Hugging Face's logo

OVWeb
/
zipformer-ov

Model card Files
xet
Community
riju19 commited on
Commit
c687768
·
verified ·
1 Parent(s): 5c3d0af

Upload 6 files

Browse files
Files changed (6) hide show
  1. decoder-epoch-30-avg-1-chunk-16-left-128.bin +3 -0
  2. decoder-epoch-30-avg-1-chunk-16-left-128.xml +492 -0
  3. encoder-epoch-30-avg-1-chunk-16-left-128.bin +3 -0
  4. encoder-epoch-30-avg-1-chunk-16-left-128.xml +0 -0
  5. joiner-epoch-30-avg-1-chunk-16-left-128.bin +3 -0
  6. joiner-epoch-30-avg-1-chunk-16-left-128.xml +178 -0
decoder-epoch-30-avg-1-chunk-16-left-128.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54936da274a6634a0e20c0d028f04255c378b8fc34ae82b3cd997015d51c2f88
3
+ size 1045556
decoder-epoch-30-avg-1-chunk-16-left-128.xml ADDED
@@ -0,0 +1,492 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="Model9" version="11">
3
+ <layers>
4
+ <layer id="0" name="y.1" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="i64" />
6
+ <output>
7
+ <port id="0" precision="I64" names="20,y.1,y.5">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="1" name="self.decoder.embedding.weight_compressed" type="Const" version="opset1">
14
+ <data element_type="f16" shape="500, 512" offset="0" size="512000" />
15
+ <output>
16
+ <port id="0" precision="FP16" names="self.decoder.embedding.weight">
17
+ <dim>500</dim>
18
+ <dim>512</dim>
19
+ </port>
20
+ </output>
21
+ </layer>
22
+ <layer id="2" name="self.decoder.embedding.weight" type="Convert" version="opset1">
23
+ <data destination_type="f32" />
24
+ <rt_info>
25
+ <attribute name="decompression" version="0" />
26
+ </rt_info>
27
+ <input>
28
+ <port id="0" precision="FP16">
29
+ <dim>500</dim>
30
+ <dim>512</dim>
31
+ </port>
32
+ </input>
33
+ <output>
34
+ <port id="1" precision="FP32">
35
+ <dim>500</dim>
36
+ <dim>512</dim>
37
+ </port>
38
+ </output>
39
+ </layer>
40
+ <layer id="3" name="Constant_256778" type="Const" version="opset1">
41
+ <data element_type="i64" shape="1, 1" offset="512000" size="8" />
42
+ <output>
43
+ <port id="0" precision="I64">
44
+ <dim>1</dim>
45
+ <dim>1</dim>
46
+ </port>
47
+ </output>
48
+ </layer>
49
+ <layer id="4" name="aten::clamp/Maximum" type="Maximum" version="opset1">
50
+ <data auto_broadcast="numpy" />
51
+ <input>
52
+ <port id="0" precision="I64">
53
+ <dim>-1</dim>
54
+ <dim>-1</dim>
55
+ </port>
56
+ <port id="1" precision="I64">
57
+ <dim>1</dim>
58
+ <dim>1</dim>
59
+ </port>
60
+ </input>
61
+ <output>
62
+ <port id="2" precision="I64" names="22">
63
+ <dim>-1</dim>
64
+ <dim>-1</dim>
65
+ </port>
66
+ </output>
67
+ </layer>
68
+ <layer id="5" name="aten::embedding/Convert" type="Convert" version="opset1">
69
+ <data destination_type="i32" />
70
+ <input>
71
+ <port id="0" precision="I64">
72
+ <dim>-1</dim>
73
+ <dim>-1</dim>
74
+ </port>
75
+ </input>
76
+ <output>
77
+ <port id="1" precision="I32">
78
+ <dim>-1</dim>
79
+ <dim>-1</dim>
80
+ </port>
81
+ </output>
82
+ </layer>
83
+ <layer id="6" name="aten::embedding/Constant" type="Const" version="opset1">
