Hugging Face's logo

Spaces:
sartifyllc
/
Swahili-Text-Embeddings-Leaderboard
Sleeping

App Files Community
Mollel commited on
Commit
970cb5a
•
1 Parent(s): 35e4214

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +99 -1
README.md CHANGED
@@ -10,4 +10,102 @@ pinned: false
10
  license: apache-2.0
11
  ---
12
 
13
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  license: apache-2.0
11
  ---
12
 
13
+ # 🏆 Swahili Text Embeddings Leaderboard (STEL)
14
+
15
+ <p align="center">
16
+ <img src="files/STEL.jpg" alt="STEL" width="300" height="auto">
17
+ </p>
18
+
19
+
20
+ Welcome to the **Swahili Text Embeddings Leaderboard (STEL)**! This is a sartify community project aimed at creating a centralized leaderboard for Swahili text embeddings. The models listed here are evaluated using various Swahili text benchmarks. Contributions and corrections are always welcome! We define a model as "open" if it can be locally deployed and used commercially.
21
+
22
+
23
+ ## Table of Contents
24
+ - [Interactive Dashboard](#-interactive-dashboard)
25
+ - [Leaderboard](#-leaderboard)
26
+ - [Evaluation](#-evaluation)
27
+ - [How to Contribute](#-how-to-contribute)
28
+ - [Sponsorship](#-sponsorship)
29
+
30
+ ## 🌐 Interactive Dashboard
31
+
32
+ Explore our interactive dashboards:
33
+
34
+ - [Streamlit Dashboard](https://swahili-llm-leaderboard.streamlit.app/)
35
+ - [Hugging Face Space](https://huggingface.co/spaces/Mollel/swahili-llm-leaderboard)
36
+
37
+ ## 📊 Leaderboard
38
+ | Model Name | Publisher | Open? | Basemodel | Matryoshka | Dimension | Average | AfriSentiClassification | AfriSentiLangClassification | MasakhaNEWSClassification | MassiveIntentClassification | MassiveScenarioClassification | SwahiliNewsClassification | NTREXBitextMining | MasakhaNEWSClusteringP2P | MasakhaNEWSClusteringS2S | XNLI | MIRACLReranking | MIRACLRetrieval |
39
+ |:-------------------------------------------------------------------------------------------------------------------------------------------------|:---------------|:--------|:--------------------------|:-------------|:------------|----------:|--------------------------:|------------------------------:|----------------------------:|------------------------------:|--------------------------------:|----------------------------:|--------------------:|---------------------------:|---------------------------:|--------:|------------------:|------------------:|
40
+ | [MultiLinguSwahili-serengeti-E250-nli-matryoshka](https://huggingface.co/sartifyllc/MultiLinguSwahili-serengeti-E250-nli-matryoshka) | sartifyllc | Yes | serengeti-E250 | Yes | 768 | 51.3997 | 45.4011 | 77.6318 | 71.4496 | 56.4492 | 61.2038 | 63.9453 | 63.4926 | 38.6383 | 32.6575 | 77.0157 | 19.287 | 9.624 |
41
+ | [MultiLinguSwahili-bert-base-sw-cased-nli-matryoshka](https://huggingface.co/sartifyllc/MultiLinguSwahili-bert-base-sw-cased-nli-matryoshka) | sartifyllc | Yes | bert-base-sw-cased | Yes | 768 | 42.2575 | 35.4278 | 82.2461 | 64.8529 | 46.1332 | 50.5649 | 57.041 | 5.16086 | 37.6302 | 19.7916 | 68.5115 | 22.57 | 17.16 |
42
+ | [MultiLinguSwahili-mxbai-embed-large-v1-nli-matryoshka](https://huggingface.co/sartifyllc/MultiLinguSwahili-mxbai-embed-large-v1-nli-matryoshka) | sartifyllc | Yes | mxbai-embed-large-v1 | Yes | 768 | 40.0694 | 36.7914 | 81.9434 | 54.2647 | 46.7182 | 51.2206 | 57.2949 | 5.4534 | 34.5571 | 9.27166 | 70.311 | 21.831 | 11.176 |
43
+ | [mxbai-embed-large-v1](https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1) | mixedbread-ai | Yes | mxbai-embed-large-v1 | | ? | 39.6734 | 35.0802 | 83.4229 | 57.416 | 43.8635 | 47.1923 | 54.4678 | 5.87399 | 27.5669 | 21.1763 | 56.0497 | 23.742 | 20.229 |
44
+ | [bert-base-uncased-swahili](https://huggingface.co/flax-community/bert-base-uncased-swahili) | flax-community | Yes | bert-base-uncased-swahili | | ? | 37.8727 | 41.123 | 80.8838 | 66.0714 | 48.302 | 51.9334 | 64.2236 | 0.400601 | 18.6071 | 2.04101 | 58.9762 | 13.807 | 8.103 |
