Spaces:
Sleeping
Sleeping
fix: use importlib to load api_client directly from file
Browse files- app/pages/compare.py +20 -49
- app/pages/monitoring.py +12 -13
- app/pages/predict.py +8 -3
app/pages/compare.py
CHANGED
|
@@ -1,55 +1,28 @@
|
|
| 1 |
-
import
|
| 2 |
from pathlib import Path
|
| 3 |
|
| 4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
|
| 6 |
import pandas as pd
|
| 7 |
import streamlit as st
|
| 8 |
|
| 9 |
-
from app.api_client import api_predict, require_api
|
| 10 |
-
|
| 11 |
require_api()
|
| 12 |
|
| 13 |
st.title("Compare Models")
|
| 14 |
st.caption("Side-by-side comparison of all trained models on CLINC150 test set.")
|
| 15 |
|
| 16 |
RESULTS = {
|
| 17 |
-
"logreg": {
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
},
|
| 23 |
-
"svm": {
|
| 24 |
-
"accuracy": 0.8002,
|
| 25 |
-
"macro_f1": 0.8474,
|
| 26 |
-
"weighted_f1": 0.7795,
|
| 27 |
-
"latency_p50_ms": 0.14,
|
| 28 |
-
},
|
| 29 |
-
"textcnn": {
|
| 30 |
-
"accuracy": 0.7740,
|
| 31 |
-
"macro_f1": 0.8244,
|
| 32 |
-
"weighted_f1": 0.7646,
|
| 33 |
-
"latency_p50_ms": 97.6,
|
| 34 |
-
},
|
| 35 |
-
"rnn": {
|
| 36 |
-
"accuracy": 0.1818,
|
| 37 |
-
"macro_f1": 0.0020,
|
| 38 |
-
"weighted_f1": 0.0560,
|
| 39 |
-
"latency_p50_ms": 124.5,
|
| 40 |
-
},
|
| 41 |
-
"lstm": {
|
| 42 |
-
"accuracy": 0.7669,
|
| 43 |
-
"macro_f1": 0.8113,
|
| 44 |
-
"weighted_f1": 0.7555,
|
| 45 |
-
"latency_p50_ms": 249.2,
|
| 46 |
-
},
|
| 47 |
-
"distilbert": {
|
| 48 |
-
"accuracy": 0.8765,
|
| 49 |
-
"macro_f1": 0.9019,
|
| 50 |
-
"weighted_f1": 0.8722,
|
| 51 |
-
"latency_p50_ms": 9.6,
|
| 52 |
-
},
|
| 53 |
}
|
| 54 |
|
| 55 |
df = pd.DataFrame(RESULTS).T
|
|
@@ -57,14 +30,12 @@ df.index.name = "model"
|
|
| 57 |
|
| 58 |
st.subheader("Metrics Table")
|
| 59 |
st.dataframe(
|
| 60 |
-
df.style.format(
|
| 61 |
-
{
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
}
|
| 67 |
-
).background_gradient(subset=["accuracy", "macro_f1", "weighted_f1"], cmap="Greens"),
|
| 68 |
use_container_width=True,
|
| 69 |
)
|
| 70 |
|
|
@@ -97,4 +68,4 @@ if st.button("Compare", type="primary", disabled=not (text and selected_models))
|
|
| 97 |
st.metric("Confidence", f"{result['confidence']:.2%}")
|
| 98 |
st.metric("Latency", f"{result['latency_ms']:.1f} ms")
|
| 99 |
if result["is_oos"]:
|
| 100 |
-
st.warning("flagged OOS")
|
|
|
|
| 1 |
+
import importlib.util
|
| 2 |
from pathlib import Path
|
| 3 |
|
| 4 |
+
_client_path = Path(__file__).parent.parent / "api_client.py"
|
| 5 |
+
_spec = importlib.util.spec_from_file_location("api_client", _client_path)
|
| 6 |
+
_module = importlib.util.module_from_spec(_spec)
|
| 7 |
+
_spec.loader.exec_module(_module)
|
| 8 |
+
api_predict = _module.api_predict
|
| 9 |
+
require_api = _module.require_api
|
| 10 |
|
| 11 |
import pandas as pd
|
| 12 |
import streamlit as st
|
| 13 |
|
|
|
|
|
|
|
| 14 |
require_api()
|
| 15 |
|
| 16 |
st.title("Compare Models")
|
| 17 |
st.caption("Side-by-side comparison of all trained models on CLINC150 test set.")
