init embedding model

.gitattributes

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+pytorch_model.bin filter=lfs diff=lfs merge=lfs -text
+.git/lfs/objects/cb/fa/cbfae04cc7f3063949a7d81258e185cd31249892768c15e27b70d4797f42b902 filter=lfs diff=lfs merge=lfs -text
+model.safetensors filter=lfs diff=lfs merge=lfs -text
+.git/lfs/objects/2d/81/2d8135cb6ff79bf4303fb0afd11808791c9123ae8ee753fccd480207e347963e filter=lfs diff=lfs merge=lfs -text

1_Pooling/config.json

@@ -0,0 +1,9 @@
+{
+  "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
+}

README.md

@@ -0,0 +1,168 @@
+---
+pipeline_tag: sentence-similarity
+tags:
+- sentence-transformers
+- feature-extraction
+- sentence-similarity
+- transformers
+library_name: generic
+language:
+- vi
+widget:
+- source_sentence: Làm thế nào Đại học Bách khoa Hà Nội thu hút sinh viên quốc tế?
+  sentences:
+  - >-
+    Đại học Bách khoa Hà Nội đã phát triển các chương trình đào tạo bằng tiếng
+    Anh để làm cho việc học tại đây dễ dàng hơn cho sinh viên quốc tế.
+  - >-
+    Môi trường học tập đa dạng và sự hỗ trợ đầy đủ cho sinh viên quốc tế tại Đại
+    học Bách khoa Hà Nội giúp họ thích nghi nhanh chóng.
+  - Hà Nội có khí hậu mát mẻ vào mùa thu.
+  - Các món ăn ở Hà Nội rất ngon và đa dạng.
+license: apache-2.0
+---
+
+# bkai-foundation-models/vietnamese-bi-encoder
+
+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.
+
+We train the model on a merged training dataset that consists of: 
+  - MS Macro (translated into Vietnamese)
+  - SQuAD v2  (translated into Vietnamese)
+  - 80% of the training set from the Legal Text Retrieval Zalo 2021 challenge
+
+We use [phobert-base-v2](https://github.com/VinAIResearch/PhoBERT) as the pre-trained backbone.
+
+Here are the results on the remaining 20% of the training set from the Legal Text Retrieval Zalo 2021 challenge:
+
+|     Pretrained Model          |     Training Datasets                  |     Acc@1    |     Acc@10    |     Acc@100    |     Pre@10    |     MRR@10    |
+|-------------------------------|---------------------------------------|:------------:|:-------------:|:--------------:|:-------------:|:-------------:|
+|     [Vietnamese-SBERT](https://huggingface.co/keepitreal/vietnamese-sbert)     |     -                                 |     32.34    |      52.97    |      89.84     |      7.05     |      45.30    |
+|     PhoBERT-base-v2           |     MSMACRO                           |     47.81    |      77.19    |      92.34     |      7.72     |      58.37    |
+|     PhoBERT-base-v2                            |     MSMACRO + SQuADv2.0 + 80% Zalo    |     73.28    |      93.59    |      98.85     |      9.36     |      80.73    |
+
+
+<!--- 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
+
+# INPUT TEXT MUST BE ALREADY WORD-SEGMENTED!
+sentences = ["Cô ấy là một người vui_tính .", "Cô ấy cười nói suốt cả ngày ."]
+
+model = SentenceTransformer('bkai-foundation-models/vietnamese-bi-encoder')
+embeddings = model.encode(sentences)
+print(embeddings)
+```
+
+
+## Usage (Widget HuggingFace)
+The widget use custom pipeline on top of the default pipeline by adding additional word segmenter before PhobertTokenizer. So you do not need to segment words before using the API:
+
+An example could be seen in Hosted inference API.
+ 
+
+## 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, we could use pyvi, underthesea, RDRSegment to segment words
+sentences = ['Cô ấy là một người vui_tính .', 'Cô ấy cười nói suốt cả ngày .']
+
+# Load model from HuggingFace Hub
+tokenizer = AutoTokenizer.from_pretrained('bkai-foundation-models/vietnamese-bi-encoder')
+model = AutoModel.from_pretrained('bkai-foundation-models/vietnamese-bi-encoder')
+
+# 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)
+```
+
+## Training
+
+The model was trained with the parameters:
+
+**DataLoader**:
+
+`torch.utils.data.dataloader.DataLoader` of length 17584 with parameters:
+
+```
+{'batch_size': 32, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}
+```
+
+**Loss**:
+
+`sentence_transformers.losses.MultipleNegativesRankingLoss.MultipleNegativesRankingLoss` with parameters:
+
+```
+{'scale': 20.0, 'similarity_fct': 'cos_sim'}
+```
+
+Parameters of the fit()-Method:
+
+```
+{
+    "epochs": 15,
+    "evaluation_steps": 0,
+    "evaluator": "NoneType",
+    "max_grad_norm": 1,
+    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
+    "optimizer_params": {
+        "lr": 2e-05
+    },
+    "scheduler": "WarmupLinear",
+    "steps_per_epoch": null,
+    "warmup_steps": 1000,
+    "weight_decay": 0.01
+}
+```
+
+## Full Model Architecture
+
+```
+SentenceTransformer(
+  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False}) with Transformer model: RobertaModel
+  (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})
+)
+```
+
+### Please cite our manuscript if this dataset is used for your work
+```
+  @article{duc2024towards,
+    title={Towards Comprehensive Vietnamese Retrieval-Augmented Generation and Large Language Models},
+    author={Nguyen Quang Duc, Le Hai Son, Nguyen Duc Nhan, Nguyen Dich Nhat Minh, Le Thanh Huong, Dinh Viet Sang},
+    journal={arXiv preprint arXiv:2403.01616},
+    year={2024}
+  }
+```