45
+ | [MultiLinguSwahili-bge-small-en-v1.5-nli-matryoshka](https://huggingface.co/sartifyllc/MultiLinguSwahili-serengeti-E250-nli-matryoshka) | sartifyllc | Yes | bge-small-en-v1.5 | Yes | 256 | 36.3029 | 35.107 | 67.3486 | 54.1597 | 38.0027 | 46.8393 | 51.2305 | 5.01061 | 21.7986 | 17.8461 | 62.3059 | 21.521 | 14.465 |
46
+ | [bert-base-sw-cased](https://huggingface.co/Geotrend/bert-base-sw-cased) | Geotrend | Yes | bert-base-sw-cased | | ? | 33.6552 | 35.3342 | 84.3066 | 62.3109 | 36.3685 | 38.7996 | 57.9199 | 0.396624 | 12.9566 | 6.77267 | 55.6602 | 10.077 | 2.959 |
47
+ | [UBC-NLPserengeti-E250](https://huggingface.co/UBC-NLP/serengeti-E250) | UCB | Yes | UBC-NLPserengeti-E250 | | ? | 33.581 | 44.0374 | 84.3848 | 42.1008 | 37.1957 | 38.2414 | 58.1592 | 12.7676 | 15.7357 | 14.7948 | 53.3967 | 2.041 | 0.117 |
48
+ ## 🧪 Evaluation
49
+ To evaluate a model on the Swahili Embeddings Text Benchmark, you can use the following Python script:
50
+ ```python
51
+ pip install mteb
52
+ pip install sentence-transformers
53
+ import mteb
54
+ from sentence_transformers import SentenceTransformer
55
+
56
+ models = ["sartifyllc/MultiLinguSwahili-bert-base-sw-cased-nli-matryoshka"]
57
+
58
+
59
+ for model_name in models:
60
+ truncate_dim = 768
61
+ language = "swa"
62
+
63
+ device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu") # if cuda available
64
+ # model = SentenceTransformer(model_name, truncate_dim = truncate_dim, device = device, trust_remote_code=True) # if you want to use matryoshka n dimension
65
+ model = SentenceTransformer(model_name, device = device, trust_remote_code=True)
66
+
67
+ tasks = [
68
+ mteb.get_task("AfriSentiClassification", languages = ["swa"]),
69
+ mteb.get_task("AfriSentiLangClassification", languages = ["swa"]),
70
+ # "LanguageClassification": "accuracy",
71
+ mteb.get_task("MasakhaNEWSClassification", languages = ["swa"]),
72
+ mteb.get_task("MassiveIntentClassification", languages = ["swa"]),
73
+ mteb.get_task("MassiveScenarioClassification", languages = ["swa"]),
74
+ mteb.get_task("SwahiliNewsClassification", languages = ["swa"]),
75
+ # mteb.get_tasks(task_types=["PairClassification", "Reranking", "BitextMining", "Clustering", "Retrieval"], languages = ["swa"]),
76
+ ]
77
+
78
+
79
+ evaluation = mteb.MTEB(tasks=tasks)
80
+ results = evaluation.run(model, output_folder=f"{model_name}")
81
+
82
+ # results = evaluation.run(model, output_folder=f"{model_name}")
83
+ tasks = mteb.get_tasks(task_types=["PairClassification", "Reranking", "BitextMining", "Clustering", "Retrieval"], languages = ["swa"])
84
+
85
+ evaluation = mteb.MTEB(tasks=tasks)
86
+ results = evaluation.run(model, output_folder=f"{model_name}")
87
+ ```
88
+
89
+
90
+ ## 🤝 How to Contribute
91
+
92
+ We welcome and appreciate all contributions! You can help by:
93
+
94
+ ### Table Work
95
+
96
+ - Filling in missing entries.
97
+ - New models are added as new rows to the leaderboard (maintaining descending order).
98
+ - Add new benchmarks as new columns in the leaderboard and include them in the benchmarks table (maintaining descending order).
99
+
100
+ ### Code Work
101
+
102
+ - Improving the existing code.
103
+ - Requesting and implementing new features.
104
+
105
+ ## 🤝 Sponsorship
106
+
107
+ This benchmark is Swahili-based, and we need support translating and curating more tasks into Swahili. Sponsorships are welcome to help advance this endeavour. Your sponsorship will facilitate essential translation efforts, bridge language barriers, and make the benchmark accessible to a broader audience. We are grateful for the dedication shown by our collaborators and aim to extend this impact further with the support of sponsors committed to advancing language technologies.
108
+
109
+ ---
110
+
111
+ Thank you for being part of this effort to advance Swahili language technologies!