84
+ <data element_type="i32" shape="" offset="512008" size="4" />
85
+ <output>
86
+ <port id="0" precision="I32" />
87
+ </output>
88
+ </layer>
89
+ <layer id="7" name="aten::embedding/Gather" type="Gather" version="opset8">
90
+ <data batch_dims="0" />
91
+ <input>
92
+ <port id="0" precision="FP32">
93
+ <dim>500</dim>
94
+ <dim>512</dim>
95
+ </port>
96
+ <port id="1" precision="I32">
97
+ <dim>-1</dim>
98
+ <dim>-1</dim>
99
+ </port>
100
+ <port id="2" precision="I32" />
101
+ </input>
102
+ <output>
103
+ <port id="3" precision="FP32" names="24">
104
+ <dim>-1</dim>
105
+ <dim>-1</dim>
106
+ <dim>512</dim>
107
+ </port>
108
+ </output>
109
+ </layer>
110
+ <layer id="8" name="Constant_256779" type="Const" version="opset1">
111
+ <data element_type="i64" shape="1, 1" offset="512000" size="8" />
112
+ <output>
113
+ <port id="0" precision="I64">
114
+ <dim>1</dim>
115
+ <dim>1</dim>
116
+ </port>
117
+ </output>
118
+ </layer>
119
+ <layer id="9" name="aten::ge/GreaterEqual" type="GreaterEqual" version="opset1">
120
+ <data auto_broadcast="numpy" />
121
+ <input>
122
+ <port id="0" precision="I64">
123
+ <dim>-1</dim>
124
+ <dim>-1</dim>
125
+ </port>
126
+ <port id="1" precision="I64">
127
+ <dim>1</dim>
128
+ <dim>1</dim>
129
+ </port>
130
+ </input>
131
+ <output>
132
+ <port id="2" precision="BOOL" names="25">
133
+ <dim>-1</dim>
134
+ <dim>-1</dim>
135
+ </port>
136
+ </output>
137
+ </layer>
138
+ <layer id="10" name="16" type="Const" version="opset1">
139
+ <data element_type="i64" shape="" offset="512012" size="8" />
140
+ <output>
141
+ <port id="0" precision="I64" names="16,16_1,16_2" />
142
+ </output>
143
+ </layer>
144
+ <layer id="11" name="aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1">
145
+ <input>
146
+ <port id="0" precision="BOOL">
147
+ <dim>-1</dim>
148
+ <dim>-1</dim>
149
+ </port>
150
+ <port id="1" precision="I64" />
151
+ </input>
152
+ <output>
153
+ <port id="2" precision="BOOL" names="26">
154
+ <dim>-1</dim>
155
+ <dim>-1</dim>
156
+ <dim>1</dim>
157
+ </port>
158
+ </output>
159
+ </layer>
160
+ <layer id="12" name="aten::mul/Convert" type="Convert" version="opset1">
161
+ <data destination_type="f32" />
162
+ <input>
163
+ <port id="0" precision="BOOL">
164
+ <dim>-1</dim>
165
+ <dim>-1</dim>
166
+ <dim>1</dim>
167
+ </port>
168
+ </input>
169
+ <output>
170
+ <port id="1" precision="FP32">
171
+ <dim>-1</dim>
172
+ <dim>-1</dim>
173
+ <dim>1</dim>
174
+ </port>
175
+ </output>
176
+ </layer>
177
+ <layer id="13" name="aten::mul/Multiply" type="Multiply" version="opset1">
178
+ <data auto_broadcast="numpy" />
179
+ <input>
180
+ <port id="0" precision="FP32">
181
+ <dim>-1</dim>
182
+ <dim>-1</dim>
183
+ <dim>512</dim>
184
+ </port>
185
+ <port id="1" precision="FP32">
186
+ <dim>-1</dim>
187
+ <dim>-1</dim>
188
+ <dim>1</dim>
189
+ </port>
190
+ </input>
191
+ <output>
192
+ <port id="2" precision="FP32" names="27_1">
193
+ <dim>-1</dim>
194
+ <dim>-1</dim>
195
+ <dim>512</dim>
196
+ </port>
197
+ </output>
