add model

.gitattributes

@@ -25,3 +25,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+training_assets/cross_silver_scores_v3.pkl filter=lfs diff=lfs merge=lfs -text
+training_assets/silver_cross_samples.pkl filter=lfs diff=lfs merge=lfs -text
+training_assets/silver_data.pkl filter=lfs diff=lfs merge=lfs -text
+training_assets/gold_eval_dataloader.pkl filter=lfs diff=lfs merge=lfs -text
+training_assets/gold_train_dataloader.pkl filter=lfs diff=lfs merge=lfs -text

1_Pooling/config.json

@@ -0,0 +1,7 @@
+{
+  "word_embedding_dimension": 768,
+  "pooling_mode_cls_token": false,
+  "pooling_mode_mean_tokens": true,
+  "pooling_mode_max_tokens": false,
+  "pooling_mode_mean_sqrt_len_tokens": false
+}

README.md

@@ -0,0 +1,126 @@
+---
+pipeline_tag: sentence-similarity
+tags:
+- sentence-transformers
+- feature-extraction
+- sentence-similarity
+- transformers
+
+---
+
+# {MODEL_NAME}
+
+This is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.
+
+<!--- Describe your model here -->
+
+## Usage (Sentence-Transformers)
+
+Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:
+
+```
+pip install -U sentence-transformers
+```
+
+Then you can use the model like this:
+
+```python
+from sentence_transformers import SentenceTransformer
+sentences = ["This is an example sentence", "Each sentence is converted"]
+
+model = SentenceTransformer('{MODEL_NAME}')
+embeddings = model.encode(sentences)
+print(embeddings)
+```
+
+
+
+## Usage (HuggingFace Transformers)
+Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings.
+
+```python
+from transformers import AutoTokenizer, AutoModel
+import torch
+
+
+#Mean Pooling - Take attention mask into account for correct averaging
+def mean_pooling(model_output, attention_mask):
+    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
+    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
+    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
+
+
+# Sentences we want sentence embeddings for
+sentences = ['This is an example sentence', 'Each sentence is converted']
+
+# Load model from HuggingFace Hub
+tokenizer = AutoTokenizer.from_pretrained('{MODEL_NAME}')
+model = AutoModel.from_pretrained('{MODEL_NAME}')
+
+# Tokenize sentences
+encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')
+
+# Compute token embeddings
+with torch.no_grad():
+    model_output = model(**encoded_input)
+
+# Perform pooling. In this case, mean pooling.
+sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
+
+print("Sentence embeddings:")
+print(sentence_embeddings)
+```
+
+
+
+## Evaluation Results
+
+<!--- Describe how your model was evaluated -->
+
+For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name={MODEL_NAME})
+
+
+## Training
+The model was trained with the parameters:
+
+**DataLoader**:
+
+`torch.utils.data.dataloader.DataLoader` of length 80805 with parameters:
+```
+{'batch_size': 8, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}
+```
+
+**Loss**:
+
+`sentence_transformers.losses.CosineSimilarityLoss.CosineSimilarityLoss` 
+
+Parameters of the fit()-Method:
+```
+{
+    "epochs": 3,
+    "evaluation_steps": 10000,
+    "evaluator": "__main__.EmbeddingSimilarityEvaluator",
+    "max_grad_norm": 1,
+    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
+    "optimizer_params": {
+        "lr": 2e-05
+    },
+    "scheduler": "WarmupLinear",
+    "steps_per_epoch": null,
+    "warmup_steps": 24242,
+    "weight_decay": 0.01
+}
+```
+
+
+## Full Model Architecture
+```
+SentenceTransformer(
+  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
+  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
+)
+```
+
+## Citing & Authors
+
+<!--- Describe where people can find more information -->

config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "/content/drive/MyDrive/Colab_Notebooks/Anatel-gdrive/anatel_train_sts_cross_bm25/augmented-bert-portuguese-anatel-last-train/",
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "directionality": "bidi",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "output_past": true,
+  "pad_token_id": 0,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.21.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 29794
+}

config_sentence_transformers.json

@@ -0,0 +1,7 @@
+{
+  "__version__": {
+    "sentence_transformers": "2.2.2",
+    "transformers": "4.21.0",
+    "pytorch": "1.12.0+cu113"
+  }
+}

modules.json

@@ -0,0 +1,14 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  }
+]