|
| 18 |
|
| 19 |
RESULTS = {
|
| 20 |
+
"logreg": {"accuracy": 0.8058, "macro_f1": 0.8416, "weighted_f1": 0.7985, "latency_p50_ms": 0.12},
|
| 21 |
+
"svm": {"accuracy": 0.8002, "macro_f1": 0.8474, "weighted_f1": 0.7795, "latency_p50_ms": 0.14},
|
| 22 |
+
"textcnn": {"accuracy": 0.7740, "macro_f1": 0.8244, "weighted_f1": 0.7646, "latency_p50_ms": 97.6},
|
| 23 |
+
"rnn": {"accuracy": 0.1818, "macro_f1": 0.0020, "weighted_f1": 0.0560, "latency_p50_ms": 124.5},
|
| 24 |
+
"lstm": {"accuracy": 0.7669, "macro_f1": 0.8113, "weighted_f1": 0.7555, "latency_p50_ms": 249.2},
|
| 25 |
+
"distilbert": {"accuracy": 0.8765, "macro_f1": 0.9019, "weighted_f1": 0.8722, "latency_p50_ms": 9.6},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
}
|
| 27 |
|
| 28 |
df = pd.DataFrame(RESULTS).T
|
|
|
|
| 30 |
|
| 31 |
st.subheader("Metrics Table")
|
| 32 |
st.dataframe(
|
| 33 |
+
df.style.format({
|
| 34 |
+
"accuracy": "{:.2%}",
|
| 35 |
+
"macro_f1": "{:.2%}",
|
| 36 |
+
"weighted_f1": "{:.2%}",
|
| 37 |
+
"latency_p50_ms": "{:.2f} ms",
|
| 38 |
+
}).background_gradient(subset=["accuracy", "macro_f1", "weighted_f1"], cmap="Greens"),
|
|
|
|
|
|
|
| 39 |
use_container_width=True,
|
| 40 |
)
|
| 41 |
|
|
|
|
| 68 |
st.metric("Confidence", f"{result['confidence']:.2%}")
|
| 69 |
st.metric("Latency", f"{result['latency_ms']:.1f} ms")
|
| 70 |
if result["is_oos"]:
|
| 71 |
+
st.warning("flagged OOS")
|
app/pages/monitoring.py
CHANGED
|
@@ -1,12 +1,17 @@
|
|
| 1 |
-
import
|
| 2 |
from pathlib import Path
|
| 3 |
|
| 4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
|
| 6 |
import streamlit as st
|
| 7 |
|
| 8 |
-
from app.api_client import api_ab_stats, api_drift, api_reset_monitoring, require_api
|
| 9 |
-
|
| 10 |
require_api()
|
| 11 |
|
| 12 |
st.title("Monitoring Dashboard")
|
|
@@ -56,19 +61,13 @@ if st.button("Run drift check"):
|
|
| 56 |
drift = api_drift()
|
| 57 |
|
| 58 |
col1, col2, col3 = st.columns(3)
|
| 59 |
-
col1.metric(
|
| 60 |
-
"Drifted Columns",
|
| 61 |
-
int(drift["drift_summary"].get("drifted_columns_count", 0)),
|
| 62 |
-
)
|
| 63 |
col2.metric(
|
| 64 |
"Confidence Drop",
|
| 65 |
f"{drift['confidence_drift']['confidence_drop']:.2%}",
|
| 66 |
delta=f"{-drift['confidence_drift']['confidence_drop']:.2%}",
|
| 67 |
)
|
| 68 |
-
col3.metric(
|
| 69 |
-
"OOS Rate Increase",
|
| 70 |
-
f"{drift['oos_rate_drift']['oos_rate_increase']:.2%}",
|
| 71 |
-
)
|
| 72 |
|
| 73 |
if drift["confidence_drift"]["is_degraded"]:
|
| 74 |
st.error("Confidence has degraded significantly vs reference traffic.")
