added initial scores

.gitignore

@@ -0,0 +1 @@
+*__pycache__

README.md

@@ -9,4 +9,4 @@ app_file: app.py
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Notebooks at https://github.com/mesolitica/llm-benchmarks

app.py

@@ -0,0 +1,51 @@
+import gradio as gr
+import pandas as pd
+from css_html_js import custom_css
+
+demo = gr.Blocks(css=custom_css)
+
+TITLE = """<h1 align="center" id="space-title">🤗 Malay LLM Leaderboard</h1>"""
+
+INTRODUCTION_TEXT = """
+📐 The 🤗 Malay LLM Leaderboard aims to track, rank and evaluate open LLMs on Malay tasks.\n
+🤗 All notebooks at https://github.com/mesolitica/llm-benchmarks, feel free to submit your own score at https://huggingface.co/spaces/mesolitica/malay-llm-leaderboard/discussions with link to the notebook.
+
+## Dataset
+
+📈 We evaluate models based on 4 datasets,
+
+1. BM-PT3 Paper 1, contains 54 questions, https://github.com/mesolitica/malaysian-dataset/tree/master/llm-benchmark/BM-pt3
+2. BM Paper 1, contains 180 questions, https://github.com/mesolitica/malaysian-dataset/tree/master/llm-benchmark/tatabahasabm.tripod.com-bm-kertas-1
+3. Tatabahasa, contains 349 questions, https://github.com/mesolitica/malaysian-dataset/tree/master/llm-benchmark/tatabahasabm.tripod.com
+4. Translated IndoNLI to Malay, tested on `test_expert` dataset, https://huggingface.co/datasets/mesolitica/translated-indonli
+"""
+
+data = [
+    {
+        'model': 'gpt-3.5-turbo-0613',
+        'BM-PT3 0-shot (% correct)': 36.53846153846153,
+        'BM-PT3 1-shot (% correct)': 28.846153846153843,
+        'BM-PT3 3-shots (% correct)': 24.528301886792452,
+    },
+    {
+        'model': 'malaysian-llama2-7b-32k',
+        'BM-PT3 0-shot (% correct)': 20.37037037037037,
+        'BM-PT3 1-shot (% correct)': 16.666666666666664,
+        'BM-PT3 3-shots (% correct)': 27.77777777777778,
+    },
+    {
+        'model': 'malaysian-llama2-13b-32k',
+        'BM-PT3 0-shot (% correct)': 33.33333333333333,
+        'BM-PT3 1-shot (% correct)': 24.074074074074073,
+        'BM-PT3 3-shots (% correct)': 25.925925925925924,
+    }
+]
+
+data = pd.DataFrame(data)
+
+with demo:
+    gr.HTML(TITLE)
+    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
+    gr.DataFrame(data)
+
+demo.launch()

css_html_js.py

@@ -0,0 +1,111 @@
+custom_css = """
+
+.markdown-text {
+    font-size: 16px !important;
+}
+
+#models-to-add-text {
+    font-size: 18px !important;
+}
+
+#citation-button span {
+    font-size: 16px !important;
+}
+
+#citation-button textarea {
+    font-size: 16px !important;
+}
+
+#citation-button > label > button {
+    margin: 6px;
+    transform: scale(1.3);
+}
+
+#leaderboard-table {
+    margin-top: 15px
+}
+
+#leaderboard-table-lite {
+    margin-top: 15px
+}
+
+#search-bar-table-box > div:first-child {
+    background: none;
+    border: none;
+}
+ 
+#search-bar {
+    padding: 0px;
+}
+
+/* Hides the final AutoEvalColumn */
+#llm-benchmark-tab-table table td:last-child,
+#llm-benchmark-tab-table table th:last-child {
+    display: none;
+}
+
+/* Limit the width of the first AutoEvalColumn so that names don't expand too much */
+table td:first-child,
+table th:first-child {
+    max-width: 400px;
+    overflow: auto;
+    white-space: nowrap;
+}
+
+.tab-buttons button {
+    font-size: 20px;
+}
+
+#scale-logo {
+    border-style: none !important;
+    box-shadow: none;
+    display: block;
+    margin-left: auto;
+    margin-right: auto;
+    max-width: 600px;
+}
+
+#scale-logo .download {
+    display: none;
+}
+#filter_type{
+    border: 0;
+    padding-left: 0;
+    padding-top: 0;
+}
+#filter_type label {
+    display: flex;
+}
+#filter_type label > span{
+    margin-top: var(--spacing-lg);
+    margin-right: 0.5em;
+}
+#filter_type label > .wrap{
+    width: 103px;
+}
+#filter_type label > .wrap .wrap-inner{  
+    padding: 2px;
+}
+#filter_type label > .wrap .wrap-inner input{
+    width: 1px
+}
+#filter-columns-type{
+    border:0;
+    padding:0.5;
+}
+#filter-columns-size{
+    border:0;
+    padding:0.5;
+}
+#box-filter > .form{
+    border: 0
+}
+"""
+
+get_window_url_params = """
+    function(url_params) {
+        const params = new URLSearchParams(window.location.search);
+        url_params = Object.fromEntries(params);
+        return url_params;
+    }
+    """

requirements.txt

@@ -0,0 +1 @@
+pandas