Commit
Β·
8cb674a
1
Parent(s):
18566fa
added more models
Browse files
app.py
CHANGED
|
@@ -2,10 +2,10 @@ import gradio as gr
|
|
| 2 |
import pandas as pd
|
| 3 |
from css_html_js import custom_css
|
| 4 |
|
| 5 |
-
TITLE = """<h1 align="center" id="space-title">π²πΎ Malaysian Embedding Leaderboard</h1>"""
|
| 6 |
|
| 7 |
INTRODUCTION_TEXT = """
|
| 8 |
-
π The π²πΎ Malaysian Embedding Leaderboard aims to track, rank and evaluate Top-k retrieval using embedding models. All notebooks at https://github.com/mesolitica/embedding-benchmarks, feel free to submit your own score at https://huggingface.co/spaces/mesolitica/Malaysian-Embedding-Leaderboard/discussions with link to the notebook.
|
| 9 |
|
| 10 |
## Dataset
|
| 11 |
|
|
@@ -52,27 +52,49 @@ open_source = [
|
|
| 52 |
'lom.agc.gov.my top-5': 0.1717699775952203,
|
| 53 |
'lom.agc.gov.my top-10': 0.23089370176748816,
|
| 54 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
{
|
| 56 |
'model': '[llama2-embedding-600m-8k-contrastive](https://huggingface.co/mesolitica/llama2-embedding-600m-8k)',
|
| 57 |
-
'Crossref Melayu top-1': 0.
|
| 58 |
-
'Crossref Melayu top-3': 0.
|
| 59 |
-
'Crossref Melayu top-5': 0.
|
| 60 |
-
'Crossref Melayu top-10': 0.
|
| 61 |
-
'lom.agc.gov.my top-1': 0.
|
| 62 |
-
'lom.agc.gov.my top-3': 0.
|
| 63 |
-
'lom.agc.gov.my top-5': 0.
|
| 64 |
-
'lom.agc.gov.my top-10': 0.
|
| 65 |
},
|
| 66 |
{
|
| 67 |
'model': '[llama2-embedding-1b-8k-contrastive](https://huggingface.co/mesolitica/llama2-embedding-1b-8k)',
|
| 68 |
-
'Crossref Melayu top-1': 0.
|
| 69 |
-
'Crossref Melayu top-3': 0.
|
| 70 |
-
'Crossref Melayu top-5': 0.
|
| 71 |
-
'Crossref Melayu top-10': 0.
|
| 72 |
-
'lom.agc.gov.my top-1': 0.
|
| 73 |
-
'lom.agc.gov.my top-3': 0.
|
| 74 |
-
'lom.agc.gov.my top-5': 0.
|
| 75 |
-
'lom.agc.gov.my top-10': 0.
|
| 76 |
},
|
| 77 |
]
|
| 78 |
|
|
|
|
| 2 |
import pandas as pd
|
| 3 |
from css_html_js import custom_css
|
| 4 |
|
| 5 |
+
TITLE = """<h1 align="center" id="space-title">π²πΎ Malaysian RAG Embedding Leaderboard</h1>"""
|
| 6 |
|
| 7 |
INTRODUCTION_TEXT = """
|
| 8 |
+
π The π²πΎ Malaysian RAG Embedding Leaderboard aims to track, rank and evaluate Top-k retrieval using embedding models. All notebooks at https://github.com/mesolitica/embedding-benchmarks, feel free to submit your own score at https://huggingface.co/spaces/mesolitica/Malaysian-Embedding-Leaderboard/discussions with link to the notebook.
|
| 9 |
|
| 10 |
## Dataset
|
| 11 |
|
|
|
|
| 52 |
'lom.agc.gov.my top-5': 0.1717699775952203,
|
| 53 |
'lom.agc.gov.my top-10': 0.23089370176748816,
|
| 54 |
},
|
| 55 |
+
{
|
| 56 |
+
'model': '[mistral-embedding-191m-8k-contrastive](https://huggingface.co/mesolitica/mistral-embedding-191m-8k-contrastive)',
|
| 57 |
+
'Crossref Melayu top-1': 0.08001654088700506,
|
| 58 |
+
'Crossref Melayu top-3': 0.17378269409697095,
|
| 59 |
+
'Crossref Melayu top-5': 0.232192701333609,
|
| 60 |
+
'Crossref Melayu top-10': 0.32482166856197664,
|
| 61 |
+
'lom.agc.gov.my top-1': 0.041075429424943986,
|
| 62 |
+
'lom.agc.gov.my top-3': 0.09148618371919343,
|
| 63 |
+
'lom.agc.gov.my top-5': 0.12758277321384118,
|
| 64 |
+
'lom.agc.gov.my top-10': 0.18707991038088126,
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
'model': '[mistral-embedding-349m-8k-contrastive](https://huggingface.co/mesolitica/mistral-embedding-349m-8k-contrastive)',
|
| 68 |
+
'Crossref Melayu top-1': 0.09045797580895276,
|
| 69 |
+
'Crossref Melayu top-3': 0.18742892587615012,
|
| 70 |
+
'Crossref Melayu top-5': 0.2444949860436266,
|
| 71 |
+
'Crossref Melayu top-10': 0.3398118474103174,
|
| 72 |
+
'lom.agc.gov.my top-1': 0.039581777445855115,
|
| 73 |
+
'lom.agc.gov.my top-3': 0.08849887976101568,
|
| 74 |
+
'lom.agc.gov.my top-5': 0.12335075927308937,
|
| 75 |
+
'lom.agc.gov.my top-10': 0.18558625840179238,
|
| 76 |
+
},
|
| 77 |
{
|
| 78 |
'model': '[llama2-embedding-600m-8k-contrastive](https://huggingface.co/mesolitica/llama2-embedding-600m-8k)',
|
| 79 |
+
'Crossref Melayu top-1': 0.11516592577276956,
|
| 80 |
+
'Crossref Melayu top-3': 0.2412901891863951,
|
| 81 |
+
'Crossref Melayu top-5': 0.3201695440918019,
|
| 82 |
+
'Crossref Melayu top-10': 0.4337847617078466,
|
| 83 |
+
'lom.agc.gov.my top-1': 0.05738113019666418,
|
| 84 |
+
'lom.agc.gov.my top-3': 0.12571570823998007,
|
| 85 |
+
'lom.agc.gov.my top-5': 0.16455065969629076,
|
| 86 |
+
'lom.agc.gov.my top-10': 0.24446104057754542,
|
| 87 |
},
|
| 88 |
{
|
| 89 |
'model': '[llama2-embedding-1b-8k-contrastive](https://huggingface.co/mesolitica/llama2-embedding-1b-8k)',
|
| 90 |
+
'Crossref Melayu top-1': 0.16489196733174816,
|
| 91 |
+
'Crossref Melayu top-3': 0.3281298459629898,
|
| 92 |
+
'Crossref Melayu top-5': 0.42623798201178537,
|
| 93 |
+
'Crossref Melayu top-10': 0.5487439263930528,
|
| 94 |
+
'lom.agc.gov.my top-1': 0.09123724172267862,
|
| 95 |
+
'lom.agc.gov.my top-3': 0.18309683843664426,
|
| 96 |
+
'lom.agc.gov.my top-5': 0.24271844660194175,
|
| 97 |
+
'lom.agc.gov.my top-10': 0.3354493403037092,
|
| 98 |
},
|
| 99 |
]
|
| 100 |
|