mssongit commited on
Commit
b927680
1 Parent(s): 9ff0374

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
.ipynb_checkpoints/README-checkpoint.md ADDED
@@ -0,0 +1,414 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - sentence-transformers
4
+ - sentence-similarity
5
+ - feature-extraction
6
+ - generated_from_trainer
7
+ - dataset_size:574325
8
+ - loss:MultipleNegativesRankingLoss
9
+ - loss:CosineSimilarityLoss
10
+ base_model: upskyy/gte-base-korean
11
+ widget:
12
+ - source_sentence: 그것은 도덕적으로 강요하지 않는다.
13
+ sentences:
14
+ - 그것은 법을 제정하고 있다.
15
+ - 시리아 야당은 회담에 참석할 것을 촉구했다.
16
+ - 한 젊은이가 기타를 연주하면서 노래를 부르고 있다.
17
+ - source_sentence: 한 여성이 무대에서 플루트를 연주하고 있다.
18
+ sentences:
19
+ - 여자가 플루트를 연주하고 있다.
20
+ - 인도, 사이클론 파일린에 대한 적색 경보 발령
21
+ - 반 데르 메르웨는 가이게스의 형을 5년 징역형으로 중지했다.
22
+ - source_sentence: 적어도 나는 이 남자가 자신의 범죄를 이해한다고 확신할 수 있었다.
23
+ sentences:
24
+ - 티셔츠와 반바지를 입고 티에서 축구를 걷어차는 남자
25
+ - 나는 그가 무엇을 잘못했는지 전혀 모른다고 생각하기 시작했다.
26
+ - 남자는 자신이 한 일을 알고 있었다.
27
+ - source_sentence: 사람은 다리로 올라갑니다.
28
+ sentences:
29
+ - 한 남자가 땅에 누워 있다.
30
+ - 자전거를 타는 한 무리의 사람들이 거리에서 돌아선다.
31
+ - 공중으로 뛰어드는 남자
32
+ - source_sentence: 모자를 쓴 남자와 여자가 거리에서 악기를 연주하고 있다.
33
+ sentences:
34
+ - 사람은 수직 물체에 받쳐진다.
35
+ - 두 남자가 길가에 서 있다.
36
+ - 두 사람이 모자를 쓰고 있다.
37
+ pipeline_tag: sentence-similarity
38
+ library_name: sentence-transformers
39
+ metrics:
40
+ - pearson_cosine
41
+ - spearman_cosine
42
+ model-index:
43
+ - name: SentenceTransformer based on upskyy/gte-base-korean
44
+ results:
45
+ - task:
46
+ type: semantic-similarity
47
+ name: Semantic Similarity
48
+ dataset:
49
+ name: sts dev
50
+ type: sts-dev
51
+ metrics:
52
+ - type: pearson_cosine
53
+ value: 0.868140244252358
54
+ name: Pearson Cosine
55
+ - type: spearman_cosine
56
+ value: 0.8689161244129222
57
+ name: Spearman Cosine
58
+ ---
59
+
60
+ # SentenceTransformer based on upskyy/gte-base-korean
61
+
62
+ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [upskyy/gte-base-korean](https://huggingface.co/upskyy/gte-base-korean). It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
63
+
64
+ ## Model Details
65
+
66
+ ### Model Description
67
+ - **Model Type:** Sentence Transformer
68
+ - **Base model:** [upskyy/gte-base-korean](https://huggingface.co/upskyy/gte-base-korean) <!-- at revision c1a18ef8326962b57c63e2d306a724a925913dfe -->
69
+ - **Maximum Sequence Length:** 8192 tokens
70
+ - **Output Dimensionality:** 768 dimensions
71
+ - **Similarity Function:** Cosine Similarity
72
+ <!-- - **Training Dataset:** Unknown -->
73
+ <!-- - **Language:** Unknown -->
74
+ <!-- - **License:** Unknown -->
75
+
76
+ ### Model Sources
77
+
78
+ - **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
79
+ - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
80
+ - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
81
+
82
+ ### Full Model Architecture
83
+
84
+ ```
85
+ SentenceTransformer(
86
+ (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: NewModel
87
+ (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, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
88
+ )
89
+ ```
90
+
91
+ ## Usage
92
+
93
+ ### Direct Usage (Sentence Transformers)
94
+
95
+ First install the Sentence Transformers library:
96
+
97
+ ```bash
98
+ pip install -U sentence-transformers
99
+ ```
100
+
101
+ Then you can load this model and run inference.
102
+ ```python
103
+ from sentence_transformers import SentenceTransformer
104
+
105
+ # Download from the 🤗 Hub
106
+ model = SentenceTransformer("sentence_transformers_model_id")
107
+ # Run inference
108
+ sentences = [
109
+ '모자를 쓴 남자와 여자가 거리에서 악기를 연주하고 있다.',
110
+ '두 사람이 모자를 쓰고 있다.',
111
+ '두 남자가 길가에 서 있다.',
112
+ ]
113
+ embeddings = model.encode(sentences)
114
+ print(embeddings.shape)
115
+ # [3, 768]
116
+
117
+ # Get the similarity scores for the embeddings
118
+ similarities = model.similarity(embeddings, embeddings)
119
+ print(similarities.shape)
120
+ # [3, 3]
121
+ ```
122
+
123
+ <!--
124
+ ### Direct Usage (Transformers)
125
+
126
+ <details><summary>Click to see the direct usage in Transformers</summary>
127
+
128
+ </details>
129
+ -->
130
+
131
+ <!--
132
+ ### Downstream Usage (Sentence Transformers)
133
+
134
+ You can finetune this model on your own dataset.