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4c080e8cd37d8c8cb7b923fd7e780bfb4ad3ea20ab40edc4083f8124c8e29bd
+size 435761969

sentence_bert_config.json

@@ -0,0 +1,4 @@
+{
+  "max_seq_length": 512,
+  "do_lower_case": false
+}

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": false,
+  "mask_token": "[MASK]",
+  "name_or_path": "/content/drive/MyDrive/Colab_Notebooks/Anatel-gdrive/anatel_train_sts_cross_bm25/augmented-bert-portuguese-anatel-last-train/",
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "special_tokens_map_file": "/root/.cache/huggingface/transformers/eecc45187d085a1169eed91017d358cc0e9cbdd5dc236bcd710059dbf0a2f816.dd8bd9bfd3664b530ea4e645105f557769387b3da9f79bdb55ed556bdd80611d",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

training_assets/2_train_sts_cross_bm25.py

@@ -0,0 +1,43 @@
+#%%
+from tqdm import tqdm
+from request_solr import SilverDataset
+from sentence_transformers.cross_encoder import CrossEncoder
+import joblib
+from solr_query_params import params
+
+############################################################################
+#
+# https://github.com/UKPLab/sentence-transformers/tree/master/examples/training/data_augmentation/train_sts_indomain_bm25.py
+# Step 2: Label BM25 sampled STSb (silver dataset) using cross-encoder model
+#
+############################################################################
+
+
+cross_encoder_path = 'ennioferreirab/cross-encoder-pt-anatel-metadados-assunto'
+gold_sample_index = set()
+with open('gold_sample_index.txt', 'r') as f:
+    for line in f:
+        gold_sample_index.add(line.strip())
+        7
+try:
+    joblib.load('silver_data_v2.pkl')
+except:
+    print('Creating silver data...')
+    silver_data = SilverDataset(query_params=params,duplicated=gold_sample_index).run()
+    joblib.dump(silver_data, 'silver_data_v2.pkl')
+    print('Done!')
+
+sentences = [(sent_1,sent_2) for sent_1, sent_2, _ in silver_data]
+
+cross_encoder = CrossEncoder(cross_encoder_path,max_length=512)
+cross_silver_scores = []
+for i in tqdm(sentences):
+    cross_silver_scores.append(cross_encoder.predict(i))
+
+import numpy as np
+cross_silver_data = np.c_[np.array(silver_data),np.array(cross_silver_scores)]
+
+# All model predictions should be between [0,1]
+assert all(0.0 <= score <= 1.0 for score in cross_silver_scores)
+
+joblib.dump(cross_silver_data, 'cross_silver_scores_2.pkl')

training_assets/cross_silver_scores_v3.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd9d6a0296f0a1e9589ac8550d6095d9f53985ecd3fc3a8f1e4398426acb84d0
+size 239383791

training_assets/gold_eval_dataloader.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8901155d353af2a1fad078daafb7eabab5bc6779d69ddb3768a359ac2b50bdad
+size 127396

