Summarization
Transformers
Safetensors
File size: 23,830 Bytes
b2b0691
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "# Setting up the environment"
      ],
      "metadata": {
        "id": "g0P2N4KjS4Dz"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "wkVgiRJtStA9"
      },
      "outputs": [],
      "source": [
        "!pip install pymupdf --quiet"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import re\n",
        "import warnings\n",
        "\n",
        "import fitz\n",
        "import nltk\n",
        "import torch\n",
        "from nltk.tokenize import sent_tokenize\n",
        "from peft import AutoPeftModelForSeq2SeqLM\n",
        "from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, logging\n",
        "\n",
        "CKPT_PATH = \"facebook/bart-large-cnn\" # Path to the fine-tuned base model\n",
        "HF_REPO_PATH = \"spolivin/bart-arxiv-lora\" # Path to the repo where LoRA adapters are saved\n",
        "\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "logging.set_verbosity_error()\n",
        "nltk.download(\"punkt_tab\")\n",
        "nltk.download(\"punkt\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XfYs8aczk6QN",
        "outputId": "4be63c1f-450a-4627-c1f6-c5f5872e30c0"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "[nltk_data] Downloading package punkt_tab to /root/nltk_data...\n",
            "[nltk_data]   Package punkt_tab is already up-to-date!\n",
            "[nltk_data] Downloading package punkt to /root/nltk_data...\n",
            "[nltk_data]   Package punkt is already up-to-date!\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {},
          "execution_count": 2
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Loading models to be tested"
      ],
      "metadata": {
        "id": "MHUC9wjpS7L8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "tokenizer = AutoTokenizer.from_pretrained(CKPT_PATH)\n",
        "original_model = AutoModelForSeq2SeqLM.from_pretrained(CKPT_PATH)\n",
        "\n",
        "lora_model = AutoPeftModelForSeq2SeqLM.from_pretrained(HF_REPO_PATH)"
      ],
      "metadata": {
        "id": "LB4MtAqOSupA"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Defining functions for retrieving/preprocessing text from PDF"
      ],
      "metadata": {
        "id": "Lm_D9hkDTtOU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def prettify_summary(summary):\n",
        "    \"\"\"Prettifies the summary.\"\"\"\n",
        "    # Spliting input into sentences and capitalizing each one\n",
        "    sentences = sent_tokenize(summary)\n",
        "    prettified_summary = \" \".join(s.capitalize() for s in sentences)\n",
        "\n",
        "    # Removing unwanted spaces before punctuation\n",
        "    prettified_summary = re.sub(r'\\s+([.,!?])', r'\\1', prettified_summary)\n",
        "\n",
        "    return prettified_summary"
      ],
      "metadata": {
        "id": "jIACNyS_Sum_"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def summarize_text(\n",
        "    model,\n",
        "    tokenizer,\n",
        "    text,\n",
        "    device=\"cuda\",\n",
        "    max_input_length=1024,\n",
        "    max_output_length=250,\n",
        "):\n",
        "    \"\"\"Tokenizes and generates a summary based on input text.\"\"\"\n",
        "    inputs = tokenizer(\n",
        "        text,\n",
        "        return_tensors=\"pt\",\n",
        "        truncation=True,\n",
        "        max_length=max_input_length\n",
        "    ).to(device)\n",
        "\n",
        "    summary_ids = model.generate(\n",
        "      **inputs,\n",
        "      max_length=max_output_length,\n",
        "      num_beams=4,\n",
        "      early_stopping=True\n",
        "    )\n",
        "\n",
        "    # Decoding and prettifying output\n",
        "    raw_summary = tokenizer.decode(summary_ids[0], skip_special_tokens=True)\n",
        "\n",
        "    return prettify_summary(raw_summary)"
      ],
      "metadata": {
        "id": "rwdQqpxySuki"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def extract_main_text(pdf_path: str, beginning=\"introduction\") -> str:\n",
        "    \"\"\"Retrieves the article text after abstract and before references.\n",
        "\n",
        "       Args:\n",
        "            pdf_path (str): Path to the pdf-file with the article.\n",
        "            beginning (str, optional): The beginning from which to start retrieving text.\n",
        "                                      Defaults to \"introduction\".\n",
        "       Returns:\n",
        "            str: Extracted text.\n",
        "    \"\"\"\n",
        "    # Extracting text from each page\n",
