File size: 28,452 Bytes
d8e4d0b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "b31a5ac5-a8b7-4ef0-a5d8-41d59e5c37df",
   "metadata": {},
   "outputs": [],
   "source": [
    "from minio import Minio\n",
    "import duckdb\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "5684f535-7bbc-435d-afb2-504cbc647688",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<duckdb.duckdb.DuckDBPyConnection at 0x7fb621719a30>"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# add the access key cell here\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "f8094e79-0538-4771-af47-9a11d1c58d1b",
   "metadata": {},
   "outputs": [],
   "source": [
    "bucket_name = \"ads-with-images\"\n",
    "object_name = \"aug_ads_w_images.parquet\"\n",
    "\n",
    "df = connection.execute(f\"SELECT * FROM parquet_scan('s3://{bucket_name}/{object_name}');\").df()\n",
    "df = df.iloc[:1000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "41567a9b-3b33-4719-b044-7b7264b943f6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>url</th>\n",
       "      <th>title</th>\n",
       "      <th>text</th>\n",
       "      <th>domain</th>\n",
       "      <th>name</th>\n",
       "      <th>description</th>\n",
       "      <th>image</th>\n",
       "      <th>retrieved</th>\n",
       "      <th>production_data</th>\n",
       "      <th>category</th>\n",
       "      <th>...</th>\n",
       "      <th>loc_name</th>\n",
       "      <th>lat</th>\n",
       "      <th>lon</th>\n",
       "      <th>country</th>\n",
       "      <th>label</th>\n",
       "      <th>score</th>\n",
       "      <th>product</th>\n",
       "      <th>label_product</th>\n",
       "      <th>score_product</th>\n",
       "      <th>path</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>https://www.ebay.ie/itm/225712575933?amdata=en...</td>\n",
       "      <td>Mofusand Exhibition Mofusando Piece Set Towel ...</td>\n",
       "      <td>\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nMofusa...</td>\n",
       "      <td>ebay.ie</td>\n",
       "      <td>Mofusand Exhibition Mofusando Piece Set Towel ...</td>\n",
       "      <td>Mofusand Exhibition Mofusando Piece Set Towel ...</td>\n",
       "      <td>https://i.ebayimg.com/images/g/CUQAAOSwatNkz~R...</td>\n",
       "      <td>2023-02-10T17:05:03.245+0000</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>...</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.990042</td>\n",
       "      <td>Mofusand Exhibition Mofusando Piece Set Towel ...</td>\n",
       "      <td>an object</td>\n",
       "      <td>0.454573</td>\n",
       "      <td>images-august/None/e8bf594b-d5bb-43b5-a1b1-18f...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>https://www.ebay.nl/itm/165510736761?amdata=en...</td>\n",
       "      <td>ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...</td>\n",
       "      <td>\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nATLANT...</td>\n",
       "      <td>ebay.nl</td>\n",
       "      <td>ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...</td>\n",
       "      <td>ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...</td>\n",
       "      <td>https://i.ebayimg.com/images/g/m8cAAOSwfbNillr...</td>\n",
       "      <td>2023-02-10T17:05:03.245+0000</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>...</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.977253</td>\n",
       "      <td>ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...</td>\n",
       "      <td>an object</td>\n",
       "      <td>0.728779</td>\n",
       "      <td>images-august/None/286b368a-8adf-46ce-9331-e29...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>https://www.ebay.nl/itm/175641098586?amdata=en...</td>\n",
       "      <td>Mountain Anoa Dwarf Water Buffalo DecoBronze S...</td>\n",
       "      <td>\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nMounta...</td>\n",
       "      <td>ebay.nl</td>\n",
       "      <td>Mountain Anoa Dwarf Water Buffalo DecoBronze S...</td>\n",
       "      <td>Mountain Anoa Dwarf Water Buffalo DecoBronze S...</td>\n",
       "      <td>https://i.ebayimg.com/images/g/VAMAAOSwxNlkBe4...</td>\n",