training_assets/gold_sample_index.txt

@@ -0,0 +1,825 @@
+4460666
+4646076
+4740485
+4841542
+5137102
+5433651
+5619579
+6730044
+6960718
+6983639
+7331138
+7460176
+8137405
+5133708
+5591405
+6098115
+6200408
+6816332
+7073711
+8252830
+5333260
+5545291
+5903862
+6070901
+6269728
+6391200
+6710641
+6763601
+5808966
+6164703
+5809430
+6379512
+8182456
+5177401
+5510278
+6083218
+6270034
+6543759
+5369628
+6099099
+6174942
+7902136
+5340149
+5401404
+5981567
+6465290
+6643010
+7835158
+7889240
+8084176
+5433923
+5498005
+5516542
+5809052
+5855590
+5985223
+6110110
+6151666
+6153056
+6439888
+6525429
+6556589
+6635224
+6986996
+7080918
+7114149
+7128483
+7886766
+8131931
+4682622
+5164923
+5503680
+5920921
+6209634
+6422257
+6872536
+7427806
+7986978
+7994954
+5774177
+5989286
+7508398
+5591454
+6102860
+5637632
+5859904
+4370203
+4949468
+5426346
+5859209
+6266408
+5387595
+5465052
+5521946
+5991666
+6209282
+6491862
+6548583
+7494666
+7595318
+8133086
+5634836
+5850384
+5863288
+6398949
+6635184
+6904804
+6975680
+7969196
+8163454
+6887423
+4479704
+5804564
+4646057
+4649560
+4895445
+5182219
+5205994
+5504036
+5864758
+6960731
+7337160
+8158609
+8165279
+8192206
+5766461
+6112431
+8179460
+3674750
+5230043
+6781048
+6862078
+8133237
+5608073
+5640398
+5987322
+6103319
+7348722
+7599312
+7889440
+7954697
+8151933
+6081741
+5261625
+5856321
+6464888
+6520857
+6779057
+6913683
+7090008
+7348871
+7410868
+8091065
+8131423
+8231168
+5550518
+8163535
+7783020
+8150022
+5992730
+6866793
+8139859
+7016421
+8213148
+4740150
+5805453
+4646084
+5174687
+5550543
+6010237
+6875009
+7030819
+7421517
+7492466
+7493941
+7522041
+7602469
+7650684
+7659691
+7991553
+8090644
+4486736
+5482615
+5551413
+5208352
+5546726
+5683883
+5689258
+5871416
+5920031
+4574013
+4888346
+5183619
+5504640
+5984725
+6026816
+6415398
+8141389
+7114643
+4620965
+5164684
+5406485
+5586960
+5855047
+6209691
+6392602
+8183810
+5276276
+5795308
+5327818
+4596269
+5272298
+5430104
+5613420
+5779154
+6294837
+6860757
+8154715
+8163261
+8258448
+4007242
+5326368
+6035394
+6108765
+6317765
+6791523
+6974773
+8152457
+8190516
+6712858
+6465138
+6664825
+8163904
+7892776
+5756069
+8205392
+7203259
+5599430
+5350003
+5855756
+6270070
+7145585
+5091384
+5876381
+5957940
+6108724
+6421665
+6620134
+7209796
+8151360
+4087034
+4531381
+4596045
+4855347
+5162839
+5235040
+5253056
+5327778
+5329186
+5401428
+5617203
+5625555
+5675553
+5773205
+5843647
+5872380
+5989087
+6215135
+6439752
+6482537
+6537688
+6604972
+6729495
+8133210
+5272072
+5516186
+7629938
+7423420
+5159992
+6980613
+6627386
+3913260
+5515029
+7630311
+5685194
+6033758
+6409793
+4595837
+3862776
+5540750
+6032434
+5930452
+5971699
+6548457
+5511835
+5563594
+6405254
+6958561
+5408482
+5502073
+5659569
+6072736
+6418523
+7126761
+4418008
+4460858
+5345610
+1825442
+2051705
+7650052
+5314776
+5571539
+4165786
+4551806
+3973194
+4304284
+4663854
+5211035
+5935558
+5852416
+3866461
+1984797
+4348571
+2667288
+2875287
+2963171
+3961144
+4096818
+4209438
+4409058
+4602656
+5238188
+5240637
+5481769
+5486463
+5497117
+5502194
+5502389
+5502390
+5627659
+5641429
+5673797
+5674040
+5677426
+5722966
+5858953
+6030890
+6057740
+6081452
+6092841
+6397111
+6758807
+7760525
+7889990
+8187830
+6733763
+5256783
+5454894
+6254439
+5430395
+5564934
+5790594
+7407966
+8120458
+3809160
+3947410
+4280956
+4950329
+5159954
+7245076
+7994916
+5476366
+5392611
+7296149
+4675474
+7771746
+4162522
+4540151
+5516359
+5586556
+5512844
+5557217
+4712460
+6593592
+8050832
+8235082
+8283799
+7236674
+5887216
+5990578
+7523737
+7880724
+8199567
+5637600
+4201008
+4406618
+8249957
+5591666
+6022264
+1827408
+2605566
+3091273
+3691667
+5807405
+5034084
+6428700
+5568958
+6021279
+6275467