198
+ </layer>
199
+ <layer id="14" name="Constant_254157" type="Const" version="opset1">
200
+ <data element_type="i64" shape="3" offset="512020" size="24" />
201
+ <output>
202
+ <port id="0" precision="I64" names="31">
203
+ <dim>3</dim>
204
+ </port>
205
+ </output>
206
+ </layer>
207
+ <layer id="15" name="aten::permute/Transpose" type="Transpose" version="opset1">
208
+ <input>
209
+ <port id="0" precision="FP32">
210
+ <dim>-1</dim>
211
+ <dim>-1</dim>
212
+ <dim>512</dim>
213
+ </port>
214
+ <port id="1" precision="I64">
215
+ <dim>3</dim>
216
+ </port>
217
+ </input>
218
+ <output>
219
+ <port id="2" precision="FP32" names="32_1">
220
+ <dim>-1</dim>
221
+ <dim>512</dim>
222
+ <dim>-1</dim>
223
+ </port>
224
+ </output>
225
+ </layer>
226
+ <layer id="16" name="aten::conv1d/Reshape_compressed" type="Const" version="opset1">
227
+ <data element_type="f16" shape="128, 4, 4, 2" offset="512044" size="8192" />
228
+ <output>
229
+ <port id="0" precision="FP16">
230
+ <dim>128</dim>
231
+ <dim>4</dim>
232
+ <dim>4</dim>
233
+ <dim>2</dim>
234
+ </port>
235
+ </output>
236
+ </layer>
237
+ <layer id="17" name="aten::conv1d/Reshape" type="Convert" version="opset1">
238
+ <data destination_type="f32" />
239
+ <rt_info>
240
+ <attribute name="decompression" version="0" />
241
+ </rt_info>
242
+ <input>
243
+ <port id="0" precision="FP16">
244
+ <dim>128</dim>
245
+ <dim>4</dim>
246
+ <dim>4</dim>
247
+ <dim>2</dim>
248
+ </port>
249
+ </input>
250
+ <output>
251
+ <port id="1" precision="FP32">
252
+ <dim>128</dim>
253
+ <dim>4</dim>
254
+ <dim>4</dim>
255
+ <dim>2</dim>
256
+ </port>
257
+ </output>
258
+ </layer>
259
+ <layer id="18" name="aten::conv1d/GroupConvolution" type="GroupConvolution" version="opset1">
260
+ <data strides="1" pads_begin="0" pads_end="0" dilations="1" auto_pad="explicit" />
261
+ <input>
262
+ <port id="0" precision="FP32">
263
+ <dim>-1</dim>
264
+ <dim>512</dim>
265
+ <dim>-1</dim>
266
+ </port>
267
+ <port id="1" precision="FP32">
268
+ <dim>128</dim>
269
+ <dim>4</dim>
270
+ <dim>4</dim>
271
+ <dim>2</dim>
272
+ </port>
273
+ </input>
274
+ <output>
275
+ <port id="2" precision="FP32" names="48,embedding_out.31">
276
+ <dim>-1</dim>
277
+ <dim>512</dim>
278
+ <dim>-1</dim>
279
+ </port>
280
+ </output>
281
+ </layer>
282
+ <layer id="19" name="Constant_254287" type="Const" version="opset1">
283
+ <data element_type="i64" shape="3" offset="512020" size="24" />
284
+ <output>
285
+ <port id="0" precision="I64" names="49">
286
+ <dim>3</dim>
287
+ </port>
288
+ </output>
289
+ </layer>
290
+ <layer id="20" name="aten::permute/Transpose_1" type="Transpose" version="opset1">
291
+ <input>
292
+ <port id="0" precision="FP32">
293
+ <dim>-1</dim>
294
+ <dim>512</dim>
295
+ <dim>-1</dim>
296
+ </port>
297
+ <port id="1" precision="I64">
298
+ <dim>3</dim>
299
+ </port>
300
+ </input>
301
+ <output>
302
+ <port id="2" precision="FP32" names="50,embedding_out.37">
303
+ <dim>-1</dim>
304
+ <dim>-1</dim>
305
+ <dim>512</dim>
306
+ </port>