       "      <td>2023-02-10T17:05:03.245+0000</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>...</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.819862</td>\n",
       "      <td>Mountain Anoa Dwarf Water Buffalo DecoBronze S...</td>\n",
       "      <td>a print of an animal</td>\n",
       "      <td>0.504843</td>\n",
       "      <td>images-august/None/07c9d7ab-0264-44e5-8c1d-934...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>https://www.ebay.com.hk/itm/256119308608?amdat...</td>\n",
       "      <td>Wildlife Research Center 1255 Scent Killer Gol...</td>\n",
       "      <td>\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nWildli...</td>\n",
       "      <td>ebay.com.hk</td>\n",
       "      <td>Wildlife Research Center 1255 Scent Killer Gol...</td>\n",
       "      <td>Wildlife Research Center 1255 Scent Killer Gol...</td>\n",
       "      <td>https://i.ebayimg.com/images/g/w3IAAOSwUp9jOv-...</td>\n",
       "      <td>2023-02-10T17:05:03.245+0000</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>...</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.815647</td>\n",
       "      <td>Wildlife Research Center 1255 Scent Killer Gol...</td>\n",
       "      <td>an object</td>\n",
       "      <td>0.738707</td>\n",
       "      <td>images-august/None/871b28ed-2e92-4c3d-8e1d-2b9...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>https://www.ebay.at/itm/134672898114?amdata=en...</td>\n",
       "      <td>4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...</td>\n",
       "      <td>\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n4 \"Sam...</td>\n",
       "      <td>ebay.at</td>\n",
       "      <td>4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...</td>\n",
       "      <td>4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...</td>\n",
       "      <td>https://i.ebayimg.com/images/g/c3YAAOSwM9NiNFP...</td>\n",
       "      <td>2023-02-10T17:05:03.245+0000</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>...</td>\n",
       "      <td>None</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>None</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.734359</td>\n",
       "      <td>4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...</td>\n",
       "      <td>a print of an animal</td>\n",
       "      <td>0.538278</td>\n",
       "      <td>images-august/None/7bfee649-2c7c-46b9-b5b4-37e...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 28 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 url  \\\n",
       "0  https://www.ebay.ie/itm/225712575933?amdata=en...   \n",
       "1  https://www.ebay.nl/itm/165510736761?amdata=en...   \n",
       "2  https://www.ebay.nl/itm/175641098586?amdata=en...   \n",
       "3  https://www.ebay.com.hk/itm/256119308608?amdat...   \n",
       "4  https://www.ebay.at/itm/134672898114?amdata=en...   \n",
       "\n",
       "                                               title  \\\n",
       "0  Mofusand Exhibition Mofusando Piece Set Towel ...   \n",
       "1  ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...   \n",
       "2  Mountain Anoa Dwarf Water Buffalo DecoBronze S...   \n",
       "3  Wildlife Research Center 1255 Scent Killer Gol...   \n",
       "4  4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...   \n",
       "\n",
       "                                                text       domain  \\\n",
       "0  \\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nMofusa...      ebay.ie   \n",
       "1  \\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nATLANT...      ebay.nl   \n",
       "2  \\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nMounta...      ebay.nl   \n",
       "3  \\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\nWildli...  ebay.com.hk   \n",
       "4  \\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n\\n4 \"Sam...      ebay.at   \n",
       "\n",
       "                                                name  \\\n",
       "0  Mofusand Exhibition Mofusando Piece Set Towel ...   \n",
       "1  ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...   \n",
       "2  Mountain Anoa Dwarf Water Buffalo DecoBronze S...   \n",