|
|
@@ -87,4 +86,4 @@ if st.button("Run drift check"):
|
|
| 87 |
st.warning(
|
| 88 |
"No reference/current data available yet. Run `verify_phase11.py` first to "
|
| 89 |
f"generate baseline monitoring data. ({e})"
|
| 90 |
-
)
|
|
|
|
| 1 |
+
import importlib.util
|
| 2 |
from pathlib import Path
|
| 3 |
|
| 4 |
+
_client_path = Path(__file__).parent.parent / "api_client.py"
|
| 5 |
+
_spec = importlib.util.spec_from_file_location("api_client", _client_path)
|
| 6 |
+
_module = importlib.util.module_from_spec(_spec)
|
| 7 |
+
_spec.loader.exec_module(_module)
|
| 8 |
+
api_ab_stats = _module.api_ab_stats
|
| 9 |
+
api_drift = _module.api_drift
|
| 10 |
+
api_reset_monitoring = _module.api_reset_monitoring
|
| 11 |
+
require_api = _module.require_api
|
| 12 |
|
| 13 |
import streamlit as st
|
| 14 |
|
|
|
|
|
|
|
| 15 |
require_api()
|
| 16 |
|
| 17 |
st.title("Monitoring Dashboard")
|
|
|
|
| 61 |
drift = api_drift()
|
| 62 |
|
| 63 |
col1, col2, col3 = st.columns(3)
|
| 64 |
+
col1.metric("Drifted Columns", int(drift["drift_summary"].get("drifted_columns_count", 0)))
|
|
|
|
|
|
|
|
|
|
| 65 |
col2.metric(
|
| 66 |
"Confidence Drop",
|
| 67 |
f"{drift['confidence_drift']['confidence_drop']:.2%}",
|
| 68 |
delta=f"{-drift['confidence_drift']['confidence_drop']:.2%}",
|
| 69 |
)
|
| 70 |
+
col3.metric("OOS Rate Increase", f"{drift['oos_rate_drift']['oos_rate_increase']:.2%}")
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
if drift["confidence_drift"]["is_degraded"]:
|
| 73 |
st.error("Confidence has degraded significantly vs reference traffic.")
|
|
|
|
| 86 |
st.warning(
|
| 87 |
"No reference/current data available yet. Run `verify_phase11.py` first to "
|
| 88 |
f"generate baseline monitoring data. ({e})"
|
| 89 |
+
)
|
app/pages/predict.py
CHANGED
|
@@ -1,13 +1,18 @@
|
|
|
|
|
| 1 |
import sys
|
| 2 |
from pathlib import Path
|
| 3 |
|
| 4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
|
| 6 |
import pandas as pd
|
| 7 |
import streamlit as st
|
| 8 |
|
| 9 |
-
from app.api_client import api_predict, require_api
|
| 10 |
-
|
| 11 |
require_api()
|
| 12 |
|
| 13 |
st.title("Predict Intent")
|
|
|
|
| 1 |
+
import importlib.util
|
| 2 |
import sys
|
| 3 |
from pathlib import Path
|
| 4 |
|
| 5 |
+
# load api_client directly from file to avoid 'app' package conflict
|
| 6 |
+
_client_path = Path(__file__).parent.parent / "api_client.py"
|
| 7 |
+
_spec = importlib.util.spec_from_file_location("api_client", _client_path)
|
| 8 |
+
_module = importlib.util.module_from_spec(_spec)
|
| 9 |
+
_spec.loader.exec_module(_module)
|
| 10 |
+
api_predict = _module.api_predict
|
| 11 |
+
require_api = _module.require_api
|
| 12 |
|
| 13 |
import pandas as pd
|
| 14 |
import streamlit as st
|
| 15 |
|
|
|
|
|
|
|
| 16 |
require_api()
|
| 17 |
|
| 18 |
st.title("Predict Intent")
|