        "    doc = fitz.open(pdf_path)\n",
        "    full_text = \"\"\n",
        "    for page in doc:\n",
        "        full_text += page.get_text(\"text\") + \"\\n\"\n",
        "\n",
        "    # Finding the start of the main text (after Abstract)\n",
        "    abstract_end_idx = full_text.lower().find(beginning) + len(beginning)\n",
        "    if abstract_end_idx == -1:  # If \"Introduction\" not found, falling back to first line after \"Abstract\"\n",
        "        abstract_end_idx = full_text.lower().find(\"abstract\") + len(\"abstract\")\n",
        "\n",
        "    # Finding the end of the main text (before References)\n",
        "    references_idx = full_text.lower().find(\"references\")\n",
        "    if references_idx == -1:  # If \"References\" section not found, extracting till the end\n",
        "        references_idx = len(full_text)\n",
        "\n",
        "    # Extract only the part between Abstract and References\n",
        "    main_text = full_text[abstract_end_idx:references_idx].strip()\n",
        "    return main_text"
      ],
      "metadata": {
        "id": "s8ZbvKaMSuh1"
      },
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def clean_text(text):\n",
        "    \"\"\"Cleans the extracted paper text.\"\"\"\n",
        "    text = re.sub(r'\\s+', ' ', text)\n",
        "\n",
        "    return text.strip()"
      ],
      "metadata": {
        "id": "001u8wp8S1Zh"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Generating summaries on sample articles"
      ],
      "metadata": {
        "id": "HwDjXitFT1ta"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "**NOTE:** *In order to test the summarization models one needs to firstly download the articles in pdf and upload to Google Colab*:\n",
        "\n",
        "* [Efficient Estimation of Word Representations in Vector Space](https://arxiv.org/abs/1301.3781)\n",
        "* [Attention Is All You Need](https://arxiv.org/abs/1706.03762)\n",
        "* [Recurrent Neural Networks (RNNs): A gentle Introduction and Overview](https://arxiv.org/abs/1912.05911)"
      ],
      "metadata": {
        "id": "CMBfxKkVkKBj"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Article 1: [Efficient Estimation of Word Representations in Vector Space](https://arxiv.org/abs/1301.3781)"
      ],
      "metadata": {
        "id": "cOlVk8OyT5eO"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Retrieving article text"
      ],
      "metadata": {
        "id": "RH-HLD1eXU2w"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "article_pdf_path_1 = \"/content/1301.3781v3.pdf\""
      ],
      "metadata": {
        "id": "w3H7a1m0hO9z"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Example usage\n",
        "article_text = extract_main_text(pdf_path=article_pdf_path_1, beginning=\"introduction\")\n",
        "article_text_cleaned = clean_text(article_text)\n",
        "article_text_cleaned[:1000]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        },
        "id": "Pgb4GhWSm9KL",
        "outputId": "42b6003a-3fe2-4871-f799-007b1b8a886a"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Many current NLP systems and techniques treat words as atomic units - there is no notion of similar- ity between words, as these are represented as indices in a vocabulary. This choice has several good reasons - simplicity, robustness and the observation that simple models trained on huge amounts of data outperform complex systems trained on less data. An example is the popular N-gram model used for statistical language modeling - today, it is possible to train N-grams on virtually all available data (trillions of words [3]). However, the simple techniques are at their limits in many tasks. For example, the amount of relevant in-domain data for automatic speech recognition is limited - the performance is usually dominated by the size of high quality transcribed speech data (often just millions of words). In machine translation, the existing corpora for many languages contain only a few billions of words or less. Thus, there are situations where simple scaling up of the basic techniques'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Generating summaries"
      ],
      "metadata": {
        "id": "JJ7r-WOCXZ3y"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "lora_summary = summarize_text(model=lora_model.to(\"cuda\"), tokenizer=tokenizer, text=article_text_cleaned)\n",
        "lora_summary"
      ],
      "metadata": {
        "id": "q8-6eVcsS6H0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        },