       "3  Wildlife Research Center 1255 Scent Killer Gol...   \n",
       "4  4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...   \n",
       "\n",
       "                                         description  \\\n",
       "0  Mofusand Exhibition Mofusando Piece Set Towel ...   \n",
       "1  ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...   \n",
       "2  Mountain Anoa Dwarf Water Buffalo DecoBronze S...   \n",
       "3  Wildlife Research Center 1255 Scent Killer Gol...   \n",
       "4  4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...   \n",
       "\n",
       "                                               image  \\\n",
       "0  https://i.ebayimg.com/images/g/CUQAAOSwatNkz~R...   \n",
       "1  https://i.ebayimg.com/images/g/m8cAAOSwfbNillr...   \n",
       "2  https://i.ebayimg.com/images/g/VAMAAOSwxNlkBe4...   \n",
       "3  https://i.ebayimg.com/images/g/w3IAAOSwUp9jOv-...   \n",
       "4  https://i.ebayimg.com/images/g/c3YAAOSwM9NiNFP...   \n",
       "\n",
       "                      retrieved production_data category  ...  loc_name lat  \\\n",
       "0  2023-02-10T17:05:03.245+0000            None     None  ...      None NaN   \n",
       "1  2023-02-10T17:05:03.245+0000            None     None  ...      None NaN   \n",
       "2  2023-02-10T17:05:03.245+0000            None     None  ...      None NaN   \n",
       "3  2023-02-10T17:05:03.245+0000            None     None  ...      None NaN   \n",
       "4  2023-02-10T17:05:03.245+0000            None     None  ...      None NaN   \n",
       "\n",
       "  lon country label     score  \\\n",
       "0 NaN    None   0.0  0.990042   \n",
       "1 NaN    None   1.0  0.977253   \n",
       "2 NaN    None   0.0  0.819862   \n",
       "3 NaN    None   1.0  0.815647   \n",
       "4 NaN    None   1.0  0.734359   \n",
       "\n",
       "                                             product         label_product  \\\n",
       "0  Mofusand Exhibition Mofusando Piece Set Towel ...             an object   \n",
       "1  ATLANTIC RIGHT WHALE =Endangered= IDENTICAL St...             an object   \n",
       "2  Mountain Anoa Dwarf Water Buffalo DecoBronze S...  a print of an animal   \n",
       "3  Wildlife Research Center 1255 Scent Killer Gol...             an object   \n",
       "4  4 \"Sammeln Sie rote Kupferschnitzerei Manis pe...  a print of an animal   \n",
       "\n",
       "  score_product                                               path  \n",
       "0      0.454573  images-august/None/e8bf594b-d5bb-43b5-a1b1-18f...  \n",
       "1      0.728779  images-august/None/286b368a-8adf-46ce-9331-e29...  \n",
       "2      0.504843  images-august/None/07c9d7ab-0264-44e5-8c1d-934...  \n",
       "3      0.738707  images-august/None/871b28ed-2e92-4c3d-8e1d-2b9...  \n",
       "4      0.538278  images-august/None/7bfee649-2c7c-46b9-b5b4-37e...  \n",
       "\n",
       "[5 rows x 28 columns]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "7f1857c7-e6ad-4de6-8f27-be8e1bbc4bfa",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.utils.data import DataLoader, Dataset, random_split\n",
    "from transformers import BertTokenizer, BertForSequenceClassification, AdamW\n",
    "from torchvision.transforms import transforms\n",
    "import torch.nn as nn\n",
    "import torch\n",
    "from transformers import BertModel, RobertaModel\n",
    "from torchvision.models import resnet50\n",
    "from torch import cuda\n",
    "from PIL import Image\n",
    "import time\n",
    "from transformers import RobertaForSequenceClassification, RobertaTokenizer\n",
    "from torchvision import models\n",
    "import torch.nn.functional as F"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "614ed9c7-890c-4d7c-8510-919edceb2e04",
   "metadata": {},
   "outputs": [],
   "source": [
    "from io import BytesIO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "d520c8f6-4e2b-40a7-bbf0-c788ada1c978",
   "metadata": {},
   "outputs": [],
   "source": [
    "# We will first prepare the dataset for inference\n",
    "class InferenceDataset(Dataset):\n",
    "    def __init__(self, dataframe, tokenizer, max_len, transform=None):\n",
    "        self.dataframe = dataframe\n",