307
+ </output>
308
+ </layer>
309
+ <layer id="21" name="aten::relu/Relu" type="ReLU" version="opset1">
310
+ <input>
311
+ <port id="0" precision="FP32">
312
+ <dim>-1</dim>
313
+ <dim>-1</dim>
314
+ <dim>512</dim>
315
+ </port>
316
+ </input>
317
+ <output>
318
+ <port id="1" precision="FP32" names="51,result.3">
319
+ <dim>-1</dim>
320
+ <dim>-1</dim>
321
+ <dim>512</dim>
322
+ </port>
323
+ </output>
324
+ </layer>
325
+ <layer id="22" name="3" type="Const" version="opset1">
326
+ <data element_type="i64" shape="" offset="520236" size="8" />
327
+ <output>
328
+ <port id="0" precision="I64" names="3" />
329
+ </output>
330
+ </layer>
331
+ <layer id="23" name="aten::squeeze/Squeeze" type="Squeeze" version="opset1">
332
+ <input>
333
+ <port id="0" precision="FP32">
334
+ <dim>-1</dim>
335
+ <dim>-1</dim>
336
+ <dim>512</dim>
337
+ </port>
338
+ <port id="1" precision="I64" />
339
+ </input>
340
+ <output>
341
+ <port id="2" precision="FP32" names="6,decoder_output.5">
342
+ <dim>-1</dim>
343
+ <dim>512</dim>
344
+ </port>
345
+ </output>
346
+ </layer>
347
+ <layer id="24" name="self.decoder_proj.weight_compressed" type="Const" version="opset1">
348
+ <data element_type="f16" shape="512, 512" offset="520244" size="524288" />
349
+ <output>
350
+ <port id="0" precision="FP16" names="self.decoder_proj.weight">
351
+ <dim>512</dim>
352
+ <dim>512</dim>
353
+ </port>
354
+ </output>
355
+ </layer>
356
+ <layer id="25" name="self.decoder_proj.weight" type="Convert" version="opset1">
357
+ <data destination_type="f32" />
358
+ <rt_info>
359
+ <attribute name="decompression" version="0" />
360
+ </rt_info>
361
+ <input>
362
+ <port id="0" precision="FP16">
363
+ <dim>512</dim>
364
+ <dim>512</dim>
365
+ </port>
366
+ </input>
367
+ <output>
368
+ <port id="1" precision="FP32">
369
+ <dim>512</dim>
370
+ <dim>512</dim>
371
+ </port>
372
+ </output>
373
+ </layer>
374
+ <layer id="26" name="aten::linear/MatMul" type="MatMul" version="opset1">
375
+ <data transpose_a="false" transpose_b="true" />
376
+ <input>
377
+ <port id="0" precision="FP32">
378
+ <dim>-1</dim>
379
+ <dim>512</dim>
380
+ </port>
381
+ <port id="1" precision="FP32">
382
+ <dim>512</dim>
383
+ <dim>512</dim>
384
+ </port>
385
+ </input>
386
+ <output>
387
+ <port id="2" precision="FP32">
388
+ <dim>-1</dim>
389
+ <dim>512</dim>
390
+ </port>
391
+ </output>
392
+ </layer>
393
+ <layer id="27" name="Constant_256780_compressed" type="Const" version="opset1">
394
+ <data element_type="f16" shape="1, 512" offset="1044532" size="1024" />
395
+ <output>
396
+ <port id="0" precision="FP16">
397
+ <dim>1</dim>
398
+ <dim>512</dim>
399
+ </port>
400
+ </output>
401
+ </layer>
402
+ <layer id="28" name="Constant_256780" type="Convert" version="opset1">
403
+ <data destination_type="f32" />
404
+ <rt_info>
405
+ <attribute name="decompression" version="0" />
406
+ </rt_info>
407
+ <input>
408
+ <port id="0" precision="FP16">
409
+ <dim>1</dim>
410
+ <dim>512</dim>
411
+ </port>
412
+ </input>
413
+ <output>
414
