feat(new models): added new models, refactor the ui

app.py

@@ -5,10 +5,21 @@ import os
 from homepage2vec.model import WebsiteClassifier as Homepage2Vec
 
 EXAMPLES = [
-    ["gpt3.5", "tanjasenghaasdesigns.de"],
-    ["gpt3.5", "epfl.ch"],
-    ["gpt3.5", "cc.cz"],
-    ["gpt3.5", "promaminky.cz"]
+    # Personal site
+    ["original", "tanjasenghaasdesigns.de"],
+    ["finetuned-gpt4", "tanjasenghaasdesigns.de"],
+
+    # EPFL
+    ["finetuned-gpt3.5", "epfl.ch"],
+    ["finetuned-gpt4", "epfl.ch"],
+
+    # Czech Crunch - czech tech news
+    ["original", "cc.cz"],
+    ["finetuned-gpt4", "cc.cz"],
+
+    # Promaminky - czech site for moms
+    ["original", "promaminky.cz"],
+    ["finetuned-gpt3.5", "promaminky.cz"],
 ]
 
 
@@ -24,8 +35,11 @@ def predict(model_choice : str, url : str) -> Dict[str, float]:
         Dict[str, float]: The categories and their corresponding scores.
     """
 
-    # Define the model directory path
-    model_dir = os.path.join("models", model_choice)
+    if model_choice == "original":
+        model_dir = os.path.join("models", "homepage2vec")
+    else:
+        which_gpt = model_choice.split("-")[1]
+        model_dir = os.path.join("models", "finetuned", which_gpt)
 
     # Initialise model
     model = Homepage2Vec(model_dir=model_dir)
@@ -44,11 +58,11 @@ def predict(model_choice : str, url : str) -> Dict[str, float]:
 
 iface = gr.Interface(
     fn=predict,
-    inputs=[gr.Dropdown(choices=["gpt3.5", "gpt4"], label="Select Model"),
-            gr.Textbox(label="Enter Website URL", placeholder="www.mikasenghaas.de")],
+    inputs=[gr.Dropdown(choices=["original", "finetuned-gpt3.5", "finetuned-gpt4"], label="Select Model", show_label=True, value="finetuned-gpt4"),
+            gr.Textbox(label="Enter Website's URL or domain", placeholder="e.g. ikea.com")],
     outputs=gr.Label(num_top_classes=14, label="Predicted Labels", show_label=True),
     title="Homepage2Vec",
-    description="Use Homepage2Vec to predict the categories of any website you wish.",
+    description="Select a version of the Homepage2Vec model and enter a website's URL or domain to predict its categories. The original model was trained on 886K websites from Curlie directory. The finetuned models, in addition, were trained on GPT annotated websites. On average, the fintuned models should predict more labels than the original model while maintaining high accuracy.",
     examples=EXAMPLES,
     live=False,
     allow_flagging="never",

models/{gpt3.5 → finetuned/gpt3.5}/model.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d40bb85c577a8c0951b585714c35fa10509267f0d52ec1c6952f650e9622887
+oid sha256:f3c325b49245d46326daa3c9ff6dd4e52b806cf88b8986e27598419764aed3ab
 size 19072308

models/finetuned/gpt4/features.txt

@@ -0,0 +1,8 @@
+f_tld 27
+f_url 768
+f_metatags 30
+f_title 768
+f_description 768
+f_keywords 768
+f_links_50 768
+f_text_100 768

models/finetuned/gpt4/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b0c10a4a65ad90f87e33f7c62364b4d44991e72819d186785beff84c24f0f22
+size 19072308

models/homepage2vec/features.txt

@@ -0,0 +1,8 @@
+f_tld 27
+f_url 768
+f_metatags 30
+f_title 768
+f_description 768
+f_keywords 768
+f_links_50 768
+f_text_100 768

models/homepage2vec/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:634de448281c5026597c05b5132ac0dc802305689adc104bde71949186999215
+size 19072357