Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -48,6 +48,7 @@ def submit_model(name):
|
|
| 48 |
|
| 49 |
with gr.Blocks(css=".tab-buttons button{font-size:1.3em}") as demo:
|
| 50 |
gr.HTML('<h1 style="text-align:center"><span style="font-size:1.3em">Subquadratic LLM Leaderboard</span></h1>')
|
|
|
|
| 51 |
|
| 52 |
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
| 53 |
with gr.Tab("🏅 LLM Benchmark"):
|
|
@@ -75,12 +76,11 @@ with gr.Blocks(css=".tab-buttons button{font-size:1.3em}") as demo:
|
|
| 75 |
|
| 76 |
with gr.Tab("📝 About"):
|
| 77 |
gr.Markdown("""
|
| 78 |
-
The **Subquadratic LLM Leaderboard** evaluates LLMs with subquadratic architectures (ie RWKV & Mamba) with the goal of providing open
|
| 79 |
-
while the architectures themselves are pending inclusion in 🤗 Transformers.
|
| 80 |
-
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
This leaderboard is maintained by Devin Gulliver and is still under construction, check back regularly for further improvements!
|
| 84 |
""")
|
| 85 |
|
| 86 |
with gr.Tab("🚀 Submit here!"):
|
|
|
|
| 48 |
|
| 49 |
with gr.Blocks(css=".tab-buttons button{font-size:1.3em}") as demo:
|
| 50 |
gr.HTML('<h1 style="text-align:center"><span style="font-size:1.3em">Subquadratic LLM Leaderboard</span></h1>')
|
| 51 |
+
gr.Markdown("**REMEMBER:** If you don't see an eligible model here, make sure to submit it! We hope to incentivize subquadratic/attention-free LLM development through friendly competition.")
|
| 52 |
|
| 53 |
with gr.Tabs(elem_classes="tab-buttons") as tabs:
|
| 54 |
with gr.Tab("🏅 LLM Benchmark"):
|
|
|
|
| 76 |
|
| 77 |
with gr.Tab("📝 About"):
|
| 78 |
gr.Markdown("""
|
| 79 |
+
The **Subquadratic LLM Leaderboard** evaluates LLMs with subquadratic/attention-free architectures (ie RWKV & Mamba) with the goal of providing open
|
| 80 |
+
evaluation results while the architectures themselves are pending inclusion/release in the 🤗 Transformers library.
|
| 81 |
+
The metrics are the same as the Open LLM Leaderboard: ARC 25-shot, HellaSwag 10-shot, MMLU 5-shot, TruthfulQA zeroshot, Winogrande 5-shot, and GSM8K 5-shot.
|
| 82 |
+
This leaderboard is maintained by Devin Gulliver and is perpetually under construction, check back regularly for further improvements!
|
| 83 |
+
Compute for evaluating RWKV-5 models is generously provided by Recursal AI.
|
|
|
|
| 84 |
""")
|
| 85 |
|
| 86 |
with gr.Tab("🚀 Submit here!"):
|