135
+
136
+ <details><summary>Click to expand</summary>
137
+
138
+ </details>
139
+ -->
140
+
141
+ <!--
142
+ ### Out-of-Scope Use
143
+
144
+ *List how the model may foreseeably be misused and address what users ought not to do with the model.*
145
+ -->
146
+
147
+ ## Evaluation
148
+
149
+ ### Metrics
150
+
151
+ #### Semantic Similarity
152
+
153
+ * Dataset: `sts-dev`
154
+ * Evaluated with [<code>EmbeddingSimilarityEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.EmbeddingSimilarityEvaluator)
155
+
156
+ | Metric | Value |
157
+ |:--------------------|:-----------|
158
+ | pearson_cosine | 0.8681 |
159
+ | **spearman_cosine** | **0.8689** |
160
+
161
+ <!--
162
+ ## Bias, Risks and Limitations
163
+
164
+ *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
165
+ -->
166
+
167
+ <!--
168
+ ### Recommendations
169
+
170
+ *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
171
+ -->
172
+
173
+ ## Training Details
174
+
175
+ ### Training Datasets
176
+
177
+ #### Unnamed Dataset
178
+
179
+
180
+ * Size: 568,576 training samples
181
+ * Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>sentence_2</code>
182
+ * Approximate statistics based on the first 1000 samples:
183
+ | | sentence_0 | sentence_1 | sentence_2 |
184
+ |:--------|:-----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|
185
+ | type | string | string | string |
186
+ | details | <ul><li>min: 5 tokens</li><li>mean: 19.18 tokens</li><li>max: 111 tokens</li></ul> | <ul><li>min: 4 tokens</li><li>mean: 20.1 tokens</li><li>max: 129 tokens</li></ul> | <ul><li>min: 4 tokens</li><li>mean: 14.51 tokens</li><li>max: 42 tokens</li></ul> |
187
+ * Samples:
188
+ | sentence_0 | sentence_1 | sentence_2 |
189
+ |:----------------------------------------------------------|:------------------------------------------------------------|:------------------------------------------------|
190
+ | <code>나는 솔직히 말해서 가게에서 그것을 산 사람은 그 남자가 아니라고 말할 것이다.</code> | <code>화학자 가게에서 스트리크닌을 산 사람은 그가 아니었다는 것을 인정하겠다.</code> | <code>난 아무것도 인정하지 않을 거야, 이 모든 대화는 무의미해!</code> |
191
+ | <code>네 명의 여성이 있다.</code> | <code>검은색과 노란색 드레스를 입은 세 명의 여성과 오렌지색 머리를 가진 한 명의 여성.</code> | <code>신부 들러리 세 명이 모두 어울리지 않는 드레스를 입고 있다.</code> |
192
+ | <code>드류는 빤히 쳐다보면서 다른 사람을 생각하고 있었다.</code> | <code>하지만 다른 하나는...... 드류가 빤히 쳐다보았다.</code> | <code>드류는 다른 사람을 걱정하지 않았다.</code> |
193
+ * Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
194
+ ```json
195
+ {
196
+ "scale": 20.0,
197
+ "similarity_fct": "cos_sim"
198
+ }
199
+ ```
200
+
201
+ #### Unnamed Dataset
202
+
203
+
204
+ * Size: 5,749 training samples
205
+ * Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>label</code>
206
+ * Approximate statistics based on the first 1000 samples:
207
+ | | sentence_0 | sentence_1 | label |
208
+ |:--------|:---------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:---------------------------------------------------------------|
209
+ | type | string | string | float |
210
+ | details | <ul><li>min: 4 tokens</li><li>mean: 19.1 tokens</li><li>max: 74 tokens</li></ul> | <ul><li>min: 5 tokens</li><li>mean: 19.15 tokens</li><li>max: 83 tokens</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.53</li><li>max: 1.0</li></ul> |
211
+ * Samples:
212
+ | sentence_0 | sentence_1 | label |
213
+ |:---------------------------------------|:--------------------------------------|:-----------------|
214
+ | <code>�� 남자가 바이올린을 연주하고 있다.</code> | <code>아기가 웃고 기어가고 있다.</code> | <code>0.0</code> |
215
+ | <code>구스마오는 동티모르 선거에서 권력을 강화한다.</code> | <code>롬니가 선거에서 승리할 경우 대법원의 가능성</code> | <code>0.0</code> |
216
+ | <code>그게 아니었다는 것만 빼면.</code> | <code>그들이 할 수 없다는 것 빼고는...</code> | <code>0.2</code> |
217
+ * Loss: [<code>CosineSimilarityLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosinesimilarityloss) with these parameters:
218
+ ```json
219
+ {
220
+ "loss_fct": "torch.nn.modules.loss.MSELoss"
221
+ }
222
+ ```
223
+
224
+ ### Training Hyperparameters
225
+ #### Non-Default Hyperparameters
226
+
227
+ - `eval_strategy`: steps
228
+ - `num_train_epochs`: 1
229
+ - `batch_sampler`: no_duplicates
230
+ - `multi_dataset_batch_sampler`: round_robin
231
+
232
+ #### All Hyperparameters
233
+ <details><summary>Click to expand</summary>
234
+
235
+ - `overwrite_output_dir`: False
236
+ - `do_predict`: False
237
+ - `eval_strategy`: steps
238
+ - `prediction_loss_only`: True
239
+ - `per_device_train_batch_size`: 8
240
+ - `per_device_eval_batch_size`: 8
241
+ - `per_gpu_train_batch_size`: None
242
+ - `per_gpu_eval_batch_size`: None
243
+ - `gradient_accumulation_steps`: 1
244
+ - `eval_accumulation_steps`: None
245
+ - `torch_empty_cache_steps`: None
246
+ - `learning_rate`: 5e-05
247
+ - `weight_decay`: 0.0
248
+ - `adam_beta1`: 0.9
249
+ - `adam_beta2`: 0.999
250
+ - `adam_epsilon`: 1e-08
251
+ - `max_grad_norm`: 1.0
252
+ - `num_train_epochs`: 1
253
+ - `max_steps`: -1
254
+ - `lr_scheduler_type`: linear
255
+ - `lr_scheduler_kwargs`: {}
256
+ - `warmup_ratio`: 0.0
257
+ - `warmup_steps`: 0
258
+ - `log_level`: passive
259
+ - `log_level_replica`: warning
260
+ - `log_on_each_node`: True
261
+ - `logging_nan_inf_filter`: True
262
+ - `save_safetensors`: True
263
+ - `save_on_each_node`: False
264
+ - `save_only_model`: False
265
+ - `restore_callback_states_from_checkpoint`: False
266
+ - `no_cuda`: False
267
+ - `use_cpu`: False
268
+ - `use_mps_device`: False
269
+ - `seed`: 42
270
+ - `data_seed`: None
271
+ - `jit_mode_eval`: False
272
+ - `use_ipex`: False
273
+ - `bf16`: False
274
+ - `fp16`: False
275
+ - `fp16_opt_level`: O1
276
+ - `half_precision_backend`: auto
277
+ - `bf16_full_eval`: False
278
+ - `fp16_full_eval`: False
279
+ - `tf32`: None
280
+ - `local_rank`: 0
281
+ - `ddp_backend`: None
282
+ - `tpu_num_cores`: None
283
+ - `tpu_metrics_debug`: False
284
+ - `debug`: []
285
+ - `dataloader_drop_last`: False
286
+ - `dataloader_num_workers`: 0
287
+ - `dataloader_prefetch_factor`: None
288
+ - `past_index`: -1
289
+ - `disable_tqdm`: False
290
+ - `remove_unused_columns`: True
291
+ - `label_names`: None
292
+ - `load_best_model_at_end`: False
293
+ - `ignore_data_skip`: False
294
+ - `fsdp`: []
295
+ - `fsdp_min_num_params`: 0
296
+ - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
297
+ - `fsdp_transformer_layer_cls_to_wrap`: None
298
+ - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
299
+ - `deepspeed`: None
300
+ - `label_smoothing_factor`: 0.0
301
+ - `optim`: adamw_torch
302
+ - `optim_args`: None
303
+ - `adafactor`: False
304
+ - `group_by_length`: False
305
+ - `length_column_name`: length
306
+ - `ddp_find_unused_parameters`: None
307
+ - `ddp_bucket_cap_mb`: None
308
+ - `ddp_broadcast_buffers`: False
309
+ - `dataloader_pin_memory`: True
310
+ - `dataloader_persistent_workers`: False
311
+ - `skip_memory_metrics`: True
312
+ - `use_legacy_prediction_loop`: False