added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "<mask>": 64000
+}

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "output/train_bi-encoder-mnrl-vinai-phobert-base-v2-margin_3.0-2023-08-27_23-13-25/",
+  "architectures": [
+    "RobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 258,
+  "model_type": "roberta",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "tokenizer_class": "PhobertTokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.38.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 64001
+}

config_sentence_transformers.json

@@ -0,0 +1,7 @@
+{
+  "__version__": {
+    "sentence_transformers": "2.2.2",
+    "transformers": "4.32.0",
+    "pytorch": "2.0.0+cu117"
+  }
+}

custom_tokenizer.py

@@ -0,0 +1,11 @@
+from transformers import PhobertTokenizer
+from pyvi import ViTokenizer
+
+
+class CustomPhobertTokenizer(PhobertTokenizer):
+    def rdr_segment(self, text):
+        return ViTokenizer.tokenize(text)
+
+    def _tokenize(self, text):
+        segmented_text = self.rdr_segment(text)
+        return super()._tokenize(segmented_text)

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e681accadaec87e79901db0c3f68e33d996cba334633b6dd0b2483dba4f398e0
+size 540015464

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"
+  }
+]

pipeline.py

@@ -0,0 +1,76 @@
+from typing import Dict, List, Union
+import torch
+from transformers import AutoModel
+from custom_tokenizer import CustomPhobertTokenizer
+
+
+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
+    )
+
+
+class PreTrainedPipeline:
+    def __init__(self, path="."):
+        self.model = AutoModel.from_pretrained(path)
+        self.tokenizer = CustomPhobertTokenizer.from_pretrained(path)
+
+    def __call__(self, inputs: Dict[str, Union[str, List[str]]]) -> List[float]:
+        """
+        Args:
+            inputs (Dict[str, Union[str, List[str]]]):
+                a dictionary containing a query sentence and a list of key sentences
+        """
+
+        # Combine the query sentence and key sentences into one list
+        sentences = [inputs["source_sentence"]] + inputs["sentences"]
+
+        # Tokenize sentences
+        encoded_input = self.tokenizer(
+            sentences, padding=True, truncation=True, return_tensors="pt"
+        )
+
+        # Compute token embeddings
+        with torch.no_grad():
+            model_output = self.model(**encoded_input)
+
+        # Perform pooling to get sentence embeddings
+        sentence_embeddings = mean_pooling(
+            model_output, encoded_input["attention_mask"]
+        )
+
+        # Separate the query embedding from the key embeddings
+        query_embedding = sentence_embeddings[0]
+        key_embeddings = sentence_embeddings[1:]
+
+        # Compute cosine similarities (or any other comparison method you prefer)
+        cosine_similarities = torch.nn.functional.cosine_similarity(
+            query_embedding.unsqueeze(0), key_embeddings
+        )
+
+        # Convert the tensor of cosine similarities to a list of floats
+        scores = cosine_similarities.tolist()
+
+        return scores
+
+
+if __name__ == "__main__":
+    inputs = {
+        "source_sentence": "Anh ấy đang là sinh viên năm cuối",
+        "sentences": [
+            "Anh ấy học tại Đại học Bách khoa Hà Nội, chuyên ngành Khoa học máy tính",
+            "Anh ấy đang làm việc tại nhà máy sản xuất linh kiện điện tử",
+            "Anh ấy chuẩn bị đi du học nước ngoài",
+            "Anh ấy sắp mở cửa hàng bán mỹ phẩm",
+            "Nhà anh ấy có rất nhiều cây cảnh",
+        ],
+    }
+
+    pipeline = PreTrainedPipeline()
+    res = pipeline(inputs)

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d8135cb6ff79bf4303fb0afd11808791c9123ae8ee753fccd480207e347963e
+size 540057065

requirements.txt

@@ -0,0 +1 @@
+pyvi>=0.1.1

sentence_bert_config.json

@@ -0,0 +1,4 @@
+{
+  "max_seq_length": 256,
+  "do_lower_case": false
+}

special_tokens_map.json

@@ -0,0 +1,9 @@
+{
+  "bos_token": "<s>",
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "unk_token": "<unk>"
+}

tokenizer_config.json

@@ -0,0 +1,54 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "64000": {
+      "content": "<mask>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "tokenizer_class": "PhobertTokenizer",
+  "unk_token": "<unk>"
+}