update demo

README.md

@@ -4,7 +4,7 @@ emoji: 👀
 colorFrom: purple
 colorTo: red
 sdk: streamlit
-sdk_version: 1.17.0
+sdk_version: 1.25.0
 app_file: app.py
 pinned: false
 ---

app.py

@@ -1,35 +1,13 @@
 import copy
-import enum
 import pandas as pd
-from typing import List, Optional
+from typing import Optional
 
 import requests
 import streamlit as st
 
 
-MODEL_DESCRIPTION = """
-**keras_2_0**: Current production model\n
-**keras_image_embeddings_3_0**: same as `keras_300_epochs_3_0` but with image embedding as input\n
-**keras_300_epochs_3_0**: trained on 300 epochs with product name, ingredients, OCR-extracted ingredients and nutriments as input\n
-**keras_ingredient_ocr_3_0**: same as `keras_sota_3_0`, but trained on less epochs\n
-**keras_baseline_3_0**: model trained with product name, ingredients and nutriments as input\n
-**keras_original_3_0**: same inputs as **keras_2_0** (product name + ingredients), but retrained\n
-**keras_product_name_only_3_0**: model with only product name as input
-"""
-
 http_session = requests.Session()
 
-@enum.unique
-class NeuralCategoryClassifierModel(enum.Enum):
-    keras_2_0 = "keras-2.0"
-    keras_image_embeddings_3_0 = "keras-image-embeddings-3-0"
-    keras_300_epochs_3_0 = "keras-300-epochs-3-0"
-    keras_ingredient_ocr_3_0 = "keras-ingredient-ocr-3.0"
-    keras_baseline_3_0 = "keras-baseline-3.0"
-    keras_original_3_0 = "keras-original-3.0"
-    keras_product_name_only_3_0 = "keras-product-name-only-3.0"
-
-
 LOCAL_DB = False
 
 if LOCAL_DB:
@@ -40,9 +18,9 @@ else:
 PREDICTION_URL = ROBOTOFF_BASE_URL + "/predict/category"
 
 
-@st.cache
-def get_predictions(barcode: str, model_name: str, threshold: Optional[float] = None):
-    data = {"barcode": barcode, "predictors": ["neural"], "neural_model_name": model_name}
+@st.cache_data
+def get_predictions(barcode: str, threshold: Optional[float] = None):
+    data = {"barcode": barcode, "predictors": ["neural"]}
     if threshold is not None:
         data["threshold"] = threshold
 
@@ -52,22 +30,19 @@ def get_predictions(barcode: str, model_name: str, threshold: Optional[float] =
 
 def display_predictions(
     barcode: str,
-    model_names: List[str],
     threshold: Optional[float] = None,
 ):
     debug = None
-    for model_name in model_names:
-        response = get_predictions(barcode, model_name, threshold)
-        response = copy.deepcopy(response) 
-        if model_name != NeuralCategoryClassifierModel.keras_2_0.name and "debug" in response:
-            if debug is None:
-                debug = response["debug"]
-            response.pop("debug")
-        st.markdown(f"**{model_name}**")
-        st.write(pd.DataFrame(response["predictions"]))
+    response = get_predictions(barcode, threshold)
+    response = copy.deepcopy(response)
+    if "debug" in response:
+        if debug is None:
+            debug = response["debug"]
+        response.pop("debug")
+    st.write(pd.DataFrame(response["predictions"]))
     
     if debug is not None:
-        st.markdown("**v3 debug information**")
+        st.markdown("**Debug information**")
         st.write(debug)
 
 
@@ -81,17 +56,9 @@ barcode = st.sidebar.text_input(
 )
 threshold = st.sidebar.number_input("Threshold", format="%f", value=0.5) or None
 
-st.sidebar.write("---\n# Model description\n" + MODEL_DESCRIPTION)
-model_names = st.multiselect(
-    "Name of the model",
-    [x.name for x in NeuralCategoryClassifierModel],
-    default=[x.name for x in NeuralCategoryClassifierModel],
-)
-
 if barcode:
     barcode = barcode.strip()
     display_predictions(
         barcode=barcode,
         threshold=threshold,
-        model_names=model_names,
     )