313
+ - `push_to_hub`: False
314
+ - `resume_from_checkpoint`: None
315
+ - `hub_model_id`: None
316
+ - `hub_strategy`: every_save
317
+ - `hub_private_repo`: False
318
+ - `hub_always_push`: False
319
+ - `gradient_checkpointing`: False
320
+ - `gradient_checkpointing_kwargs`: None
321
+ - `include_inputs_for_metrics`: False
322
+ - `include_for_metrics`: []
323
+ - `eval_do_concat_batches`: True
324
+ - `fp16_backend`: auto
325
+ - `push_to_hub_model_id`: None
326
+ - `push_to_hub_organization`: None
327
+ - `mp_parameters`:
328
+ - `auto_find_batch_size`: False
329
+ - `full_determinism`: False
330
+ - `torchdynamo`: None
331
+ - `ray_scope`: last
332
+ - `ddp_timeout`: 1800
333
+ - `torch_compile`: False
334
+ - `torch_compile_backend`: None
335
+ - `torch_compile_mode`: None
336
+ - `dispatch_batches`: None
337
+ - `split_batches`: None
338
+ - `include_tokens_per_second`: False
339
+ - `include_num_input_tokens_seen`: False
340
+ - `neftune_noise_alpha`: None
341
+ - `optim_target_modules`: None
342
+ - `batch_eval_metrics`: False
343
+ - `eval_on_start`: False
344
+ - `use_liger_kernel`: False
345
+ - `eval_use_gather_object`: False
346
+ - `average_tokens_across_devices`: False
347
+ - `prompts`: None
348
+ - `batch_sampler`: no_duplicates
349
+ - `multi_dataset_batch_sampler`: round_robin
350
+
351
+ </details>
352
+
353
+ ### Training Logs
354
+ | Epoch | Step | Training Loss | sts-dev_spearman_cosine |
355
+ |:------:|:----:|:-------------:|:-----------------------:|
356
+ | 0.3477 | 500 | 0.1296 | - |
357
+ | 0.6954 | 1000 | 0.1192 | 0.8689 |
358
+
359
+
360
+ ### Framework Versions
361
+ - Python: 3.11.10
362
+ - Sentence Transformers: 3.3.0
363
+ - Transformers: 4.46.2
364
+ - PyTorch: 2.4.0+cu121
365
+ - Accelerate: 1.1.1
366
+ - Datasets: 3.1.0
367
+ - Tokenizers: 0.20.3
368
+
369
+ ## Citation
370
+
371
+ ### BibTeX
372
+
373
+ #### Sentence Transformers
374
+ ```bibtex
375
+ @inproceedings{reimers-2019-sentence-bert,
376
+ title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
377
+ author = "Reimers, Nils and Gurevych, Iryna",
378
+ booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
379
+ month = "11",
380
+ year = "2019",
381
+ publisher = "Association for Computational Linguistics",
382
+ url = "https://arxiv.org/abs/1908.10084",
383
+ }
384
+ ```
385
+
386
+ #### MultipleNegativesRankingLoss
387
+ ```bibtex
388
+ @misc{henderson2017efficient,
389
+ title={Efficient Natural Language Response Suggestion for Smart Reply},
390
+ author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
391
+ year={2017},
392
+ eprint={1705.00652},
393
+ archivePrefix={arXiv},
394
+ primaryClass={cs.CL}
395
+ }
396
+ ```
397
+
398
+ <!--
399
+ ## Glossary
400
+
401
+ *Clearly define terms in order to be accessible across audiences.*
402
+ -->
403
+
404
+ <!--
405
+ ## Model Card Authors
406
+
407
+ *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
408
+ -->
409
+
410
+ <!--
411
+ ## Model Card Contact
412
+
413
+ *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
414
+ -->
.ipynb_checkpoints/config-checkpoint.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "Alibaba-NLP/gte-multilingual-base",
3
+ "architectures": [
4
+ "NewModel"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "Alibaba-NLP/new-impl--configuration.NewConfig",
9
+ "AutoModel": "Alibaba-NLP/new-impl--modeling.NewModel",
10
+ "AutoModelForMaskedLM": "Alibaba-NLP/new-impl--modeling.NewForMaskedLM",
11
+ "AutoModelForMultipleChoice": "Alibaba-NLP/new-impl--modeling.NewForMultipleChoice",
12
+ "AutoModelForQuestionAnswering": "Alibaba-NLP/new-impl--modeling.NewForQuestionAnswering",
13
+ "AutoModelForSequenceClassification": "Alibaba-NLP/new-impl--modeling.NewForSequenceClassification",
14
+ "AutoModelForTokenClassification": "Alibaba-NLP/new-impl--modeling.NewForTokenClassification"
15
+ },
16
+ "classifier_dropout": 0.0,
17
+ "hidden_act": "gelu",
18
+ "hidden_dropout_prob": 0.1,
19
+ "hidden_size": 768,
20
+ "id2label": {
21
+ "0": "LABEL_0"
22
+ },
23
+ "initializer_range": 0.02,
24
+ "intermediate_size": 3072,
25
+ "label2id": {
26
+ "LABEL_0": 0
27
+ },
28
+ "layer_norm_eps": 1e-12,
29
+ "layer_norm_type": "layer_norm",
30
+ "logn_attention_clip1": false,
31
+ "logn_attention_scale": false,
32
+ "max_position_embeddings": 8192,
33
+ "model_type": "new",
34
+ "num_attention_heads": 12,
35
+ "num_hidden_layers": 12,
36
+ "pack_qkv": true,
37
+ "pad_token_id": 1,
38
+ "position_embedding_type": "rope",
39
+ "rope_scaling": {
40
+ "factor": 8.0,
41
+ "type": "ntk"
42
+ },
43
+ "rope_theta": 20000,
44
+ "torch_dtype": "float32",
45
+ "transformers_version": "4.46.2",
46
+ "type_vocab_size": 1,
47
+ "unpad_inputs": false,
48
+ "use_memory_efficient_attention": false,
49
+ "vocab_size": 250048
50
+ }
.ipynb_checkpoints/configuration-checkpoint.py ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2024 The GTE Team Authors and Alibaba Group.
3
+ # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
4
+ #
5
+ # Licensed under the Apache License, Version 2.0 (the "License");
6
+ # you may not use this file except in compliance with the License.
7
+ # You may obtain a copy of the License at
8
+ #
9
+ # http://www.apache.org/licenses/LICENSE-2.0
10
+ #
11
+ # Unless required by applicable law or agreed to in writing, software
12
+ # distributed under the License is distributed on an "AS IS" BASIS,
13
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ # See the License for the specific language governing permissions and
15
+ # limitations under the License.
16
+ """ NEW model configuration"""
17
+ from transformers.configuration_utils import PretrainedConfig
18
+ from transformers.utils import logging
19
+
20
+ logger = logging.get_logger(__name__)
21
+
22
+
23
+ class NewConfig(PretrainedConfig):
24
+ r"""
25
+ This is the configuration class to store the configuration of a [`NewModel`] or a [`TFNewModel`]. It is used to
26
+ instantiate a NEW model according to the specified arguments, defining the model architecture. Instantiating a
27
+ configuration with the defaults will yield a similar configuration to that of the NEW
28
+ [izhx/new-base-en](https://huggingface.co/izhx/new-base-en) architecture.
29
+
30
+ Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
31
+ documentation from [`PretrainedConfig`] for more information.
32
+
33
+
34
+ Args:
35
+ vocab_size (`int`, *optional*, defaults to 30522):
36
+ Vocabulary size of the NEW model. Defines the number of different tokens that can be represented by the
37
+ `inputs_ids` passed when calling [`NewModel`] or [`TFNewModel`].
38
+ hidden_size (`int`, *optional*, defaults to 768):
39
+ Dimensionality of the encoder layers and the pooler layer.
40
+ num_hidden_layers (`int`, *optional*, defaults to 12):
41
+ Number of hidden layers in the Transformer encoder.