+ <port id="1" precision="FP32">
415
+ <dim>1</dim>
416
+ <dim>512</dim>
417
+ </port>
418
+ </output>
419
+ </layer>
420
+ <layer id="29" name="aten::linear/Add" type="Add" version="opset1">
421
+ <data auto_broadcast="numpy" />
422
+ <input>
423
+ <port id="0" precision="FP32">
424
+ <dim>-1</dim>
425
+ <dim>512</dim>
426
+ </port>
427
+ <port id="1" precision="FP32">
428
+ <dim>1</dim>
429
+ <dim>512</dim>
430
+ </port>
431
+ </input>
432
+ <output>
433
+ <port id="2" precision="FP32" names="output.1">
434
+ <dim>-1</dim>
435
+ <dim>512</dim>
436
+ </port>
437
+ </output>
438
+ </layer>
439
+ <layer id="30" name="Result_254756" type="Result" version="opset1" output_names="output.1">
440
+ <input>
441
+ <port id="0" precision="FP32">
442
+ <dim>-1</dim>
443
+ <dim>512</dim>
444
+ </port>
445
+ </input>
446
+ </layer>
447
+ </layers>
448
+ <edges>
449
+ <edge from-layer="0" from-port="0" to-layer="4" to-port="0" />
450
+ <edge from-layer="0" from-port="0" to-layer="9" to-port="0" />
451
+ <edge from-layer="1" from-port="0" to-layer="2" to-port="0" />
452
+ <edge from-layer="2" from-port="1" to-layer="7" to-port="0" />
453
+ <edge from-layer="3" from-port="0" to-layer="4" to-port="1" />
454
+ <edge from-layer="4" from-port="2" to-layer="5" to-port="0" />
455
+ <edge from-layer="5" from-port="1" to-layer="7" to-port="1" />
456
+ <edge from-layer="6" from-port="0" to-layer="7" to-port="2" />
457
+ <edge from-layer="7" from-port="3" to-layer="13" to-port="0" />
458
+ <edge from-layer="8" from-port="0" to-layer="9" to-port="1" />
459
+ <edge from-layer="9" from-port="2" to-layer="11" to-port="0" />
460
+ <edge from-layer="10" from-port="0" to-layer="11" to-port="1" />
461
+ <edge from-layer="11" from-port="2" to-layer="12" to-port="0" />
462
+ <edge from-layer="12" from-port="1" to-layer="13" to-port="1" />
463
+ <edge from-layer="13" from-port="2" to-layer="15" to-port="0" />
464
+ <edge from-layer="14" from-port="0" to-layer="15" to-port="1" />
465
+ <edge from-layer="15" from-port="2" to-layer="18" to-port="0" />
466
+ <edge from-layer="16" from-port="0" to-layer="17" to-port="0" />
467
+ <edge from-layer="17" from-port="1" to-layer="18" to-port="1" />
468
+ <edge from-layer="18" from-port="2" to-layer="20" to-port="0" />
469
+ <edge from-layer="19" from-port="0" to-layer="20" to-port="1" />
470
+ <edge from-layer="20" from-port="2" to-layer="21" to-port="0" />
471
+ <edge from-layer="21" from-port="1" to-layer="23" to-port="0" />
472
+ <edge from-layer="22" from-port="0" to-layer="23" to-port="1" />
473
+ <edge from-layer="23" from-port="2" to-layer="26" to-port="0" />
474
+ <edge from-layer="24" from-port="0" to-layer="25" to-port="0" />
475
+ <edge from-layer="25" from-port="1" to-layer="26" to-port="1" />
476
+ <edge from-layer="26" from-port="2" to-layer="29" to-port="0" />
477
+ <edge from-layer="27" from-port="0" to-layer="28" to-port="0" />
478
+ <edge from-layer="28" from-port="1" to-layer="29" to-port="1" />
479
+ <edge from-layer="29" from-port="2" to-layer="30" to-port="0" />
