Spaces:
Running
Running
Added Atla Selene / Selene Mini API
Browse files
app.py
CHANGED
|
@@ -13,7 +13,7 @@ from gen_api_answer import (
|
|
| 13 |
get_model_response,
|
| 14 |
parse_model_response,
|
| 15 |
prometheus_parse_model_response,
|
| 16 |
-
|
| 17 |
flow_judge_parse_model_response
|
| 18 |
)
|
| 19 |
|
|
@@ -706,7 +706,7 @@ with gr.Blocks(theme="default", css=CSS_STYLES) as demo:
|
|
| 706 |
if info.get("active", True) is True] # Explicitly check for True
|
| 707 |
|
| 708 |
# Define new models list
|
| 709 |
-
new_models = ["Atla Selene 1
|
| 710 |
|
| 711 |
# New models appear 40% of the time
|
| 712 |
if random.random() < 0.4:
|
|
@@ -742,8 +742,6 @@ with gr.Blocks(theme="default", css=CSS_STYLES) as demo:
|
|
| 742 |
|
| 743 |
is_prometheus_a = model_data.get(model_a, {}).get('organization') == 'Prometheus'
|
| 744 |
is_prometheus_b = model_data.get(model_b, {}).get('organization') == 'Prometheus'
|
| 745 |
-
is_atla_a = model_data.get(model_a, {}).get('organization') == 'Atla'
|
| 746 |
-
is_atla_b = model_data.get(model_b, {}).get('organization') == 'Atla'
|
| 747 |
is_flow_judge_a = model_data.get(model_a, {}).get('organization') == 'Flow AI'
|
| 748 |
is_flow_judge_b = model_data.get(model_b, {}).get('organization') == 'Flow AI'
|
| 749 |
is_salesforce_a = model_data.get(model_a, {}).get('organization') == 'Salesforce'
|
|
@@ -753,8 +751,8 @@ with gr.Blocks(theme="default", css=CSS_STYLES) as demo:
|
|
| 753 |
if is_prometheus_a:
|
| 754 |
score_a_val, critique_a_val = prometheus_parse_model_response(response_a)
|
| 755 |
score_a_val = f"{score_a_val} / 5"
|
| 756 |
-
elif
|
| 757 |
-
score_a_val, critique_a_val =
|
| 758 |
score_a_val = f"{score_a_val} / 5"
|
| 759 |
elif is_flow_judge_a:
|
| 760 |
score_a_val, critique_a_val = flow_judge_parse_model_response(response_a)
|
|
@@ -766,8 +764,8 @@ with gr.Blocks(theme="default", css=CSS_STYLES) as demo:
|
|
| 766 |
if is_prometheus_b:
|
| 767 |
score_b_val, critique_b_val = prometheus_parse_model_response(response_b)
|
| 768 |
score_b_val = f"{score_b_val} / 5"
|
| 769 |
-
elif
|
| 770 |
-
score_b_val, critique_b_val =
|
| 771 |
score_b_val = f"{score_b_val} / 5"
|
| 772 |
elif is_flow_judge_b:
|
| 773 |
score_b_val, critique_b_val = flow_judge_parse_model_response(response_b)
|
|
|
|
| 13 |
get_model_response,
|
| 14 |
parse_model_response,
|
| 15 |
prometheus_parse_model_response,
|
| 16 |
+
salesforce_parse_model_response,
|
| 17 |
flow_judge_parse_model_response
|
| 18 |
)
|
| 19 |
|
|
|
|
| 706 |
if info.get("active", True) is True] # Explicitly check for True
|
| 707 |
|
| 708 |
# Define new models list
|
| 709 |
+
new_models = ["Atla Selene 1", "SFR-LLaMA-3.1-70B-Judge"]
|
| 710 |
|
| 711 |
# New models appear 40% of the time
|
| 712 |
if random.random() < 0.4:
|
|
|
|
| 742 |
|
| 743 |
is_prometheus_a = model_data.get(model_a, {}).get('organization') == 'Prometheus'
|
| 744 |
is_prometheus_b = model_data.get(model_b, {}).get('organization') == 'Prometheus'
|
|
|
|
|
|
|
| 745 |
is_flow_judge_a = model_data.get(model_a, {}).get('organization') == 'Flow AI'
|
| 746 |
is_flow_judge_b = model_data.get(model_b, {}).get('organization') == 'Flow AI'
|
| 747 |
is_salesforce_a = model_data.get(model_a, {}).get('organization') == 'Salesforce'
|
|
|
|
| 751 |
if is_prometheus_a:
|
| 752 |
score_a_val, critique_a_val = prometheus_parse_model_response(response_a)
|
| 753 |
score_a_val = f"{score_a_val} / 5"
|
| 754 |
+
elif is_salesforce_a: # Same parser for Atla and Salesforce
|
| 755 |
+
score_a_val, critique_a_val = salesforce_parse_model_response(response_a)
|
| 756 |
score_a_val = f"{score_a_val} / 5"
|
| 757 |
elif is_flow_judge_a:
|
| 758 |
score_a_val, critique_a_val = flow_judge_parse_model_response(response_a)
|
|
|
|
| 764 |
if is_prometheus_b:
|
| 765 |
score_b_val, critique_b_val = prometheus_parse_model_response(response_b)
|
| 766 |
score_b_val = f"{score_b_val} / 5"
|
| 767 |
+
elif is_salesforce_b: # Same parser for Atla and Salesforce
|
| 768 |
+
score_b_val, critique_b_val = salesforce_parse_model_response(response_b)
|
| 769 |
score_b_val = f"{score_b_val} / 5"
|
| 770 |
elif is_flow_judge_b:
|
| 771 |
score_b_val, critique_b_val = flow_judge_parse_model_response(response_b)
|