42
+ num_attention_heads (`int`, *optional*, defaults to 12):
43
+ Number of attention heads for each attention layer in the Transformer encoder.
44
+ intermediate_size (`int`, *optional*, defaults to 3072):
45
+ Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
46
+ hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
47
+ The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
48
+ `"relu"`, `"silu"` and `"gelu_new"` are supported.
49
+ hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
50
+ The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
51
+ attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
52
+ The dropout ratio for the attention probabilities.
53
+ max_position_embeddings (`int`, *optional*, defaults to 512):
54
+ The maximum sequence length that this model might ever be used with. Typically set this to something large
55
+ just in case (e.g., 512 or 1024 or 2048).
56
+ type_vocab_size (`int`, *optional*, defaults to 2):
57
+ The vocabulary size of the `token_type_ids` passed when calling [`NewModel`] or [`TFNewModel`].
58
+ initializer_range (`float`, *optional*, defaults to 0.02):
59
+ The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
60
+ layer_norm_eps (`float`, *optional*, defaults to 1e-12):
61
+ The epsilon used by the layer normalization layers.
62
+ position_embedding_type (`str`, *optional*, defaults to `"rope"`):
63
+ Type of position embedding. Choose one of `"absolute"`, `"rope"`.
64
+ rope_theta (`float`, *optional*, defaults to 10000.0):
65
+ The base period of the RoPE embeddings.
66
+ rope_scaling (`Dict`, *optional*):
67
+ Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
68
+ strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
69
+ `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
70
+ `max_position_embeddings` to the expected new maximum. See the following thread for more information on how
71
+ these scaling strategies behave:
72
+ https://www.reddit.com/r/LocalLLaMA/comments/14mrgpr/dynamically_scaled_rope_further_increases/. This is an
73
+ experimental feature, subject to breaking API changes in future versions.
74
+ classifier_dropout (`float`, *optional*):
75
+ The dropout ratio for the classification head.
76
+
77
+ Examples:
78
+
79
+ ```python
80
+ >>> from transformers import NewConfig, NewModel
81
+
82
+ >>> # Initializing a NEW izhx/new-base-en style configuration
83
+ >>> configuration = NewConfig()
84
+
85
+ >>> # Initializing a model (with random weights) from the izhx/new-base-en style configuration
86
+ >>> model = NewModel(configuration)
87
+
88
+ >>> # Accessing the model configuration
89
+ >>> configuration = model.config
90
+ ```"""
91
+
92
+ model_type = "new"
93
+
94
+ def __init__(
95
+ self,
96
+ vocab_size=30528,
97
+ hidden_size=768,
98
+ num_hidden_layers=12,
99
+ num_attention_heads=12,
100
+ intermediate_size=3072,
101
+ hidden_act="gelu",
102
+ hidden_dropout_prob=0.1,
103
+ attention_probs_dropout_prob=0.0,
104
+ max_position_embeddings=2048,
105
+ type_vocab_size=1,
106
+ initializer_range=0.02,
107
+ layer_norm_type='layer_norm',
108
+ layer_norm_eps=1e-12,
109
+ # pad_token_id=0,
110
+ position_embedding_type="rope",
111
+ rope_theta=10000.0,
112
+ rope_scaling=None,
113
+ classifier_dropout=None,
114
+ pack_qkv=True,
115
+ unpad_inputs=False,
116
+ use_memory_efficient_attention=False,
117
+ logn_attention_scale=False,
118
+ logn_attention_clip1=False,
119
+ **kwargs,
120
+ ):
121
+ super().__init__(**kwargs)
122
+
123
+ self.vocab_size = vocab_size
124
+ self.hidden_size = hidden_size
125
+ self.num_hidden_layers = num_hidden_layers
126
+ self.num_attention_heads = num_attention_heads
127
+ self.hidden_act = hidden_act
128
+ self.intermediate_size = intermediate_size
129
+ self.hidden_dropout_prob = hidden_dropout_prob
130
+ self.attention_probs_dropout_prob = attention_probs_dropout_prob
131
+ self.max_position_embeddings = max_position_embeddings
132
+ self.type_vocab_size = type_vocab_size
133
+ self.initializer_range = initializer_range
134
+ self.layer_norm_type = layer_norm_type
135
+ self.layer_norm_eps = layer_norm_eps
136
+ self.position_embedding_type = position_embedding_type
137
+ self.rope_theta = rope_theta
138
+ self.rope_scaling = rope_scaling
139
+ self.classifier_dropout = classifier_dropout
140
+
141
+ self.pack_qkv = pack_qkv
142
+ self.unpad_inputs = unpad_inputs
143
+ self.use_memory_efficient_attention = use_memory_efficient_attention
144
+ self.logn_attention_scale = logn_attention_scale
145
+ self.logn_attention_clip1 = logn_attention_clip1
.ipynb_checkpoints/modules-checkpoint.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "idx": 0,
4
+ "name": "0",
5
+ "path": "",
6
+ "type": "sentence_transformers.models.Transformer"
7
+ },
8
+ {
9
+ "idx": 1,
10
+ "name": "1",
11
+ "path": "1_Pooling",
12
+ "type": "sentence_transformers.models.Pooling"
13
+ }
14
+ ]
1_Pooling/config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "word_embedding_dimension": 768,
3
+ "pooling_mode_cls_token": true,
4
+ "pooling_mode_mean_tokens": false,
5
+ "pooling_mode_max_tokens": false,
6
+ "pooling_mode_mean_sqrt_len_tokens": false,
7
+ "pooling_mode_weightedmean_tokens": false,
8
+ "pooling_mode_lasttoken": false,
9
+ "include_prompt": true
10
+ }
README.md ADDED
@@ -0,0 +1,421 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - sentence-transformers
4
+ - sentence-similarity
5
+ - feature-extraction
6
+ - generated_from_trainer
7
+ - dataset_size:579077
8
+ - loss:MultipleNegativesRankingLoss
9
+ - loss:CosineSimilarityLoss
10
+ base_model: Alibaba-NLP/gte-multilingual-base
11
+ widget:
12
+ - source_sentence: 공공부문 채용의 경우 안전·건강 등 국민생활과 밀접한 서비스 중심으로 국가공무원을 1만 6000명 증원하고, 공공기관
13
+ 필수인력 확충을 추진한다.
14
+ sentences:
15
+ - 공공부문 채용의 경우 안전보건 등 국민생활과 밀접한 서비스를 중심으로 국가공무원을 1만6000명 늘리고 공공기관 필수인력 확충을 추진하기로
16
+ 했습니다.
17
+ - 백열등보단 간접 조명을 켜두고 독서를 하는게 좋을 것 같아
18
+ - 이번에 공개한 기관별 정규직 전환 실적은 ‘공공부문 비정규직 고용개선 시스템’(http://public.moel.go.kr)에서 확인할 수
19
+ 있다.
20
+ - source_sentence: 런던 여행을 하려는 분들에게 추천하고 싶은 곳 입니다.
21
+ sentences:
22
+ - 만약 내가 파리에 다시 온다면, 나는 여기에 머무를 것입니다.
23
+ - 지금의 위기를 새로운 기회와 발전의 원동력으로 삼겠습니다.
24
+ - 런던을 여행하고 싶은 분들에게 추천해 드리고 싶은 곳이에요.
25
+ - source_sentence: 이 절에서는 지불 과정에서 내부 통제의 중요성을 강조한다.