    "        self.tokenizer = tokenizer\n",
    "        self.max_len = max_len\n",
    "        self.transform = transform\n",
    "    \n",
    "    def __len__(self):\n",
    "        return len(self.dataframe)\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        row = self.dataframe.iloc[idx]\n",
    "        text = row['combined_text']\n",
    "        img_path = row['path']\n",
    "        \n",
    "        # Processing text\n",
    "        encoding = self.tokenizer.encode_plus(\n",
    "            text,\n",
    "            add_special_tokens=True,\n",
    "            max_length=self.max_len,\n",
    "            return_token_type_ids=False,\n",
    "            padding='max_length',\n",
    "            return_attention_mask=True,\n",
    "            return_tensors='pt',\n",
    "            truncation=True\n",
    "        )\n",
    "        \n",
    "        bucket_name, _, object_name = img_path.partition('/')\n",
    "        image_data = client.get_object(bucket_name, object_name)\n",
    "        image_bytes = image_data.read()\n",
    "        image = Image.open(BytesIO(image_bytes)).convert(\"RGB\")\n",
    "        \n",
    "        # Processing images\n",
    "        # image = Image.open(img_path).convert(\"RGB\")\n",
    "        if self.transform:\n",
    "            image = self.transform(image)\n",
    "        \n",
    "        return {\n",
    "            'input_ids': encoding['input_ids'].flatten(),\n",
    "            'attention_mask': encoding['attention_mask'].flatten(),\n",
    "            'image': image\n",
    "        }\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "ba1acf51-dfeb-41dc-a925-2e7f7505e989",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Constants\n",
    "MAX_LEN = 128\n",
    "BATCH_SIZE = 16\n",
    "\n",
    "tokenizer = RobertaTokenizer.from_pretrained('roberta-base')\n",
    "\n",
    "transform = transforms.Compose([\n",
    "    transforms.RandomHorizontalFlip(),\n",
    "    transforms.RandomRotation(10),\n",
    "    transforms.RandomResizedCrop(224, scale=(0.8, 1.0)),\n",
    "    transforms.ColorJitter(brightness=0.2, contrast=0.2, saturation=0.2, hue=0.1),\n",
    "    transforms.ToTensor(),\n",
    "])\n",
    "\n",
    "\n",
    "# Combine text fields\n",
    "df['description'] = df['description'].fillna('')\n",
    "df['combined_text'] = df['title'] + ' ' + df['text'] + ' ' + df['description']\n",
    "\n",
    "# Prepare dataset and dataloader for inference\n",
    "inference_dataset = InferenceDataset(\n",
    "    dataframe=df,\n",
    "    tokenizer=tokenizer,  \n",
    "    max_len=MAX_LEN,\n",
    "    transform=transform\n",
    ")\n",
    "\n",
    "inference_dataloader = DataLoader(inference_dataset, batch_size=BATCH_SIZE, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "a779c8bf-fd40-4558-9e15-41599e96ed59",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of RobertaModel were not initialized from the model checkpoint at roberta-base and are newly initialized: ['roberta.pooler.dense.bias', 'roberta.pooler.dense.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
      "/home/jovyan/.local/lib/python3.9/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
      "  warnings.warn(\n",
      "/home/jovyan/.local/lib/python3.9/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=Swin_V2_B_Weights.IMAGENET1K_V1`. You can also use `weights=Swin_V2_B_Weights.DEFAULT` to get the most up-to-date weights.\n",
      "  warnings.warn(msg)\n"
     ]
    }
   ],
   "source": [
    "text = RobertaModel.from_pretrained('roberta-base')\n",
    "\n",
    "img = models.swin_v2_b(pretrained=True)\n",
    "\n",
    "class MultiHeadCrossAttention(nn.Module):\n",
    "    def __init__(self, text_dim, image_dim, num_heads, hidden_dim, output_dim):\n",
    "        super(MultiHeadCrossAttention, self).__init__()\n",
    "        self.num_heads = num_heads\n",
    "        self.head_dim = hidden_dim // num_heads\n",
    "\n",
    "        # These linear layers project the inputs to multiple heads\n",
    "        self.text_query = nn.Linear(text_dim, hidden_dim, bias=False)\n",
    "        self.text_key = nn.Linear(text_dim, hidden_dim, bias=False)\n",