480
+ </edges>
481
+ <rt_info>
482
+ <Runtime_version value="2025.2.0-18688-37496659f41" />
483
+ <conversion_parameters>
484
+ <framework value="pytorch" />
485
+ <is_python_object value="True" />
486
+ </conversion_parameters>
487
+ <metadata>
488
+ <context_size value="2" />
489
+ <vocab_size value="500" />
490
+ </metadata>
491
+ </rt_info>
492
+ </net>
encoder-epoch-30-avg-1-chunk-16-left-128.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c211ce1aa94c710050757deff797ec7c066065b1d049e59ede0cfaebceceee35
3
+ size 129495373
encoder-epoch-30-avg-1-chunk-16-left-128.xml ADDED
The diff for this file is too large to render. See raw diff
 
joiner-epoch-30-avg-1-chunk-16-left-128.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:876cfded52cc6cee2258bef21a9a9f46d03820a6b9af54c129573876a43aef98
3
+ size 513000
joiner-epoch-30-avg-1-chunk-16-left-128.xml ADDED
@@ -0,0 +1,178 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="Model12" version="11">
3
+ <layers>
4
+ <layer id="1" name="encoder_out" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="f32" />
6
+ <output>
7
+ <port id="0" precision="FP32" names="encoder_out">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="0" name="decoder_out" type="Parameter" version="opset1">
14
+ <data shape="?,?" element_type="f32" />
15
+ <output>
16
+ <port id="0" precision="FP32" names="decoder_out">
17
+ <dim>-1</dim>
18
+ <dim>-1</dim>
19
+ </port>
20
+ </output>
21
+ </layer>
22
+ <layer id="2" name="aten::add/Add" type="Add" version="opset1">
23
+ <data auto_broadcast="numpy" />
24
+ <input>
25
+ <port id="0" precision="FP32">
26
+ <dim>-1</dim>
27
+ <dim>-1</dim>
28
+ </port>
29
+ <port id="1" precision="FP32">
30
+ <dim>-1</dim>
31
+ <dim>-1</dim>
32
+ </port>
33
+ </input>
34
+ <output>
35
+ <port id="2" precision="FP32" names="5,logit">
36
+ <dim>-1</dim>
37
+ <dim>-1</dim>
38
+ </port>
39
+ </output>
40
+ </layer>
41
+ <layer id="3" name="aten::tanh/Tanh" type="Tanh" version="opset1">
42
+ <input>
43
+ <port id="0" precision="FP32">
44
+ <dim>-1</dim>
45
+ <dim>-1</dim>
46
+ </port>
47
+ </input>
48
+ <output>
49
+ <port id="1" precision="FP32" names="6,input">
50
+ <dim>-1</dim>
51
+ <dim>-1</dim>
52
+ </port>
53
+ </output>
54
+ </layer>
55
+ <layer id="4" name="self.output_linear.weight_compressed" type="Const" version="opset1">
56
+ <data element_type="f16" shape="500, 512" offset="0" size="512000" />
57
+ <output>
58
+ <port id="0" precision="FP16" names="self.output_linear.weight">
59
+ <dim>500</dim>
60
+ <dim>512</dim>
61
+ </port>
62
+ </output>
63
+ </layer>
64
+ <layer id="5" name="self.output_linear.weight" type="Convert" version="opset1">
65
+ <data destination_type="f32" />
66
+ <rt_info>
67
+ <attribute name="decompression" version="0" />
68
+ </rt_info>
69
+ <input>
70
+ <port id="0" precision="FP16">
71
+ <dim>500</dim>
72
+ <dim>512</dim>
73
+ </port>
74
+ </input>
75
+ <output>
76
+ <port id="1" precision="FP32">
77
+ <dim>500</dim>
78
+ <dim>512</dim>
79
+ </port>