26
+ sentences:
27
+ - 그들은 스스로 세금을 부과함으로써 고속도로를 건설하고 새로운 버스 노선을 만들 것인가?
28
+ - 이 섹션에서는 전통적인 지불 프로세스, 전통적인 지불 프로세스 수정 및 지불 프로세스를 효과적으로 관리하기 위한 내부 제어의 중요성에 대해
29
+ 논의합니다.
30
+ - 이 절은 전통적인 지불 절차에 대한 조정을 다루지 않을 것이다.
31
+ - source_sentence: 스케이트보드를 타고 건물 계단을 내려가는 스케이트보드 타는 사람.
32
+ sentences:
33
+ - 그는 긴장이나 피로의 한계에 도달한 후 해시 물체를 얻기 시작했다.
34
+ - 스케이트보더가 목을 부러뜨린다
35
+ - 스케이트보드 타는 사람이 건물 계단을 타고 내려간다
36
+ - source_sentence: 1896년, 경제 및 행정 조직이 조정되었다.
37
+ sentences:
38
+ - 세 명의 여자가 밖에 있다.
39
+ - 1896년에 아무 관심도 없었다.
40
+ - 말레이 주 Selangor, Perak, Negeri Sembilan 및 Pahang의 연맹은 1896년에 경제 및 행정 조직을 조정하기 위해
41
+ 선포되었습니다.
42
+ pipeline_tag: sentence-similarity
43
+ library_name: sentence-transformers
44
+ metrics:
45
+ - pearson_cosine
46
+ - spearman_cosine
47
+ model-index:
48
+ - name: SentenceTransformer based on Alibaba-NLP/gte-multilingual-base
49
+ results:
50
+ - task:
51
+ type: semantic-similarity
52
+ name: Semantic Similarity
53
+ dataset:
54
+ name: sts dev
55
+ type: sts-dev
56
+ metrics:
57
+ - type: pearson_cosine
58
+ value: 0.9347680624097541
59
+ name: Pearson Cosine
60
+ - type: spearman_cosine
61
+ value: 0.8993438650317843
62
+ name: Spearman Cosine
63
+ ---
64
+
65
+ # SentenceTransformer based on Alibaba-NLP/gte-multilingual-base
66
+
67
+ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [Alibaba-NLP/gte-multilingual-base](https://huggingface.co/Alibaba-NLP/gte-multilingual-base). It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
68
+
69
+ ## Model Details
70
+
71
+ ### Model Description
72
+ - **Model Type:** Sentence Transformer
73
+ - **Base model:** [Alibaba-NLP/gte-multilingual-base](https://huggingface.co/Alibaba-NLP/gte-multilingual-base) <!-- at revision 7fc06782350c1a83f88b15dd4b38ef853d3b8503 -->
74
+ - **Maximum Sequence Length:** 8192 tokens
75
+ - **Output Dimensionality:** 768 dimensions
76
+ - **Similarity Function:** Cosine Similarity
77
+ <!-- - **Training Dataset:** Unknown -->
78
+ <!-- - **Language:** Unknown -->
79
+ <!-- - **License:** Unknown -->
80
+
81
+ ### Model Sources
82
+
83
+ - **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
84
+ - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
85
+ - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
86
+
87
+ ### Full Model Architecture
88
+
89
+ ```
90
+ SentenceTransformer(
91
+ (0): Transformer({'max_seq_length': 8192, 'do_lower_case': False}) with Transformer model: NewModel
92
+ (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
93
+ (2): Normalize()
94
+ )
95
+ ```
96
+
97
+ ## Usage
98
+
99
+ ### Direct Usage (Sentence Transformers)
100
+
101
+ First install the Sentence Transformers library:
102
+
103
+ ```bash
104
+ pip install -U sentence-transformers
105
+ ```
106
+
107
+ Then you can load this model and run inference.
108
+ ```python
109
+ from sentence_transformers import SentenceTransformer
110
+
111
+ # Download from the 🤗 Hub
112
+ model = SentenceTransformer("sentence_transformers_model_id")
113
+ # Run inference
114
+ sentences = [
115
+ '1896년, 경제 및 행정 조직이 조정되었다.',
116
+ '말레이 주 Selangor, Perak, Negeri Sembilan 및 Pahang의 연맹은 1896년에 경제 및 행정 조직을 조정하기 위해 선포되었습니다.',
117
+ '1896년에 아무 관심도 없었다.',
118
+ ]
119
+ embeddings = model.encode(sentences)
120
+ print(embeddings.shape)
121
+ # [3, 768]
122
+
123
+ # Get the similarity scores for the embeddings
124
+ similarities = model.similarity(embeddings, embeddings)
125
+ print(similarities.shape)
126
+ # [3, 3]
127
+ ```
128
+
129
+ <!--
130
+ ### Direct Usage (Transformers)
131
+
132
+ <details><summary>Click to see the direct usage in Transformers</summary>
133
+
134
+ </details>
135
+ -->
136
+
137
+ <!--
138
+ ### Downstream Usage (Sentence Transformers)
139
+
140
+ You can finetune this model on your own dataset.
141
+
142
+ <details><summary>Click to expand</summary>
143
+
144
+ </details>
145
+ -->
146
+
147
+ <!--
148
+ ### Out-of-Scope Use
149
+
150
+ *List how the model may foreseeably be misused and address what users ought not to do with the model.*
151
+ -->
152
+
153
+ ## Evaluation
154
+
155
+ ### Metrics
156
+
157
+ #### Semantic Similarity
158
+
159
+ * Dataset: `sts-dev`
160
+ * Evaluated with [<code>EmbeddingSimilarityEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.EmbeddingSimilarityEvaluator)
161
+
162
+ | Metric | Value |
163
+ |:--------------------|:-----------|
164
+ | pearson_cosine | 0.9348 |
165
+ | **spearman_cosine** | **0.8993** |
166
+
167
+ <!--
168
+ ## Bias, Risks and Limitations
169
+
170
+ *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
171
+ -->
172
+
173
+ <!--
174
+ ### Recommendations
175
+
176
+ *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
177
+ -->
178
+
179
+ ## Training Details
180
+
181
+ ### Training Datasets
182
+
183
+ #### Unnamed Dataset
184
+
185
+
186
+ * Size: 568,576 training samples
187
+ * Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>sentence_2</code>
188
+ * Approximate statistics based on the first 1000 samples:
189
+ | | sentence_0 | sentence_1 | sentence_2 |
190
+ |:--------|:-----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|
191
+ | type | string | string | string |
192
+ | details | <ul><li>min: 5 tokens</li><li>mean: 20.03 tokens</li><li>max: 122 tokens</li></ul> | <ul><li>min: 5 tokens</li><li>mean: 19.48 tokens</li><li>max: 88 tokens</li></ul> | <ul><li>min: 5 tokens</li><li>mean: 14.72 tokens</li><li>max: 47 tokens</li></ul> |
193
+ * Samples:
194
+ | sentence_0 | sentence_1 | sentence_2 |
195
+ |:-------------------------------------------------------|:-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:---------------------------------------------------|
196
+ | <code>사람들이 자동차를 좋아한다.</code> | <code>사람들은 클래식 자동차를 존경한다.</code> | <code>사람들이 줄을 서서 콘서트를 기다리고 있다.</code> |
197
+ | <code>그가 말을 타고 가면서 피의 강물이 흐르고 남자는 안장에 털썩 주저앉았다.</code> | <code>그 남자는 말을 타다가 피를 흘리고 있었다.</code> | <code>남자는 안장에 똑바로 앉았다.</code> |
198