    "        self.text_value = nn.Linear(text_dim, hidden_dim, bias=False)\n",
    "\n",
    "        self.image_query = nn.Linear(image_dim, hidden_dim, bias=False)\n",
    "        self.image_key = nn.Linear(image_dim, hidden_dim, bias=False)\n",
    "        self.image_value = nn.Linear(image_dim, hidden_dim, bias=False)\n",
    "\n",
    "        # Final projection layer\n",
    "        self.out_proj = nn.Linear(hidden_dim, output_dim, bias=False)\n",
    "\n",
    "    def forward(self, text_features, image_features):\n",
    "        Q_text = self.text_query(text_features)\n",
    "        K_text = self.text_key(text_features)\n",
    "        V_text = self.text_value(text_features)\n",
    "\n",
    "        Q_image = self.image_query(image_features)\n",
    "        K_image = self.image_key(image_features)\n",
    "        V_image = self.image_value(image_features)\n",
    "\n",
    "        # Split the hidden dimension into num_heads\n",
    "        Q_text = Q_text.view(Q_text.size(0), -1, self.num_heads, self.head_dim).transpose(1, 2)\n",
    "        K_text = K_text.view(K_text.size(0), -1, self.num_heads, self.head_dim).transpose(1, 2)\n",
    "        V_text = V_text.view(V_text.size(0), -1, self.num_heads, self.head_dim).transpose(1, 2)\n",
    "\n",
    "        Q_image = Q_image.view(Q_image.size(0), -1, self.num_heads, self.head_dim).transpose(1, 2)\n",
    "        K_image = K_image.view(K_image.size(0), -1, self.num_heads, self.head_dim).transpose(1, 2)\n",
    "        V_image = V_image.view(V_image.size(0), -1, self.num_heads, self.head_dim).transpose(1, 2)\n",
    "\n",
    "        # Calculate the attention scores\n",
    "        attn_scores_text_image = torch.matmul(Q_text, K_image.transpose(-1, -2)) / (self.head_dim ** 0.5)\n",
    "        attn_scores_image_text = torch.matmul(Q_image, K_text.transpose(-1, -2)) / (self.head_dim ** 0.5)\n",
    "\n",
    "        # Normalize scores\n",
    "        attn_probs_text_image = F.softmax(attn_scores_text_image, dim=-1)\n",
    "        attn_probs_image_text = F.softmax(attn_scores_image_text, dim=-1)\n",
    "\n",
    "        # Apply attention\n",
    "        attn_output_text_image = torch.matmul(attn_probs_text_image, V_image)\n",
    "        attn_output_image_text = torch.matmul(attn_probs_image_text, V_text)\n",
    "\n",
    "        # Concatenate the results across the heads\n",
    "        attn_output_text_image = attn_output_text_image.transpose(1, 2).contiguous().view(text_features.size(0), -1)\n",
    "        attn_output_image_text = attn_output_image_text.transpose(1, 2).contiguous().view(image_features.size(0), -1)\n",
    "\n",
    "        # Project to output dimension\n",
    "        output_text_image = self.out_proj(attn_output_text_image)\n",
    "        output_image_text = self.out_proj(attn_output_image_text)\n",
    "\n",
    "        return output_text_image, output_image_text\n",
    "\n",
    "\n",
    "class MultiModalModel(nn.Module):\n",
    "            def __init__(self, num_labels):\n",
    "                super(MultiModalModel, self).__init__()\n",
    "\n",
    "                # Load pre-trained models\n",
    "                self.bert = text\n",
    "                self.resnet = img\n",
    "\n",
    "                # Remove the final classification layer of ResNet\n",
    "                self.resnet = nn.Sequential(*list(self.resnet.children())[:-1])\n",
    "                self.mhca = MultiHeadCrossAttention(text_dim=768, image_dim=1024, num_heads=4, hidden_dim=512, output_dim=2048)\n",
    "                \n",
    "                \n",
    "                self.classifier = nn.Sequential(\n",
    "                    nn.Linear(2816, 512), \n",
    "                    nn.ReLU(),\n",
    "                    nn.Dropout(0.2),\n",
    "                    nn.Linear(512, num_labels)\n",
    "                )\n",
    "\n",
    "            def forward(self, input_ids, attention_mask, image):\n",
    "                # Forward pass through BERT\n",
    "                outputs = self.bert(input_ids=input_ids, attention_mask=attention_mask)\n",