+7075117
+6088741
+6446079
+3678407
+3892052
+3802952
+4660263
+5523325
+5476893
+5956550
+6074852
+6391212
+0888264
+1226636
+1325045
+1796335
+3863377
+4190087
+4872263
+5639415
+4598502
+4938680
+5394225
+5941839
+5986252
+6831056
+7389491
+5518440
+5328013
+5795782
+6405189
+6910804
+7428552
+3912783
+4334946
+4750544
+5428246
+5534921
+8232931
+5627612
+5882930
+5924285
+5564068
+6115268
+6603458
+5503797
+5591747
+6009845
+8161801
+4287397
+6508273
+6707346
+8196196
+5608601
+6020852
+6105662
+6529304
+4024170
+8132569
+5236024
+5403472
+5397274
+6410230
+6939796
+8182990
+6538571
+3717823
+4141332
+4475895
+6015699
+6090450
+6594521
+6969505
+8131792
+7759380
+5532712
+5886827
+3174130
+3714303
+5016494
+5745123
+6091768
+8164430
+5878425
+4791647
+7517583
+5662890
+3974603
+4086799
+4282564
+5344043
+5414706
+5416591
+5808227
+0975861
+1538893
+1591036
+1838447
+5138501
+6715682
+5863991
+5135182
+5221262
+4922845
+6731643
+7907513
+8053350
+5228986
+6415154
+5141202
+4646096
+5167166
+4772438
+5309058
+5342240
+6400759
+6881639
+6075097
+6614341
+5468988
+5551297
+3503104
+3963399
+4274953
+6825998
+7434061
+7668958
+4870966
+5110222
+5418746
+5418901
+5489424
+6005690
+6106680
+6450841
+6022525
+5171532
+7504807
+7752278
+6420131
+4779688
+4898526
+5134283
+5262888
+4648476
+7521721
+5208120
+6792659
+8085896
+4848964
+7435248
+6161277
+7887662
+6742588
+5626956
+5618086
+5734885
+5090949
+5116239
+5124741
+5159732
+5219306
+5591519
+5676606
+5690219
+5786819
+6041528
+6072192
+7884624
+7981332
+8084930
+5612301
+4731346
+5623910
+6091875
+6281436
+5213783
+6009557
+7039607
+7679918
+5176028
+5560512
+5874435
+6520636
+5468132
+5556839
+6075388
+6167589
+6402939
+6439732
+6604418
+5205819
+5335110
+5384794
+6173780
+7613196
+5161886
+5426013
+5520987
+5559035
+5931478
+4540213
+5167812
+5364524
+5400717
+5547227
+5605539
+5622937
+5639412
+5795790
+6906068
+6947670
+7963695
+6178278
+5101120
+5862886
+5470473
+6002283
+5186421
+5196824
+5226107
+5231329
+5235885
+5245418
+5247129
+5328934
+5338196
+5344932
+5347088
+5355720
+5362812
+5365107
+5376109
+5376117
+5380762
+5393873
+5394467
+5395955
+5415840
+5425280
+5449732
+5503636
+5541083
+5743983
+5747840
+5792689
+5794851
+5839502
+5887460
+5926330
+6009768
+7039563
+8044928
+8216267
+3730523
+4786125
+5211966
+5469180
+6897698
+5138518
+5401485
+5753393
+6180767
+4106645
+6164679
+6391213
+6401090
+7373471
+6916874
+5103812
+5744044
+6021351
+8010147
+4813869
+7949719
+5349964
+5846693
+8064168
+6812336
+5674311
+5153043
+5159853
+5399855
+5424931
+5444239
+5727887
+5945666
+6092852
+2010236
+2097934
+3756637
+3874160
+3976877
+4301951
+5091159
+5420402
+5466778
+5476112
+6158436
+5545661
+6726156
+5920353
+6538842
+6519579
+6880547
+5206272
+6106472
+6516584
+6732221
+3872348
+5116412
+5563281
+3856904
+4428505
+5345839
+6094897
+6419864
+5945139
+4706163
+3725580
+4003539
+4027979
+4736809
+5025370
+5084254
+5211319
+5257759
+5429093
+5441857
+5505327
+5560235
+5602619
+5606581
+5793811
+5807776
+6010825
+6059919
+6081869
+6137061
+6549955
+6614527
+7246382
+4741669
+0908100
+4486339
+5162489
+5493532
+5636837
+6049972
+4047135
+4385386
+8248665
+1402356
+4967941
+5092529
+5214722
+5228115
+5291887
+5519267
+5537326
+5795976
+5796440
+5803348
+5810185
+5852507
+5909067
+5977933
+6796500
+6863557
+7428664
+8152906
+4404049
+2607396
+3183641
+3184217
+3449359
+3727592
+3734172
+3966722
+4137035
+4224913
+4902962
+4969387
+5163826
+5206856
+5239793
+5269996
+5311495
+5338031
+5370069
+5399410
+5425394
+5440378
+5446751
+5456903
+5458292
+5476100
+5482266
+5504203
+5504582
+5524801
+5551055
+5553070
+5554819
+5567170
+5568809
+5669114
+5717111
+5992402
+6000491
+6059906
+6146140
+6351943
+6467278
+6780508
+8197798

