 <data axis="0" />330 <input>331 <port id="0" precision="I64">332 <dim>1</dim>333 </port>334 <port id="1" precision="I64">335 <dim>1</dim>336 </port>337 <port id="2" precision="I64">338 <dim>1</dim>339 </port>340 <port id="3" precision="I64">341 <dim>1</dim>342 </port>343 </input>344 <output>345 <port id="4" precision="I64">346 <dim>4</dim>347 </port>348 </output>349 </layer>350 <layer id="33" name="Broadcast_18387" type="Broadcast" version="opset3">351 <data mode="numpy" />352 <input>353 <port id="0" precision="FP32" />354 <port id="1" precision="I64">355 <dim>4</dim>356 </port>357 </input>358 <output>359 <port id="2" precision="FP32">360 <dim>-1</dim>361 <dim>2</dim>362 <dim>0</dim>363 <dim>12</dim>364 </port>365 </output>366 </layer>367 <layer id="34" name="ReadValue_18180" type="ReadValue" version="opset6">368 <data variable_id="past_key_values.1.decoder.valuepresent.0.encoder.value" variable_type="f32" variable_shape="?,2,?,12" />369 <input>370 <port id="0" precision="FP32">371 <dim>-1</dim>372 <dim>2</dim>373 <dim>0</dim>374 <dim>12</dim>375 </port>376 </input>377 <output>378 <port id="1" precision="FP32" names="past_key_values.1.decoder.value">379 <dim>-1</dim>380 <dim>2</dim>381 <dim>-1</dim>382 <dim>12</dim>383 </port>384 </output>385 </layer>386 <layer id="35" name="Constant_17982" type="Const" version="opset1">387 <data element_type="i64" shape="" offset="0" size="8" />388 <output>389 <port id="0" precision="I64" />390 </output>391 </layer>392 <layer id="36" name="Gather_17983" type="Gather" version="opset8">393 <data batch_dims="0" />394 <input>395 <port id="0" precision="FP32">396 <dim>-1</dim>397 <dim>2</dim>398 <dim>-1</dim>399 <dim>12</dim>400 </port>401 <port id="1" precision="I32">402 <dim>-1</dim>403 </port>404 <port id="2" precision="I64" />405 </input>406 <output>407 <port id="3" precision="FP32">408 <dim>-1</dim>409 <dim>2</dim>410 <dim>-1</dim>411 <dim>12</dim>412 </port>413 </output>414 </layer>415 <layer id="37" name="self.speecht5.decoder.prenet.layers.0.weight" type="Const" version="opset1">416 <data element_type="f32" shape="256, 80" offset="9244" size="81920" />417 <output>418 <port id="0" precision="FP32" names="self.speecht5.decoder.prenet.layers.0.weight">419 <dim>256</dim>420 <dim>80</dim>421 </port>422 </output>423 </layer>424 <layer id="38" name="__module.speecht5.decoder.prenet.layers.0/aten::linear/MatMul" type="MatMul" version="opset1">425 <data transpose_a="false" transpose_b="true" />426 <input>427 <port id="0" precision="FP32">428 <dim>-1</dim>429 <dim>-1</dim>430 <dim>80</dim>431 </port>432 <port id="1" precision="FP32">433 <dim>256</dim>434 <dim>80</dim>435 </port>436 </input>437 <output>438 <port id="2" precision="FP32" names="106">439 <dim>-1</dim>440 <dim>-1</dim>441 <dim>256</dim>442 </port>443 </output>444 </layer>445 <layer id="39" name="__module.speecht5.decoder.prenet/aten::relu/Relu" type="ReLU" version="opset1">446 <input>447 <port id="0" precision="FP32">448 <dim>-1</dim>449 <dim>-1</dim>450 <dim>256</dim>451 </port>452 </input>453 <output>454 <port id="1" precision="FP32" names="107,inputs_embeds.3">455 <dim>-1</dim>456 <dim>-1</dim>457 <dim>256</dim>458 </port>459 </output>460 </layer>461 <layer id="40" name="ShapeOf_16336" type="ShapeOf" version="opset3">462 <data output_type="i32" />463 <input>464 <port id="0" precision="FP32">465 <dim>-1</dim>466 <dim>-1</dim>467 <dim>256</dim>468 </port>469 </input>470 <output>471 <port id="1" precision="I32">472 <dim>3</dim>473 </port>474 </output>475 </layer>476 <layer id="41" name="Constant_16337" type="Const" version="opset1">477 <data element_type="i64" shape="2" offset="91164" size="16" />478 <output>479 <port id="0" precision="I64">480 <dim>2</dim>481 </port>482 </output>483 </layer>484 <layer id="42" name="Constant_16335" type="Const" version="opset1">485 <data element_type="i64" shape="" offset="0" size="8" />486 <output>487 <port id="0" precision="I64" />488 </output>489 </layer>490 <layer id="43" name="__module.speecht5.decoder.prenet/aten::bernoulli/ShapeOf" type="Gather" version="opset1">491 <input>492 <port id="0" precision="I32">493 <dim>3</dim>494 </port>495 <port id="1" precision="I64">496 <dim>2</dim>497 </port>498 <port id="2" precision="I64" />499 </input>500 <output>501 <port id="3" precision="I32">502 <dim>2</dim>503 </port>504 </output>505 </layer>506 <layer id="44" name="__module.speecht5.decoder.prenet/aten::bernoulli/Constant_2" type="Const" version="opset1">507 <data element_type="f32" shape="" offset="9224" size="4" />508 <output>509 <port id="0" precision="FP32" />510 </output>511 </layer>512 <layer id="45" name="__module.speecht5.decoder.prenet/aten::bernoulli/Constant_3" type="Const" version="opset1">513 <data element_type="f32" shape="" offset="91180" size="4" />514 <output>515 <port id="0" precision="FP32" />516 </output>517 </layer>518 <layer id="46" name="__module.speecht5.decoder.prenet/aten::bernoulli/RandomUniform_1" type="RandomUniform" version="opset8">519 <data output_type="f32" op_seed="0" global_seed="0" alignment="pytorch" />520 <input>521 <port id="0" precision="I32">522 <dim>2</dim>523 </port>524 <port id="1" precision="FP32" />525 <port id="2" precision="FP32" />526 </input>527 <output>528 <port id="3" precision="FP32">529 <dim>-1</dim>530 <dim>256</dim>531 </port>532 </output>533 </layer>534 <layer id="47" name="Constant_16648" type="Const" version="opset1">535 <data element_type="f32" shape="1, 1" offset="91184" size="4" />536 <output>537 <port id="0" precision="FP32">538 <dim>1</dim>539 <dim>1</dim>540 </port>541 </output>542 </layer>543 <layer id="48" name="__module.speecht5.decoder.prenet/aten::bernoulli/Less_1" type="Less" version="opset1">544 <data auto_broadcast="numpy" />545 <input>546 <port id="0" precision="FP32">547 <dim>-1</dim>548 <dim>256</dim>549 </port>550 <port id="1" precision="FP32">551 <dim>1</dim>552 <dim>1</dim>553 </port>554 </input>555 <output>556 <port id="2" precision="BOOL">557 <dim>-1</dim>558 <dim>256</dim>559 </port>560 </output>561 </layer>562 <layer id="49" name="__module.speecht5.decoder.prenet/aten::bernoulli/ConvertLike_1" type="Convert" version="opset1">563 <data destination_type="f32" />564 <input>565 <port id="0" precision="BOOL">566 <dim>-1</dim>567 <dim>256</dim>568 </port>569 </input>570 <output>571 <port id="1" precision="FP32" names="109,123,mask,mask.1">572 <dim>-1</dim>573 <dim>256</dim>574 </port>575 </output>576 </layer>577 <layer id="50" name="96" type="Const" version="opset1">578 <data element_type="i64" shape="" offset="0" size="8" />579 <output>580 <port id="0" precision="I64" names="96" />581 </output>582 </layer>583 <layer id="51" name="__module.speecht5.decoder.prenet/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1">584 <input>585 <port id="0" precision="FP32">586 <dim>-1</dim>587 <dim>256</dim>588 </port>589 <port id="1" precision="I64" />590 </input>591 <output>592 <port id="2" precision="FP32" names="110,124">593 <dim>1</dim>594 <dim>-1</dim>595 <dim>256</dim>596 </port>597 </output>598 </layer>599 <layer id="52" name="ShapeOf_17141" type="ShapeOf" version="opset3">600 <data output_type="i64" />601 <input>602 <port id="0" precision="FP32">603 <dim>-1</dim>604 <dim>-1</dim>605 <dim>256</dim>606 </port>607 </input>608 <output>609 <port id="1" precision="I64">610 <dim>3</dim>611 </port>612 </output>613 </layer>614 <layer id="53" name="Constant_17142" type="Const" version="opset1">615 <data element_type="i64" shape="1" offset="0" size="8" />616 <output>617 <port id="0" precision="I64">618 <dim>1</dim>619 </port>620 </output>621 </layer>622 <layer id="54" name="Constant_17143" type="Const" version="opset1">623 <data element_type="i64" shape="" offset="0" size="8" />624 <output>625 <port id="0" precision="I64" />626 </output>627 </layer>628 <layer id="55" name="Gather_17144" type="Gather" version="opset8">629 <data batch_dims="0" />630 <input>631 <port id="0" precision="I64">632 <dim>3</dim>633 </port>634 <port id="1" precision="I64">635 <dim>1</dim>636 </port>637 <port id="2" precision="I64" />638 </input>639 <output>640 <port id="3" precision="I64" names="111,125,184,206,27,273,29,30,365,432">641 <dim>1</dim>642 </port>643 </output>644 </layer>645 <layer id="56" name="Constant_9641" type="Const" version="opset1">646 <data element_type="i64" shape="1" offset="91188" size="8" />647 <output>648 <port id="0" precision="I64">649 <dim>1</dim>650 </port>651 </output>652 </layer>653 <layer id="57" name="Constant_9643" type="Const" version="opset1">654 <data element_type="i64" shape="1" offset="91188" size="8" />655 <output>656 <port id="0" precision="I64">657 <dim>1</dim>658 </port>659 </output>660 </layer>661 <layer id="58" name="__module.speecht5.decoder.prenet/prim::ListConstruct/Concat" type="Concat" version="opset1">662 <data axis="0" />663 <input>664 <port id="0" precision="I64">665 <dim>1</dim>666 </port>667 <port id="1" precision="I64">668 <dim>1</dim>669 </port>670 <port id="2" precision="I64">671 <dim>1</dim>672 </port>673 </input>674 <output>675 <port id="3" precision="I64" names="112,126">676 <dim>3</dim>677 </port>678 </output>679 </layer>680 <layer id="59" name="__module.speecht5.decoder.prenet/aten::repeat/Tile_1" type="Tile" version="opset1">681 <input>682 <port id="0" precision="FP32">683 <dim>1</dim>684 <dim>-1</dim>685 <dim>256</dim>686 </port>687 <port id="1" precision="I64">688 <dim>3</dim>689 </port>690 </input>691 <output>692 <port id="2" precision="FP32" names="113,127,all_masks,all_masks.1">693 <dim>-1</dim>694 <dim>-1</dim>695 <dim>256</dim>696 </port>697 </output>698 </layer>699 <layer id="60" name="Constant_16649" type="Const" version="opset1">700 <data element_type="f32" shape="1, 1, 1" offset="91180" size="4" />701 <output>702 <port id="0" precision="FP32">703 <dim>1</dim>704 <dim>1</dim>705 <dim>1</dim>706 </port>707 </output>708 </layer>709 <layer id="61" name="__module.speecht5.decoder.prenet/aten::eq/Equal_1" type="Equal" version="opset1">710 <data auto_broadcast="numpy" />711 <input>712 <port id="0" precision="FP32">713 <dim>-1</dim>714 <dim>-1</dim>715 <dim>256</dim>716 </port>717 <port id="1" precision="FP32">718 <dim>1</dim>719 <dim>1</dim>720 <dim>1</dim>721 </port>722 </input>723 <output>724 <port id="2" precision="BOOL" names="114,128">725 <dim>-1</dim>726 <dim>-1</dim>727 <dim>256</dim>728 </port>729 </output>730 </layer>731 <layer id="62" name="__module.speecht5.decoder.prenet/aten::where/Convert_1" type="Const" version="opset1">732 <data element_type="f32" shape="" offset="9224" size="4" />733 <output>734 <port id="0" precision="FP32" />735 </output>736 </layer>737 <layer id="63" name="__module.speecht5.decoder.prenet/aten::where/Select" type="Select" version="opset1">738 <data auto_broadcast="numpy" />739 <input>740 <port id="0" precision="BOOL">741 <dim>-1</dim>742 <dim>-1</dim>743 <dim>256</dim>744 </port>745 <port id="1" precision="FP32">746 <dim>-1</dim>747 <dim>-1</dim>748 <dim>256</dim>749 </port>750 <port id="2" precision="FP32" />751 </input>752 <output>753 <port id="3" precision="FP32" names="115,116">754 <dim>-1</dim>755 <dim>-1</dim>756 <dim>256</dim>757 </port>758 </output>759 </layer>760 <layer id="64" name="Constant_16647" type="Const" version="opset1">761 <data element_type="f32" shape="1, 1, 1" offset="91184" size="4" />762 <output>763 <port id="0" precision="FP32">764 <dim>1</dim>765 <dim>1</dim>766 <dim>1</dim>767 </port>768 </output>769 </layer>770 <layer id="65" name="__module.speecht5.decoder.prenet/aten::div/Divide" type="Divide" version="opset1">771 <data auto_broadcast="numpy" m_pythondiv="true" />772 <input>773 <port id="0" precision="FP32">774 <dim>-1</dim>775 <dim>-1</dim>776 <dim>256</dim>777 </port>778 <port id="1" precision="FP32">779 <dim>1</dim>780 <dim>1</dim>781 <dim>1</dim>782 </port>783 </input>784 <output>785 <port id="2" precision="FP32" names="117">786 <dim>-1</dim>787 <dim>-1</dim>788 <dim>256</dim>789 </port>790 </output>791 </layer>792 <layer id="66" name="self.speecht5.decoder.prenet.layers.1.weight" type="Const" version="opset1">793 <data element_type="f32" shape="256, 256" offset="91196" size="262144" />794 <output>795 <port id="0" precision="FP32" names="self.speecht5.decoder.prenet.layers.1.weight">796 <dim>256</dim>797 <dim>256</dim>798 </port>799 </output>800 </layer>801 <layer id="67" name="__module.speecht5.decoder.prenet.layers.1/aten::linear/MatMul" type="MatMul" version="opset1">802 <data transpose_a="false" transpose_b="true" />803 <input>804 <port id="0" precision="FP32">805 <dim>-1</dim>806 <dim>-1</dim>807 <dim>256</dim>808 </port>809 <port id="1" precision="FP32">810 <dim>256</dim>811 <dim>256</dim>812 </port>813 </input>814 <output>815 <port id="2" precision="FP32" names="120">816 <dim>-1</dim>817 <dim>-1</dim>818 <dim>256</dim>819 </port>820 </output>821 </layer>822 <layer id="68" name="__module.speecht5.decoder.prenet/aten::relu/Relu_1" type="ReLU" version="opset1">823 <input>824 <port id="0" precision="FP32">825 <dim>-1</dim>826 <dim>-1</dim>827 <dim>256</dim>828 </port>829 </input>830 <output>831 <port id="1" precision="FP32" names="121,inputs_embeds.5">832 <dim>-1</dim>833 <dim>-1</dim>834 <dim>256</dim>835 </port>836 </output>837 </layer>838 <layer id="69" name="__module.speecht5.decoder.prenet/aten::where/Select_1" type="Select" version="opset1">839 <data auto_broadcast="numpy" />840 <input>841 <port id="0" precision="BOOL">842 <dim>-1</dim>843 <dim>-1</dim>844 <dim>256</dim>845 </port>846 <port id="1" precision="FP32">847 <dim>-1</dim>848 <dim>-1</dim>849 <dim>256</dim>850 </port>851 <port id="2" precision="FP32" />852 </input>853 <output>854 <port id="3" precision="FP32" names="129,130">855 <dim>-1</dim>856 <dim>-1</dim>857 <dim>256</dim>858 </port>859 </output>860 </layer>861 <layer id="70" name="Constant_16650" type="Const" version="opset1">862 <data element_type="f32" shape="1, 1, 1" offset="91184" size="4" />863 <output>864 <port id="0" precision="FP32">865 <dim>1</dim>866 <dim>1</dim>867 <dim>1</dim>868 </port>869 </output>870 </layer>871 <layer id="71" name="__module.speecht5.decoder.prenet/aten::div/Divide_1" type="Divide" version="opset1">872 <data auto_broadcast="numpy" m_pythondiv="true" />873 <input>874 <port id="0" precision="FP32">875 <dim>-1</dim>876 <dim>-1</dim>877 <dim>256</dim>878 </port>879 <port id="1" precision="FP32">880 <dim>1</dim>881 <dim>1</dim>882 <dim>1</dim>883 </port>884 </input>885 <output>886 <port id="2" precision="FP32" names="131">887 <dim>-1</dim>888 <dim>-1</dim>889 <dim>256</dim>890 </port>891 </output>892 </layer>893 <layer id="72" name="self.speecht5.decoder.prenet.final_layer.weight" type="Const" version="opset1">894 <data element_type="f32" shape="24, 256" offset="353340" size="24576" />895 <output>896 <port id="0" precision="FP32" names="self.speecht5.decoder.prenet.final_layer.weight">897 <dim>24</dim>898 <dim>256</dim>899 </port>900 </output>901 </layer>902 <layer id="73" name="__module.speecht5.decoder.prenet.final_layer/aten::linear/MatMul" type="MatMul" version="opset1">903 <data transpose_a="false" transpose_b="true" />904 <input>905 <port id="0" precision="FP32">906 <dim>-1</dim>907 <dim>-1</dim>908 <dim>256</dim>909 </port>910 <port id="1" precision="FP32">911 <dim>24</dim>912 <dim>256</dim>913 </port>914 </input>915 <output>916 <port id="2" precision="FP32" names="134,emb">917 <dim>-1</dim>918 <dim>-1</dim>919 <dim>24</dim>920 </port>921 </output>922 </layer>923 <layer id="74" name="__module.speecht5.decoder.prenet.encode_positions/aten::slice/Slice" type="Const" version="opset1">924 <data element_type="f32" shape="1, 4000, 24" offset="377916" size="384000" />925 <output>926 <port id="0" precision="FP32" names="138">927 <dim>1</dim>928 <dim>4000</dim>929 <dim>24</dim>930 </port>931 </output>932 </layer>933 <layer id="75" name="Constant_5029" type="Const" version="opset1">934 <data element_type="i64" shape="1" offset="0" size="8" />935 <output>936 <port id="0" precision="I64">937 <dim>1</dim>938 </port>939 </output>940 </layer>941 <layer id="76" name="Constant_17159" type="Const" version="opset1">942 <data element_type="i64" shape="1" offset="91188" size="8" />943 <output>944 <port id="0" precision="I64">945 <dim>1</dim>946 </port>947 </output>948 </layer>949 <layer id="77" name="Constant_17160" type="Const" version="opset1">950 <data element_type="i64" shape="" offset="0" size="8" />951 <output>952 <port