test metrics for best models

app_implementation.py

@@ -26,7 +26,7 @@ class RetrievalApp:
 
     def get_device_options(self):
         if self.is_cuda_available():
-            return ["cpu", "cuda"]
+            return ["cuda", "cpu"]
         else:
             return ["cpu"]
 

assets/nbow_dependencies-nbow-nbow-mnrl

@@ -0,0 +1 @@
+{"accuracy@k": {"1": 0.3388, "3": 0.4821, "5": 0.5212, "10": 0.5896}, "precision@k": {"1": 0.3388, "3": 0.2888, "5": 0.2502, "10": 0.2088}, "recall@k": {"1": 0.0481, "3": 0.0953, "5": 0.1204, "10": 0.1725}, "ndcg@k": {"10": 0.2811}, "mrr@k": {"10": 0.419}, "map@k": {"10": 0.2044}, "columns": "dependencies", "model_name": "nbow_dependencies-nbow-nbow-mnrl"}

assets/nbow_readme-nbow-nbow-mnrl

@@ -0,0 +1 @@
+{"accuracy@k": {"1": 0.3225, "3": 0.4821, "5": 0.5537, "10": 0.6189}, "precision@k": {"1": 0.3225, "3": 0.2899, "5": 0.2495, "10": 0.2007}, "recall@k": {"1": 0.0451, "3": 0.0982, "5": 0.1281, "10": 0.1706}, "ndcg@k": {"10": 0.2748}, "mrr@k": {"10": 0.4175}, "map@k": {"10": 0.1938}, "columns": "readme", "model_name": "nbow_readme-nbow-nbow-mnrl"}

assets/nbow_readme_dependencies-nbow-nbow-mnrl

@@ -0,0 +1 @@
+{"accuracy@k": {"1": 0.4235, "3": 0.5375, "5": 0.5896, "10": 0.6482}, "precision@k": {"1": 0.4235, "3": 0.342, "5": 0.3199, "10": 0.2557}, "recall@k": {"1": 0.063, "3": 0.1208, "5": 0.1642, "10": 0.2273}, "ndcg@k": {"10": 0.3508}, "mrr@k": {"10": 0.4955}, "map@k": {"10": 0.2688}, "columns": "readme_dependencies", "model_name": "nbow_readme_dependencies-nbow-nbow-mnrl"}

assets/nbow_titles-nbow-nbow-mnrl

@@ -0,0 +1 @@
+{"accuracy@k": {"1": 0.6645, "3": 0.7622, "5": 0.8046, "10": 0.8306}, "precision@k": {"1": 0.6645, "3": 0.5765, "5": 0.5121, "10": 0.4075}, "recall@k": {"1": 0.1201, "3": 0.2775, "5": 0.3462, "10": 0.4477}, "ndcg@k": {"10": 0.6024}, "mrr@k": {"10": 0.7198}, "map@k": {"10": 0.5262}, "columns": "titles", "model_name": "nbow_titles-nbow-nbow-mnrl"}

assets/nbow_titles_dependencies-nbow-nbow-mnrl

@@ -0,0 +1 @@
+{"accuracy@k": {"1": 0.4169, "3": 0.5179, "5": 0.57, "10": 0.6352}, "precision@k": {"1": 0.4169, "3": 0.3333, "5": 0.3003, "10": 0.241}, "recall@k": {"1": 0.0537, "3": 0.107, "5": 0.1518, "10": 0.2045}, "ndcg@k": {"10": 0.3294}, "mrr@k": {"10": 0.4847}, "map@k": {"10": 0.2493}, "columns": "titles_dependencies", "model_name": "nbow_titles_dependencies-nbow-nbow-mnrl"}

pages/2_Statistics.py

@@ -1,6 +1,30 @@
 import pandas as pd
 import streamlit as st
 import config
+from pathlib import Path as P
+import json
+
+
+nbow_results_path = P("assets").glob("nbow*")
+
+def display_metrics_dict(metrics, display_only_accuracy):
+    model_name = metrics.pop("model_name")
+    columns = metrics.pop("columns").split("_")
+    st.markdown(f"### columns: {columns}")
+    st.markdown(f"best model {model_name}")
+    if not display_only_accuracy:
+        st.json(metrics)
+    else:
+        st.json({"accuracy@10": metrics["accuracy@k"]["10"]})
+
+def display_metrics():
+    display_only_accuracy = st.sidebar.checkbox("display only accuracy@10", value=True)
+    st.markdown("## Test metrics for best validation modelon given columns")
+    for p in nbow_results_path:
+        metrics = json.loads(open(p, "r").read())
+        display_metrics_dict(metrics, display_only_accuracy)
+
+display_metrics()
 
 best_results_df = pd.read_csv(config.best_tasks_path)