training_assets/gold_train_dataloader.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c53ce2f26aab328f08d2db38d11718bb3579048ced91a9acb6a607b95228eaa2
+size 3586422

training_assets/request_solr.py

@@ -0,0 +1,64 @@
+import httpx
+from pydantic import BaseModel
+import tqdm
+'''
+gold_sample_index = set()
+with open('gold_sample_index.txt', 'r') as f:
+    for line in f:
+        gold_sample_index.add(line.strip())
+'''
+
+class SilverDataset(BaseModel):
+    '''
+    Classe para retornar o dataset silver para ser utilizado pela estratégia do Augmented SBERT
+    ref: https://github.com/UKPLab/sentence-transformers/tree/master/examples/training/data_augmentation/train_sts_indomain_bm25.py
+    '''
+    query_params: dict
+    duplicated: set = set()
+    solr_url: str = "http://localhost:8983/solr/sei_similaridade_augmented_sbert"
+    
+    def get_ids_list(self):
+        query = f'{self.solr_url}/select?q=*&fl=id&rows=999999999'
+        r = httpx.get(query).json()['response']['docs']
+        return [doc['id'] for doc in r]
+
+    def get_data(self,id):
+        q_id = f"id:{id}"
+        self.query_params['q'] = q_id
+        r = httpx.post(f'{self.solr_url}/mlt', data=self.query_params).json()
+        maxscore = r['response']['maxScore']
+        response_docs =  r['response']['docs']
+        response_docs = self.remove_duplicated(response_docs)
+        return {'query_id': id,
+                'query_doc': r['match']['docs'][0][self.query_params['mlt.qf']],
+                'docs': response_docs,
+                'maxscore': maxscore}
+    
+    def remove_duplicated(self,docs):
+        '''
+        remove os documentos que são iguais aos documentos do dataset Gold
+        '''
+        return [doc for doc in docs if doc['id'] not in self.duplicated]
+
+
+    @staticmethod
+    def create_sentence_pairs(queries):
+        '''
+        cria os pares de frases para o dataset silver
+        '''
+        pairs = set()
+        for query in queries:
+            for doc in query['docs']:
+                pairs.add(
+                    (query['query_doc'],
+                    doc['assunto_text'],
+                    doc['score']/query['maxscore']))
+        return pairs
+    
+    def run(self):
+        queries = []
+        list_ids = self.get_ids_list()
+        for id in tqdm.tqdm(list_ids):
+            queries.append(self.get_data(id))
+        pairs = self.create_sentence_pairs(queries)
+        return pairs

training_assets/silver_cross_samples.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6a9119d292328000dd27b5f674e0cf86c708d1b9042a9b8911c03a6726c2e50
+size 239072747

training_assets/silver_data.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68e0f9acd5aea86b2e45ae6eee6840c1df4f70705b21b1b9a535ecae0580e5fc
+size 303024365