id="0" precision="I64" />953 </output>954 </layer>955 <layer id="78" name="Gather_17161" type="Gather" version="opset8">956 <data batch_dims="0" />957 <input>958 <port id="0" precision="I64">959 <dim>3</dim>960 </port>961 <port id="1" precision="I64">962 <dim>1</dim>963 </port>964 <port id="2" precision="I64" />965 </input>966 <output>967 <port id="3" precision="I64" names="137">968 <dim>1</dim>969 </port>970 </output>971 </layer>972 <layer id="79" name="Constant_5032" type="Const" version="opset1">973 <data element_type="i64" shape="1" offset="91188" size="8" />974 <output>975 <port id="0" precision="I64">976 <dim>1</dim>977 </port>978 </output>979 </layer>980 <layer id="80" name="Constant_5027" type="Const" version="opset1">981 <data element_type="i64" shape="1" offset="91188" size="8" />982 <output>983 <port id="0" precision="I64">984 <dim>1</dim>985 </port>986 </output>987 </layer>988 <layer id="81" name="__module.speecht5.decoder.prenet.encode_positions/aten::slice/Slice_1" type="Slice" version="opset8">989 <input>990 <port id="0" precision="FP32">991 <dim>1</dim>992 <dim>4000</dim>993 <dim>24</dim>994 </port>995 <port id="1" precision="I64">996 <dim>1</dim>997 </port>998 <port id="2" precision="I64">999 <dim>1</dim>1000 </port>1001 <port id="3" precision="I64">1002 <dim>1</dim>1003 </port>1004 <port id="4" precision="I64">1005 <dim>1</dim>1006 </port>1007 </input>1008 <output>1009 <port id="5" precision="FP32" names="139,140">1010 <dim>1</dim>1011 <dim>-1</dim>1012 <dim>24</dim>1013 </port>1014 </output>1015 </layer>1016 <layer id="82" name="__module.speecht5.decoder.prenet.encode_positions/aten::add/Add" type="Add" version="opset1">1017 <data auto_broadcast="numpy" />1018 <input>1019 <port id="0" precision="FP32">1020 <dim>-1</dim>1021 <dim>-1</dim>1022 <dim>24</dim>1023 </port>1024 <port id="1" precision="FP32">1025 <dim>1</dim>1026 <dim>-1</dim>1027 <dim>24</dim>1028 </port>1029 </input>1030 <output>1031 <port id="2" precision="FP32" names="141,input.1">1032 <dim>-1</dim>1033 <dim>-1</dim>1034 <dim>24</dim>1035 </port>1036 </output>1037 </layer>1038 <layer id="83" name="Constant_5047" type="Const" version="opset1">1039 <data element_type="i64" shape="1" offset="91188" size="8" />1040 <output>1041 <port id="0" precision="I64" names="143">1042 <dim>1</dim>1043 </port>1044 </output>1045 </layer>1046 <layer id="84" name="__module.speecht5.decoder.prenet/aten::linalg_vector_norm/ReduceL2" type="ReduceL2" version="opset4">1047 <data keep_dims="true" />1048 <input>1049 <port id="0" precision="FP32">1050 <dim>1</dim>1051 <dim>512</dim>1052 </port>1053 <port id="1" precision="I64">1054 <dim>1</dim>1055 </port>1056 </input>1057 <output>1058 <port id="2" precision="FP32" names="144">1059 <dim>1</dim>1060 <dim>1</dim>1061 </port>1062 </output>1063 </layer>1064 <layer id="85" name="Constant_16651" type="Const" version="opset1">1065 <data element_type="f32" shape="1, 1" offset="761916" size="4" />1066 <output>1067 <port id="0" precision="FP32">1068 <dim>1</dim>1069 <dim>1</dim>1070 </port>1071 </output>1072 </layer>1073 <layer id="86" name="__module.speecht5.decoder.prenet/aten::clamp_min/Maximum" type="Maximum" version="opset1">1074 <data auto_broadcast="numpy" />1075 <input>1076 <port id="0" precision="FP32">1077 <dim>1</dim>1078 <dim>1</dim>1079 </port>1080 <port id="1" precision="FP32">1081 <dim>1</dim>1082 <dim>1</dim>1083 </port>1084 </input>1085 <output>1086 <port id="2" precision="FP32" names="145,146,denom">1087 <dim>1</dim>1088 <dim>1</dim>1089 </port>1090 </output>1091 </layer>1092 <layer id="87" name="__module.speecht5.decoder.prenet/aten::div/Divide_2" type="Divide" version="opset1">1093 <data auto_broadcast="numpy" m_pythondiv="true" />1094 <input>1095 <port id="0" precision="FP32">1096 <dim>1</dim>1097 <dim>512</dim>1098 </port>1099 <port id="1" precision="FP32">1100 <dim>1</dim>1101 <dim>1</dim>1102 </port>1103 </input>1104 <output>1105 <port id="2" precision="FP32" names="147,speaker_embeddings_norm">1106 <dim>1</dim>1107 <dim>512</dim>1108 </port>1109 </output>1110 </layer>1111 <layer id="88" name="93" type="Const" version="opset1">1112 <data element_type="i64" shape="" offset="91188" size="8" />1113 <output>1114 <port id="0" precision="I64" names="93" />1115 </output>1116 </layer>1117 <layer id="89" name="__module.speecht5.decoder.prenet/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1">1118 <input>1119 <port id="0" precision="FP32">1120 <dim>1</dim>1121 <dim>512</dim>1122 </port>1123 <port id="1" precision="I64" />1124 </input>1125 <output>1126 <port id="2" precision="FP32" names="148">1127 <dim>1</dim>1128 <dim>1</dim>1129 <dim>512</dim>1130 </port>1131 </output>1132 </layer>1133 <layer id="90" name="Constant_15463" type="Const" version="opset1">1134 <data element_type="i64" shape="1" offset="91188" size="8" />1135 <output>1136 <port id="0" precision="I64">1137 <dim>1</dim>1138 </port>1139 </output>1140 </layer>1141 <layer id="91" name="ShapeOf_17164" type="ShapeOf" version="opset3">1142 <data output_type="i64" />1143 <input>1144 <port id="0" precision="FP32">1145 <dim>-1</dim>1146 <dim>-1</dim>1147 <dim>24</dim>1148 </port>1149 </input>1150 <output>1151 <port id="1" precision="I64">1152 <dim>3</dim>1153 </port>1154 </output>1155 </layer>1156 <layer id="92" name="Constant_17165" type="Const" version="opset1">1157 <data element_type="i64" shape="1" offset="91188" size="8" />1158 <output>1159 <port id="0" precision="I64">1160 <dim>1</dim>1161 </port>1162 </output>1163 </layer>1164 <layer id="93" name="Constant_17166" type="Const" version="opset1">1165 <data element_type="i64" shape="" offset="0" size="8" />1166 <output>1167 <port id="0" precision="I64" />1168 </output>1169 </layer>1170 <layer id="94" name="Gather_17167" type="Gather" version="opset8">1171 <data batch_dims="0" />1172 <input>1173 <port id="0" precision="I64">1174 <dim>3</dim>1175 </port>1176 <port id="1" precision="I64">1177 <dim>1</dim>1178 </port>1179 <port id="2" precision="I64" />1180 </input>1181 <output>1182 <port id="3" precision="I64" names="149">1183 <dim>1</dim>1184 </port>1185 </output>1186 </layer>1187 <layer id="95" name="Constant_15466" type="Const" version="opset1">1188 <data element_type="i64" shape="1" offset="91188" size="8" />1189 <output>1190 <port id="0" precision="I64">1191 <dim>1</dim>1192 </port>1193 </output>1194 </layer>1195 <layer id="96" name="__module.speecht5.decoder.prenet/prim::ListConstruct" type="Concat" version="opset1">1196 <data axis="0" />1197 <input>1198 <port id="0" precision="I64">1199 <dim>1</dim>1200 </port>