Spaces:
Running
Running
Renamed accordion to "Edit Judge Prompt" & added messaging around "Turbo" models
Browse files
app.py
CHANGED
|
@@ -386,7 +386,7 @@ with gr.Blocks(theme="default", css=CSS_STYLES) as demo:
|
|
| 386 |
gr.Markdown("<br>")
|
| 387 |
|
| 388 |
# Update Evaluator Prompt Accordion
|
| 389 |
-
with gr.Accordion("📝 Judge Prompt", open=False):
|
| 390 |
eval_prompt_editable = gr.TextArea(
|
| 391 |
value=DEFAULT_EVAL_PROMPT_EDITABLE,
|
| 392 |
label="Evaluation Criteria",
|
|
@@ -413,6 +413,13 @@ with gr.Blocks(theme="default", css=CSS_STYLES) as demo:
|
|
| 413 |
datatype=["str", "number", "str", "number", "str", "str", "str"],
|
| 414 |
)
|
| 415 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 416 |
# Add change handler for checkbox
|
| 417 |
show_preliminary.change(
|
| 418 |
fn=refresh_leaderboard,
|
|
|
|
| 386 |
gr.Markdown("<br>")
|
| 387 |
|
| 388 |
# Update Evaluator Prompt Accordion
|
| 389 |
+
with gr.Accordion("📝 Edit Judge Prompt", open=False):
|
| 390 |
eval_prompt_editable = gr.TextArea(
|
| 391 |
value=DEFAULT_EVAL_PROMPT_EDITABLE,
|
| 392 |
label="Evaluation Criteria",
|
|
|
|
| 413 |
datatype=["str", "number", "str", "number", "str", "str", "str"],
|
| 414 |
)
|
| 415 |
|
| 416 |
+
gr.Markdown("""<br>
|
| 417 |
+
<br>
|
| 418 |
+
Judge Arena uses Together AI for inference of open-source models. FP8 models are named as -- "Turbo" where the performance of the FP16 reference models is closely matched:
|
| 419 |
+
|
| 420 |
+
[*"Together Turbo achieves this performance while maintaining full accuracy compared to Meta's reference implementation across all models. Llama-3.1-405B-Instruct-Turbo matches the accuracy of Meta reference models."*](https://www.together.ai/blog/together-inference-engine-2)
|
| 421 |
+
""")
|
| 422 |
+
|
| 423 |
# Add change handler for checkbox
|
| 424 |
show_preliminary.change(
|
| 425 |
fn=refresh_leaderboard,
|