    "                text_features = outputs['last_hidden_state'][:, 0, :]  # CLS token output as text feature\n",
    "                \n",
    "                # Forward pass through ResNet\n",
    "                image_features = self.resnet(image)\n",
    "                image_features = image_features.view(image_features.size(0), -1)  # Flatten the output\n",
    "                \n",
    "                if text_features.dim() == 2:\n",
    "                    text_features = text_features.unsqueeze(1)\n",
    "                if image_features.dim() == 2:\n",
    "                    image_features = image_features.unsqueeze(1)\n",
    "                \n",
    "                attended_text, attended_image = self.mhca(text_features, image_features)\n",
    "                \n",
    "                attended_text = attended_text.squeeze(1)  # shape: [16, 768]\n",
    "                attended_image = attended_image.squeeze(1) # shape: [16, 2048]\n",
    "\n",
    "                self.image_projection = torch.nn.Linear(2048, 768).to(device)\n",
    "                attended_image = self.image_projection(attended_image)\n",
    "                combined_features = torch.cat((attended_text, attended_image), dim=-1)\n",
    "\n",
    "                logits = self.classifier(combined_features)\n",
    "\n",
    "                return logits\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "568e1b32-8505-4c2e-9b44-89563ebb7724",
   "metadata": {},
   "outputs": [],
   "source": [
    "        import urllib3, socket\n",
    "        from urllib3.connection import HTTPConnection\n",
    "    \n",
    "        HTTPConnection.default_socket_options = ( \n",
    "            HTTPConnection.default_socket_options + [\n",
    "            (socket.SOL_SOCKET, socket.SO_SNDBUF, 1000000), #1MB in byte\n",
    "            (socket.SOL_SOCKET, socket.SO_RCVBUF, 1000000)\n",
    "        ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "09306857-c737-4380-a377-a6f71b0ef0ff",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to run inference\n",
    "def run_inference(model, dataloader, device):\n",
    "    model.eval()\n",
    "    predictions = []\n",
    "\n",
    "    with torch.no_grad():\n",
    "        for batch in dataloader:\n",
    "            input_ids = batch['input_ids'].to(device)\n",
    "            attention_mask = batch['attention_mask'].to(device)\n",
    "            images = batch['image'].to(device)\n",
    "            \n",
    "            outputs = model(input_ids=input_ids, attention_mask=attention_mask, image=images)\n",
    "            \n",
    "            preds = torch.argmax(outputs, dim=1)\n",
    "            predictions.extend(preds.cpu().numpy())\n",
    "    \n",
    "    return predictions\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e67750b4-b917-4d26-8cfe-19913dc21496",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Initialize an empty model\n",
    "loaded_model = MultiModalModel(num_labels=2)\n",
    "\n",
    "# Load the state dictionary\n",
    "model_load_path = \"./model.pth\"\n",
    "\n",
    "# Check device\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "\n",
    "# Load the model weights\n",
    "if device == torch.device('cpu'):\n",
    "    loaded_model.load_state_dict(torch.load(model_load_path, map_location=device), strict=False)\n",
    "else:\n",
    "    loaded_model.load_state_dict(torch.load(model_load_path), strict=False)\n",
    "\n",
    "# Move model to evaluation mode and to the device\n",
    "loaded_model.eval()\n",
    "loaded_model = loaded_model.to(device)\n",
    "\n",
    "start_time = time.time()\n",
    "\n",
    "# Get the predictions\n",
    "predictions = run_inference(loaded_model, inference_dataloader, device)\n",
    "\n",
    "end_time = time.time()\n",
    "\n",
    "elapsed_time = end_time - start_time\n",
    "print(f\"Inference took {elapsed_time:.2f} seconds\")\n",
    "\n",
    "\n",
    "# Add predictions to the dataframe\n",
    "df['predicted_label'] = predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "296a17d9-f8bd-4e15-bb65-f0a2900a529d",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.head()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}