+ | <code>그 자료는 일년 중 일부만을 다루었다.</code> | <code>올해 3월 보고된 2001년 자료는 예비 자료로 간주해야 하지만(반년만 다뤄지고 새로운 데이터 시스템에 기대되는 통상적인 종류의 스타트업 문제를 반영했다), 이미 공사가 그 어느 때보다 전국적으로 가능한 법률 서비스 관행에 대한 완전한 그림을 제공할 수 있는 풍부한 정보를 만들어냈다.</code> | <code>그 자료는 일년 중 일부만을 다루었을 뿐 전혀 도움이 되지 않았다.</code> |
199
+ * Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
200
+ ```json
201
+ {
202
+ "scale": 20.0,
203
+ "similarity_fct": "cos_sim"
204
+ }
205
+ ```
206
+
207
+ #### Unnamed Dataset
208
+
209
+
210
+ * Size: 10,501 training samples
211
+ * Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>label</code>
212
+ * Approximate statistics based on the first 1000 samples:
213
+ | | sentence_0 | sentence_1 | label |
214
+ |:--------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:---------------------------------------------------------------|
215
+ | type | string | string | float |
216
+ | details | <ul><li>min: 7 tokens</li><li>mean: 20.82 tokens</li><li>max: 73 tokens</li></ul> | <ul><li>min: 8 tokens</li><li>mean: 19.95 tokens</li><li>max: 63 tokens</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.44</li><li>max: 1.0</li></ul> |
217
+ * Samples:
218
+ | sentence_0 | sentence_1 | label |
219
+ |:------------------------------------------|:----------------------------------------------|:------------------|
220
+ | <code>제 학교 성적표를 받기로한 메일을 알 수 있을까요?</code> | <code>쿠팡은 여태까지 배송 주문 확인 메일을 몇 통 보냈어?</code> | <code>0.04</code> |
221
+ | <code>지냈던 숙소 중에서 제일 마음에 들었습니다.</code> | <code>지금 까지 이용한 에어비앤비 중에서 제일 마음에 들었어요.</code> | <code>0.6</code> |
222
+ | <code>눈 내릴 때 운전은 안됩니다.</code> | <code>눈 내릴 때 운전은 위험해서 안돼.</code> | <code>0.74</code> |
223
+ * Loss: [<code>CosineSimilarityLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosinesimilarityloss) with these parameters:
224
+ ```json
225
+ {
226
+ "loss_fct": "torch.nn.modules.loss.MSELoss"
227
+ }
228
+ ```
229
+
230
+ ### Training Hyperparameters
231
+ #### Non-Default Hyperparameters
232
+
233
+ - `eval_strategy`: steps
234
+ - `per_device_train_batch_size`: 32
235
+ - `per_device_eval_batch_size`: 32
236
+ - `batch_sampler`: no_duplicates
237
+ - `multi_dataset_batch_sampler`: round_robin
238
+
239
+ #### All Hyperparameters
240
+ <details><summary>Click to expand</summary>
241
+
242
+ - `overwrite_output_dir`: False
243
+ - `do_predict`: False
244
+ - `eval_strategy`: steps
245
+ - `prediction_loss_only`: True
246
+ - `per_device_train_batch_size`: 32
247
+ - `per_device_eval_batch_size`: 32
248
+ - `per_gpu_train_batch_size`: None
249
+ - `per_gpu_eval_batch_size`: None
250
+ - `gradient_accumulation_steps`: 1
251
+ - `eval_accumulation_steps`: None
252
+ - `torch_empty_cache_steps`: None
253
+ - `learning_rate`: 5e-05
254
+ - `weight_decay`: 0.0
255
+ - `adam_beta1`: 0.9
256
+ - `adam_beta2`: 0.999
257
+ - `adam_epsilon`: 1e-08
258
+ - `max_grad_norm`: 1.0
259
+ - `num_train_epochs`: 3
260
+ - `max_steps`: -1
261
+ - `lr_scheduler_type`: linear
262
+ - `lr_scheduler_kwargs`: {}
263
+ - `warmup_ratio`: 0.0
264
+ - `warmup_steps`: 0
265
+ - `log_level`: passive
266
+ - `log_level_replica`: warning
267
+ - `log_on_each_node`: True
268
+ - `logging_nan_inf_filter`: True
269
+ - `save_safetensors`: True
270
+ - `save_on_each_node`: False
271
+ - `save_only_model`: False
272
+ - `restore_callback_states_from_checkpoint`: False
273
+ - `no_cuda`: False
274
+ - `use_cpu`: False
275
+ - `use_mps_device`: False
276
+ - `seed`: 42
277
+ - `data_seed`: None
278
+ - `jit_mode_eval`: False
279
+ - `use_ipex`: False
280
+ - `bf16`: False
281
+ - `fp16`: False
282
+ - `fp16_opt_level`: O1
283
+ - `half_precision_backend`: auto
284
+ - `bf16_full_eval`: False
285
+ - `fp16_full_eval`: False
286
+ - `tf32`: None
287
+ - `local_rank`: 0
288
+ - `ddp_backend`: None
289
+ - `tpu_num_cores`: None
290
+ - `tpu_metrics_debug`: False
291
+ - `debug`: []
292
+ - `dataloader_drop_last`: False
293
+ - `dataloader_num_workers`: 0
294
+ - `dataloader_prefetch_factor`: None
295
+ - `past_index`: -1
296
+ - `disable_tqdm`: False
297
+ - `remove_unused_columns`: True
298
+ - `label_names`: None
299
+ - `load_best_model_at_end`: False
300
+ - `ignore_data_skip`: False
301
+ - `fsdp`: []
302
+ - `fsdp_min_num_params`: 0
303
+ - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
304
+ - `fsdp_transformer_layer_cls_to_wrap`: None
305
+ - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
306
+ - `deepspeed`: None
307
+ - `label_smoothing_factor`: 0.0
308
+ - `optim`: adamw_torch
309
+ - `optim_args`: None
310
+ - `adafactor`: False
311
+ - `group_by_length`: False
312
+ - `length_column_name`: length
313
+ - `ddp_find_unused_parameters`: None
314
+ - `ddp_bucket_cap_mb`: None
315
+ - `ddp_broadcast_buffers`: False
316
+ - `dataloader_pin_memory`: True
317
+ - `dataloader_persistent_workers`: False
318
+ - `skip_memory_metrics`: True
319
+ - `use_legacy_prediction_loop`: False
320
+ - `push_to_hub`: False
321
+ - `resume_from_checkpoint`: None
322
+ - `hub_model_id`: None
323
+ - `hub_strategy`: every_save
324
+ - `hub_private_repo`: False
325
+ - `hub_always_push`: False
326
+ - `gradient_checkpointing`: False
327
+ - `gradient_checkpointing_kwargs`: None
328
+ - `include_inputs_for_metrics`: False
329
+ - `include_for_metrics`: []
330
+ - `eval_do_concat_batches`: True
331
+ - `fp16_backend`: auto
332
+ - `push_to_hub_model_id`: None
333
+ - `push_to_hub_organization`: None
334
+ - `mp_parameters`:
335
+ - `auto_find_batch_size`: False
336
+ - `full_determinism`: False
337
+ - `torchdynamo`: None
338
+ - `ray_scope`: last
339
+ - `ddp_timeout`: 1800
340
+ - `torch_compile`: False
341
+ - `torch_compile_backend`: None
342
+ - `torch_compile_mode`: None
343
+ - `dispatch_batches`: None
344
+ - `split_batches`: None
345
+ - `include_tokens_per_second`: False
346
+ - `include_num_input_tokens_seen`: False
347
+ - `neftune_noise_alpha`: None
348
+ - `optim_target_modules`: None
349
+ - `batch_eval_metrics`: False
350
+ - `eval_on_start`: False
351
+ - `use_liger_kernel`: False
352