        "outputId": "ffeafb71-ea36-43fb-b2bf-9551d4ffa7ff"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'We develop new techniques for learning high-quality word vectors from huge data sets with billions of words, and with millions of words in the vocabulary. We use recently proposed techniques for measuring the quality of the resulting vector representa- tions, with the expectation that not only will similar words tend to be close to each other, but that words can have multiple degrees of similarity. We design a new comprehensive test set for measuring both syntactic and semantic regularities1, and show that many such regularities can be learned with high accuracy. We discuss how training time and accuracy depends on the dimensionality of the word vectors and on the amount of the training data.'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "original_summary = summarize_text(model=original_model.to(\"cuda\"), tokenizer=tokenizer, text=article_text_cleaned)\n",
        "original_summary"
      ],
      "metadata": {
        "id": "IbLkYIRVS6Fu",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 87
        },
        "outputId": "f5ba105d-cff8-43fc-e6e6-9aa090f714ec"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Many current nlp systems and techniques treat words as atomic units - there is no notion of similar- ity between words. This choice has several good reasons - simplicity, robustness and the observation that simple models trained on huge amounts of data outperform complex systems trained on less data. The main goal of this paper is to introduce techniques that can be used for learning high-quality word vectors from huge data sets with billions of words.'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Article 2: [Attention Is All You Need](https://arxiv.org/abs/1706.03762)"
      ],
      "metadata": {
        "id": "w-FlJMTlT8cy"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "article_pdf_path_2 = \"/content/1706.03762v7.pdf\""
      ],
      "metadata": {
        "id": "8RCD2sWChM-g"
      },
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Example usage\n",
        "article_text = extract_main_text(article_pdf_path_2)\n",
        "article_text_cleaned = clean_text(article_text)\n",
        "article_text_cleaned[:1000]"
      ],
      "metadata": {
        "id": "7f54JBALonIT",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        },
        "outputId": "4007007f-9ec2-461e-94ec-778a8732f239"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Recurrent neural networks, long short-term memory [13] and gated recurrent [7] neural networks in particular, have been firmly established as state of the art approaches in sequence modeling and transduction problems such as language modeling and machine translation [35, 2, 5]. Numerous efforts have since continued to push the boundaries of recurrent language models and encoder-decoder architectures [38, 24, 15]. Recurrent models typically factor computation along the symbol positions of the input and output sequences. Aligning the positions to steps in computation time, they generate a sequence of hidden states ht, as a function of the previous hidden state ht−1 and the input for position t. This inherently sequential nature precludes parallelization within training examples, which becomes critical at longer sequence lengths, as memory constraints limit batching across examples. Recent work has achieved significant improvements in computational efficiency through factorization tricks '"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "lora_summary = summarize_text(model=lora_model.to(\"cuda\"), tokenizer=tokenizer, text=article_text_cleaned)\n",
        "lora_summary"
      ],
      "metadata": {
        "id": "0ac3VOgvT-AN",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        },
        "outputId": "d318c5b1-dc39-492e-d04c-f99c41e817ca"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'The transformer is the first transduction model relying entirely on self-attention to compute representations of its input and output without using sequence- aligned rnns or convolutional neural networks as basic building block, computing hidden representations in parallel for all input andoutput positions. The transformer allows for significantly more parallelization and can reach a new state of the art in translation quality after being trained for as little as twelve hours on eight p100 gpus. In the following sections, we describe the model architecture, motivate selfattention and discuss its advantages over models such as [17, 18] and convs2s [9].'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "original_summary = summarize_text(model=original_model.to(\"cuda\"), tokenizer=tokenizer, text=article_text_cleaned)\n",
        "original_summary"
      ],
      "metadata": {