80
+ </output>
81
+ </layer>
82
+ <layer id="6" name="__module.output_linear/aten::linear/MatMul" type="MatMul" version="opset1">
83
+ <data transpose_a="false" transpose_b="true" />
84
+ <input>
85
+ <port id="0" precision="FP32">
86
+ <dim>-1</dim>
87
+ <dim>-1</dim>
88
+ </port>
89
+ <port id="1" precision="FP32">
90
+ <dim>500</dim>
91
+ <dim>512</dim>
92
+ </port>
93
+ </input>
94
+ <output>
95
+ <port id="2" precision="FP32">
96
+ <dim>-1</dim>
97
+ <dim>500</dim>
98
+ </port>
99
+ </output>
100
+ </layer>
101
+ <layer id="7" name="Constant_258850_compressed" type="Const" version="opset1">
102
+ <data element_type="f16" shape="1, 500" offset="512000" size="1000" />
103
+ <output>
104
+ <port id="0" precision="FP16">
105
+ <dim>1</dim>
106
+ <dim>500</dim>
107
+ </port>
108
+ </output>
109
+ </layer>
110
+ <layer id="8" name="Constant_258850" type="Convert" version="opset1">
111
+ <data destination_type="f32" />
112
+ <rt_info>
113
+ <attribute name="decompression" version="0" />
114
+ </rt_info>
115
+ <input>
116
+ <port id="0" precision="FP16">
117
+ <dim>1</dim>
118
+ <dim>500</dim>
119
+ </port>
120
+ </input>
121
+ <output>
122
+ <port id="1" precision="FP32">
123
+ <dim>1</dim>
124
+ <dim>500</dim>
125
+ </port>
126
+ </output>
127
+ </layer>
128
+ <layer id="9" name="__module.output_linear/aten::linear/Add" type="Add" version="opset1">
129
+ <data auto_broadcast="numpy" />
130
+ <input>
131
+ <port id="0" precision="FP32">
132
+ <dim>-1</dim>
133
+ <dim>500</dim>
134
+ </port>
135
+ <port id="1" precision="FP32">
136
+ <dim>1</dim>
137
+ <dim>500</dim>
138
+ </port>
139
+ </input>
140
+ <output>
141
+ <port id="2" precision="FP32">
142
+ <dim>-1</dim>
143
+ <dim>500</dim>
144
+ </port>
145
+ </output>
146
+ </layer>
147
+ <layer id="10" name="Result_256965" type="Result" version="opset1">
148
+ <input>
149
+ <port id="0" precision="FP32">
150
+ <dim>-1</dim>
151
+ <dim>500</dim>
152
+ </port>
153
+ </input>
154
+ </layer>
155
+ </layers>
156
+ <edges>
157
+ <edge from-layer="0" from-port="0" to-layer="2" to-port="1" />
158
+ <edge from-layer="1" from-port="0" to-layer="2" to-port="0" />
159
+ <edge from-layer="2" from-port="2" to-layer="3" to-port="0" />
160
+ <edge from-layer="3" from-port="1" to-layer="6" to-port="0" />
161
+ <edge from-layer="4" from-port="0" to-layer="5" to-port="0" />
162
+ <edge from-layer="5" from-port="1" to-layer="6" to-port="1" />
163
+ <edge from-layer="6" from-port="2" to-layer="9" to-port="0" />
164
+ <edge from-layer="7" from-port="0" to-layer="8" to-port="0" />
165
+ <edge from-layer="8" from-port="1" to-layer="9" to-port="1" />
166
+ <edge from-layer="9" from-port="2" to-layer="10" to-port="0" />
167
+ </edges>
168
+ <rt_info>
169
+ <Runtime_version value="2025.2.0-18688-37496659f41" />
170
+ <conversion_parameters>
171
+ <framework value="pytorch" />
172
+ <is_python_object value="True" />
173
+ </conversion_parameters>
174
+ <metadata>
175
+ <joiner_dim value="512" />
176
+ </metadata>
177
+ </rt_info>
178
+ </net>