+ - `eval_use_gather_object`: False
353
+ - `average_tokens_across_devices`: False
354
+ - `prompts`: None
355
+ - `batch_sampler`: no_duplicates
356
+ - `multi_dataset_batch_sampler`: round_robin
357
+
358
+ </details>
359
+
360
+ ### Training Logs
361
+ | Epoch | Step | Training Loss | sts-dev_spearman_cosine |
362
+ |:------:|:----:|:-------------:|:-----------------------:|
363
+ | 0.7599 | 500 | 0.324 | - |
364
+ | 1.0015 | 659 | - | 0.8993 |
365
+
366
+
367
+ ### Framework Versions
368
+ - Python: 3.11.10
369
+ - Sentence Transformers: 3.3.0
370
+ - Transformers: 4.46.2
371
+ - PyTorch: 2.4.0+cu121
372
+ - Accelerate: 1.1.1
373
+ - Datasets: 3.1.0
374
+ - Tokenizers: 0.20.3
375
+
376
+ ## Citation
377
+
378
+ ### BibTeX
379
+
380
+ #### Sentence Transformers
381
+ ```bibtex
382
+ @inproceedings{reimers-2019-sentence-bert,
383
+ title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
384
+ author = "Reimers, Nils and Gurevych, Iryna",
385
+ booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
386
+ month = "11",
387
+ year = "2019",
388
+ publisher = "Association for Computational Linguistics",
389
+ url = "https://arxiv.org/abs/1908.10084",
390
+ }
391
+ ```
392
+
393
+ #### MultipleNegativesRankingLoss
394
+ ```bibtex
395
+ @misc{henderson2017efficient,
396
+ title={Efficient Natural Language Response Suggestion for Smart Reply},
397
+ author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
398
+ year={2017},
399
+ eprint={1705.00652},
400
+ archivePrefix={arXiv},
401
+ primaryClass={cs.CL}
402
+ }
403
+ ```
404
+
405
+ <!--
406
+ ## Glossary
407
+
408
+ *Clearly define terms in order to be accessible across audiences.*
409
+ -->
410
+
411
+ <!--
412
+ ## Model Card Authors
413
+
414
+ *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
415
+ -->
416
+
417
+ <!--
418
+ ## Model Card Contact
419
+
420
+ *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
421
+ -->
config.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "Alibaba-NLP/gte-multilingual-base",
3
+ "architectures": [
4
+ "NewModel"
5
+ ],
6
+ "attention_probs_dropout_prob": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "configuration.NewConfig",
9
+ "AutoModel": "Alibaba-NLP/new-impl--modeling.NewModel",
10
+ "AutoModelForMaskedLM": "Alibaba-NLP/new-impl--modeling.NewForMaskedLM",
11
+ "AutoModelForMultipleChoice": "Alibaba-NLP/new-impl--modeling.NewForMultipleChoice",
12
+ "AutoModelForQuestionAnswering": "Alibaba-NLP/new-impl--modeling.NewForQuestionAnswering",
13
+ "AutoModelForSequenceClassification": "Alibaba-NLP/new-impl--modeling.NewForSequenceClassification",
14
+ "AutoModelForTokenClassification": "Alibaba-NLP/new-impl--modeling.NewForTokenClassification"
15
+ },
16
+ "classifier_dropout": 0.0,
17
+ "hidden_act": "gelu",
18
+ "hidden_dropout_prob": 0.1,
19
+ "hidden_size": 768,
20
+ "id2label": {
21
+ "0": "LABEL_0"
22
+ },
23
+ "initializer_range": 0.02,
24
+ "intermediate_size": 3072,
25
+ "label2id": {
26
+ "LABEL_0": 0
27
+ },
28
+ "layer_norm_eps": 1e-12,
29
+ "layer_norm_type": "layer_norm",
30
+ "logn_attention_clip1": false,
31
+ "logn_attention_scale": false,
32
+ "max_position_embeddings": 8192,
33
+ "model_type": "new",
34
+ "num_attention_heads": 12,
35
+ "num_hidden_layers": 12,
36
+ "pack_qkv": true,
37
+ "pad_token_id": 1,
38
+ "position_embedding_type": "rope",
39
+ "rope_scaling": {
40
+ "factor": 8.0,
41
+ "type": "ntk"
42
+ },
43
+ "rope_theta": 20000,
44
+ "torch_dtype": "float32",
45
+ "transformers_version": "4.46.2",
46
+ "type_vocab_size": 1,
47
+ "unpad_inputs": false,
48
+ "use_memory_efficient_attention": false,
49
+ "vocab_size": 250048
50
+ }
config_sentence_transformers.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "__version__": {
3
+ "sentence_transformers": "3.3.0",
4
+ "transformers": "4.46.2",
5
+ "pytorch": "2.4.0+cu121"
6
+ },
7
+ "prompts": {},
8
+ "default_prompt_name": null,
9
+ "similarity_fn_name": "cosine"
10
+ }
configuration.py ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2024 The GTE Team Authors and Alibaba Group.
3
+ # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
4
+ #
5
+ # Licensed under the Apache License, Version 2.0 (the "License");
6
+ # you may not use this file except in compliance with the License.
7
+ # You may obtain a copy of the License at
8
+ #
9
+ # http://www.apache.org/licenses/LICENSE-2.0
10
+ #
11
+ # Unless required by applicable law or agreed to in writing, software
12
+ # distributed under the License is distributed on an "AS IS" BASIS,
13
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ # See the License for the specific language governing permissions and
15
+ # limitations under the License.
16
+ """ NEW model configuration"""
17
+ from transformers.configuration_utils import PretrainedConfig
18
+ from transformers.utils import logging
19
+
20
+ logger = logging.get_logger(__name__)
21
+
22
+
23
+ class NewConfig(PretrainedConfig):
24
+ r"""
25
+ This is the configuration class to store the configuration of a [`NewModel`] or a [`TFNewModel`]. It is used to
26
+ instantiate a NEW model according to the specified arguments, defining the model architecture. Instantiating a
27
+ configuration with the defaults will yield a similar configuration to that of the NEW
28
+ [izhx/new-base-en](https://huggingface.co/izhx/new-base-en) architecture.
29
+
30
+ Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
31
+ documentation from [`PretrainedConfig`] for more information.
32
+
33
+
34
+ Args:
35
+ vocab_size (`int`, *optional*, defaults to 30522):
36
+ Vocabulary size of the NEW model. Defines the number of different tokens that can be represented by the
37
+ `inputs_ids` passed when calling [`NewModel`] or [`TFNewModel`].
38
+ hidden_size (`int`, *optional*, defaults to 768):
39
+ Dimensionality of the encoder layers and the pooler layer.
40
+ num_hidden_layers (`int`, *optional*, defaults to 12):
41
+ Number of hidden layers in the Transformer encoder.
