Upload 10 files

1_Pooling/config.json

@@ -0,0 +1,7 @@
+{
+  "word_embedding_dimension": 312,
+  "pooling_mode_cls_token": true,
+  "pooling_mode_mean_tokens": false,
+  "pooling_mode_max_tokens": false,
+  "pooling_mode_mean_sqrt_len_tokens": false
+}

README.md

@@ -1,3 +1,84 @@
 ---
+language:
+- ru
+- en
+
+pipeline_tag: sentence-similarity
+
+tags:
+- russian
+- pretraining
+- embeddings
+- tiny
+- feature-extraction
+- sentence-similarity
+- sentence-transformers
+- transformers
+- mteb
+
+datasets:
+- IlyaGusev/gazeta
+- zloelias/lenta-ru
+- HuggingFaceFW/fineweb-2
+- HuggingFaceFW/fineweb
+
 license: mit
+
+base_model: sergeyzh/rubert-tiny-turbo
+
 ---
+
+Быстрая модель BERT для задач симметричного перефразирования (STS, поиск парафраз, дедупликация) и логического вывода (NLI). Получена дистилляцией эмбеддингов русских и английских текстов [Qwen/Qwen3-Embedding-4B](https://huggingface.co/Qwen/Qwen3-Embedding-4B) в [rubert-tiny-turbo](https://huggingface.co/sergeyzh/rubert-tiny-turbo).
+
+Модель может использоваться в качестве базовой для дообучения под пользовательские задачи классификации и кластеризации.
+
+Основные характеристики модели: 
+ - размер ембеддинга - 312, 
+ - длина контекста - 512, 
+ - слоёв - 3,
+ - префиксы - не требуются. 
+
+
+
+## Использование
+
+```Python
+from sentence_transformers import SentenceTransformer
+
+model = SentenceTransformer('sergeyzh/rubert-tiny-sts-v2')
+
+sentences = ["привет мир", "hello world", "здравствуй вселенная"]
+embeddings = model.encode(sentences)
+print(model.similarity(embeddings, embeddings))
+```
+
+
+
+## Метрики
+
+Оценки модели на задачах для русского языка:
+
+| Model Name | RuSTS Benchmark STS | RU ParaPhraser STS | STS22,v2 | TERRa Classification | Average |
+| :--- | :---: | :---: | :---: | :---: | :---: |
+| Qwen3-Embedding-4B | 0,888 | 0,766 | 0,701 | 0,666 | 0,755 |
+| multilingual-e5-large-instruct | 0,840 | 0,754 | 0,706 | 0,639 | 0,735 |
+| Qwen3-Embedding-0.6B | 0,842 | 0,721 | 0,662 | 0,607 | 0,708 |
+| **rubert-tiny-sts-v2** | 0,830 | 0,736 | 0,646 | 0,616 | 0,707 |
+| bge-m3 | 0,797 | 0,749 | 0,663 | 0,607 | 0,704 |
+| rubert-tiny-turbo | 0,787 | 0,721 | 0,646 | 0,563 | 0,679 |
+| multilingual-e5-base | 0,796 | 0,702 | 0,607 | 0,550 | 0,664 |
+
+
+
+Оценки модели на задачах для английского языка:
+
+| Model Name | STS12 | STS13 | STS14 | STS15 | STS17 | STS22,v2 | STS Benchmark | Average |
+| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
+| Qwen3-Embedding-4B | 0,866 | 0,944 | 0,909 | 0,938 | 0,918 | 0,730 | 0,937 | 0,892 |
+| Qwen3-Embedding-0.6B | 0,830 | 0,918 | 0,871 | 0,914 | 0,855 | 0,718 | 0,911 | 0,860 |
+| multilingual-e5-large-instruct | 0,825 | 0,881 | 0,848 | 0,910 | 0,860 | 0,690 | 0,884 | 0,842 |
+| bge-m3 | 0,787 | 0,796 | 0,790 | 0,878 | 0,796 | 0,700 | 0,849 | 0,800 |
+| multilingual-e5-base | 0,767 | 0,780 | 0,766 | 0,882 | 0,783 | 0,646 | 0,856 | 0,783 |
+| **rubert-tiny-sts-v2** | 0,761 | 0,836 | 0,804 | 0,856 | 0,821 | 0,547 | 0,833 | 0,780 |
+| rubert-tiny-turbo | 0,662 | 0,603 | 0,661 | 0,774 | 0,272 | 0,334 | 0,690 | 0,571 |
+

config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "sergeyzh/rubert-tiny-sts-v2",
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "emb_size": 312,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 312,
+  "initializer_range": 0.02,
+  "intermediate_size": 600,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 2048,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 3,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "transformers_version": "4.57.6",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 83828
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44dbeacf5eae4e38e51ea44d19b8e7c40b350e24546390c402a798cf6d363150
+size 116781184

modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.models.Normalize"
+  }
+]

sentence_bert_config.json

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

special_tokens_map.json

@@ -0,0 +1,37 @@
+{
+  "cls_token": {
+    "content": "[CLS]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "[MASK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "[PAD]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "[SEP]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "[UNK]",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,65 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": false,
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "max_length": 512,
+  "model_max_length": 2048,
+  "never_split": null,
+  "pad_to_multiple_of": null,
+  "pad_token": "[PAD]",
+  "pad_token_type_id": 0,
+  "padding_side": "right",
+  "sep_token": "[SEP]",
+  "stride": 0,
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "[UNK]"
+}