        "id": "DrFG4oJVT993",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 87
        },
        "outputId": "85727c11-441e-45c1-e70e-cc76d60e1cd8"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Recurrent neural networks, long short-term memory [13] and gated recurrent [7] neural networks have been firmly established as state of the art approaches in sequence modeling and transduction problems. In this work we propose the transformer, a model architecture eschewing recurrence and relying entirely on an attention mechanism to draw global dependencies between input and output. The transformer allows for significantly more parallelization and can reach a new state-of-the-art in translation quality after being trained for as little as twelve hours on eight p100 gpus.'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Article 3: [Recurrent Neural Networks (RNNs): A gentle Introduction and Overview](https://arxiv.org/abs/1912.05911)"
      ],
      "metadata": {
        "id": "lVcXPuUYT-aR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "article_pdf_path_3 = \"/content/1912.05911v1.pdf\""
      ],
      "metadata": {
        "id": "gjeMgg4ahbzk"
      },
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Example usage\n",
        "article_text = extract_main_text(pdf_path=article_pdf_path_3, beginning=\"introduction & notation\")\n",
        "article_text_cleaned = clean_text(article_text)\n",
        "article_text_cleaned[:1000]"
      ],
      "metadata": {
        "id": "iedhi7XyUAGP",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        },
        "outputId": "b97fa906-5da3-469a-af59-b507b1182266"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Recurrent Neural Networks (RNNs) are a type of neural network architecture which is mainly used to detect patterns in a sequence of data. Such data can be handwriting, genomes, text or numerical time series which are often produced in industry settings (e.g. stock markets or sensors) [7, 12]. However, they are also applicable to images if these get respectively decomposed into a series of patches and treated as a sequence [12]. On a higher level, RNNs find applications in Language Modelling & Generating Text, Speech Recognition, Generating Image Descriptions or Video Tagging. What differentiates Recurrent Neural Networks from Feedforward Neural Networks also known as Multi-Layer Perceptrons (MLPs) is how information gets passed through the network. While Feedforward Networks pass information through the network without cycles, the RNN has cycles and transmits information back into itself. This enables them to extend the functionality of Feedforward Networks to also take into account pre'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "lora_summary = summarize_text(model=lora_model.to(\"cuda\"), tokenizer=tokenizer, text=article_text_cleaned)\n",
        "lora_summary"
      ],
      "metadata": {
        "id": "zdAHolweUABG",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        },
        "outputId": "4db40a39-0da8-473b-bb07-383b8cfa4739"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Recurrent neural networks (rnns) are a type of neural network architecture which is mainly used to detect patterns in a sequence of data such as handwriting, genomes, text or numerical time series. While feedforward networks pass information through the network without cycles, the rnn has cycles and transmits information back into itself. This enables them to extend the functionality of feedforward neural networks to also take into account previous inputs x0:t−1 and not only the current input xt. In this paper, we show how to construct a recurrent neural network using the backpropagation through time (bptt) and truncated bptt algorithm for rnns.'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "original_summary = summarize_text(model=original_model.to(\"cuda\"), tokenizer=tokenizer, text=article_text_cleaned)\n",
        "original_summary"
      ],
      "metadata": {
        "id": "f0CwXPt8T__N",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 70
        },
        "outputId": "e57dc83f-b3a3-4b22-dccd-a5e2edf1f416"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Recurrent neural networks (rnns) are a type of neural network architecture. They are mainly used to detect patterns in a sequence of data. Rnns have applications in language modelling & generating text, speech recognition, generating image descriptions or video tagging. Backpropagation through time (bptt) is used to backpropagate the error through a rnn.'"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "string"
            }
          },
          "metadata": {},
          "execution_count": 19
        }
      ]
    }
  ]
}