42
+ num_attention_heads (`int`, *optional*, defaults to 12):
43
+ Number of attention heads for each attention layer in the Transformer encoder.
44
+ intermediate_size (`int`, *optional*, defaults to 3072):
45
+ Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
46
+ hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`):
47
+ The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
48
+ `"relu"`, `"silu"` and `"gelu_new"` are supported.
49
+ hidden_dropout_prob (`float`, *optional*, defaults to 0.1):
50
+ The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
51
+ attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):
52
+ The dropout ratio for the attention probabilities.
53
+ max_position_embeddings (`int`, *optional*, defaults to 512):
54
+ The maximum sequence length that this model might ever be used with. Typically set this to something large
55
+ just in case (e.g., 512 or 1024 or 2048).
56
+ type_vocab_size (`int`, *optional*, defaults to 2):
57
+ The vocabulary size of the `token_type_ids` passed when calling [`NewModel`] or [`TFNewModel`].
58
+ initializer_range (`float`, *optional*, defaults to 0.02):
59
+ The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
60
+ layer_norm_eps (`float`, *optional*, defaults to 1e-12):
61
+ The epsilon used by the layer normalization layers.
62
+ position_embedding_type (`str`, *optional*, defaults to `"rope"`):
63
+ Type of position embedding. Choose one of `"absolute"`, `"rope"`.
64
+ rope_theta (`float`, *optional*, defaults to 10000.0):
65
+ The base period of the RoPE embeddings.
66
+ rope_scaling (`Dict`, *optional*):
67
+ Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
68
+ strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
69
+ `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
70
+ `max_position_embeddings` to the expected new maximum. See the following thread for more information on how
71
+ these scaling strategies behave:
72
+ https://www.reddit.com/r/LocalLLaMA/comments/14mrgpr/dynamically_scaled_rope_further_increases/. This is an
73
+ experimental feature, subject to breaking API changes in future versions.
74
+ classifier_dropout (`float`, *optional*):
75
+ The dropout ratio for the classification head.
76
+
77
+ Examples:
78
+
79
+ ```python
80
+ >>> from transformers import NewConfig, NewModel
81
+
82
+ >>> # Initializing a NEW izhx/new-base-en style configuration
83
+ >>> configuration = NewConfig()
84
+
85
+ >>> # Initializing a model (with random weights) from the izhx/new-base-en style configuration
86
+ >>> model = NewModel(configuration)
87
+
88
+ >>> # Accessing the model configuration
89
+ >>> configuration = model.config
90
+ ```"""
91
+
92
+ model_type = "new"
93
+
94
+ def __init__(
95
+ self,
96
+ vocab_size=30528,
97
+ hidden_size=768,
98
+ num_hidden_layers=12,
99
+ num_attention_heads=12,
100
+ intermediate_size=3072,
101
+ hidden_act="gelu",
102
+ hidden_dropout_prob=0.1,
103
+ attention_probs_dropout_prob=0.0,
104
+ max_position_embeddings=2048,
105
+ type_vocab_size=1,
106
+ initializer_range=0.02,
107
+ layer_norm_type='layer_norm',
108
+ layer_norm_eps=1e-12,
109
+ # pad_token_id=0,
110
+ position_embedding_type="rope",
111
+ rope_theta=10000.0,
112
+ rope_scaling=None,
113
+ classifier_dropout=None,
114
+ pack_qkv=True,
115
+ unpad_inputs=False,
116
+ use_memory_efficient_attention=False,
117
+ logn_attention_scale=False,
118
+ logn_attention_clip1=False,
119
+ **kwargs,
120
+ ):
121
+ super().__init__(**kwargs)
122
+
123
+ self.vocab_size = vocab_size
124
+ self.hidden_size = hidden_size
125
+ self.num_hidden_layers = num_hidden_layers
126
+ self.num_attention_heads = num_attention_heads
127
+ self.hidden_act = hidden_act
128
+ self.intermediate_size = intermediate_size
129
+ self.hidden_dropout_prob = hidden_dropout_prob
130
+ self.attention_probs_dropout_prob = attention_probs_dropout_prob
131
+ self.max_position_embeddings = max_position_embeddings
132
+ self.type_vocab_size = type_vocab_size
133
+ self.initializer_range = initializer_range
134
+ self.layer_norm_type = layer_norm_type
135
+ self.layer_norm_eps = layer_norm_eps
136
+ self.position_embedding_type = position_embedding_type
137
+ self.rope_theta = rope_theta
138
+ self.rope_scaling = rope_scaling
139
+ self.classifier_dropout = classifier_dropout
140
+
141
+ self.pack_qkv = pack_qkv
142
+ self.unpad_inputs = unpad_inputs
143
+ self.use_memory_efficient_attention = use_memory_efficient_attention
144
+ self.logn_attention_scale = logn_attention_scale
145
+ self.logn_attention_clip1 = logn_attention_clip1
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5b9411237518e546ca2e4159b79328a76af59bee7351e250227fd9b4924e93a
3
+ size 1221487872
modules.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "idx": 0,
4
+ "name": "0",
5
+ "path": "",
6
+ "type": "sentence_transformers.models.Transformer"
7
+ },
8
+ {
9
+ "idx": 1,
10
+ "name": "1",
11
+ "path": "1_Pooling",
12
+ "type": "sentence_transformers.models.Pooling"
13
+ },
14
+ {
15
+ "idx": 2,
16
+ "name": "2",
17
+ "path": "2_Normalize",
18
+ "type": "sentence_transformers.models.Normalize"
19
+ }
20
+ ]
sentence_bert_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "max_seq_length": 8192,
3
+ "do_lower_case": false
4
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "cls_token": {
10
+ "content": "<s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "eos_token": {
17
+ "content": "</s>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "mask_token": {
24
+ "content": "<mask>",
25
+ "lstrip": true,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ },
30
+ "pad_token": {
31
+ "content": "<pad>",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false
36
+ },
37
+ "sep_token": {
38
+ "content": "</s>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false
43
+ },
44
+ "unk_token": {
45
+ "content": "<unk>",
46
+ "lstrip": false,
47
+ "normalized": false,
48
+ "rstrip": false,
49
+ "single_word": false
50
+ }
51
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa7a6ad87a7ce8fe196787355f6af7d03aee94d19c54a5eb1392ed18c8ef451a
3
+ size 17082988
tokenizer_config.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "0": {
4
+ "content": "<s>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false,
9
+ "special": true
10
+ },
11
+ "1": {
12
+ "content": "<pad>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false,
17
+ "special": true
18
+ },
19
+ "2": {
20
+ "content": "</s>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "3": {
28
+ "content": "<unk>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ },
35
+ "250001": {
36
+ "content": "<mask>",
37
+ "lstrip": true,
38
+ "normalized": false,
39
+ "rstrip": false,
40
+ "single_word": false,
41
+ "special": true
42
+ }
43
+ },
44
+ "bos_token": "<s>",
45
+ "clean_up_tokenization_spaces": true,
46
+ "cls_token": "<s>",
47
+ "eos_token": "</s>",
48
+ "mask_token": "<mask>",
49
+ "model_max_length": 8192,
50
+ "pad_token": "<pad>",
51
+ "sep_token": "</s>",
52
+ "tokenizer_class": "XLMRobertaTokenizer",
53
+ "unk_token": "<unk>"
54
+ }