Sync from GitHub Actions

README.md

@@ -18,11 +18,18 @@ This is a specialized version of **MiniEmbed**, fine-tuned exclusively for **hig
 Unlike general-purpose embedding models, this model is designed to determine if two product listings—often with different titles, specifications, or formatting—refer to the **exact same physical item**.
 
 ## Use Case
+**E-commerce Product Matching & Entity Resolution**
 
-**Cross-Catalog Product Matching**
-*   **Scenario**: You have a catalog (Site A) and want to find matching products in a competitor's catalog (Site B).
-*   **Challenge**: Titles differ ("iPhone 14 128GB" vs "Apple iPhone 14 Midnight 128GB"), specs are formatted differently, and noise/distractors exist.
-*   **Solution**: This model maps semantically identical products to the same vector space, ignoring irrelevant noise while paying attention to critical specs (GB, Model Number, Color).
+This model is fine-tuned to solve the "Same Product, Different Description" problem in e-commerce:
+
+*   **Marketplace Aggregation**: Unifying listings from Amazon, Walmart, and eBay into a single catalog.
+*   **Competitor Analysis**: Matching your inventory against competitors to track pricing.
+*   **Data Cleaning**: Removing duplicates in databases where titles vary slightly (e.g., "Nike Air Max" vs "Nike Men's Air Max Shoe").
+
+**Example Challenges Handled:**
+*   **Variations**: "iPhone 14 128GB" vs "Apple iPhone 14 Midnight 128GB"
+*   **Missing Attributes**: "Sony Headphones" vs "Sony WH-1000XM5 Noise Canceling Headphones"
+*   **Formatting Differences**: "5-Pack T-Shirts" vs "T-Shirt (Pack of 5)"
 
 ## Interactive Demo
 

config.json

@@ -4,6 +4,5 @@
     "num_heads": 4,
     "num_layers": 4,
     "d_ff": 1024,
-    "max_seq_len": 128,
-    "pad_token_id": 0
+    "max_seq_len": 128
 }

demo.py

@@ -108,6 +108,33 @@ if st.button("Load Large Benchmark Dataset (100+ items)"):
         ("Lenovo ThinkPad X1 Carbon Gen 11", "Lenovo ThinkPad X1 Carbon Gen 11 14 inch", "Lenovo ThinkPad T14s"),
         ("HP Spectre x360 14", "HP Spectre x360 2-in-1 Laptop 13.5t", "HP Envy x360 15"),
         ("Microsoft Surface Pro 9", "Microsoft Surface Pro 9 (2022), 13 2-in-1", "Microsoft Surface Laptop 5"),
+        # New E-commerce items
+        ("Levis 501 Original Fit Jeans", "Levi's Men's 501 Original Fit Jeans - Dark Stonewash", "Levi's 511 Slim Fit Jeans"),
+        ("Adidas Ultraboost Light", "adidas Men's Ultraboost Light Running Shoe", "Adidas Ultraboost 1.0"),
+        ("North Face McMurdo Parka", "The North Face Men's McMurdo Parka Coat", "North Face Gotham Jacket"),
+        ("Patagonia Better Sweater", "Patagonia Men's Better Sweater 1/4-Zip Fleece", "Patagonia Synchilla Snap-T"),
+        ("Ray-Ban Aviator Classic", "Ray-Ban RB3025 Classic Aviator Sunglasses", "Ray-Ban Wayfarer"),
+        ("Herschel Little America", "Herschel Supply Co. Little America Backpack", "Herschel Heritage Backpack"),
+        ("Hydro Flask 32 oz Wide Mouth", "Hydro Flask Azure 32 oz Wide Mouth Bottle with Flex Cap", "Hydro Flask 21 oz Standard Mouth"),
+        ("Lululemon Align Leggings 25", "lululemon Align High-Rise Pant 25 inch", "Lululemon Wunder Train"),
+        ("Birkenstock Arizona Sandals", "Birkenstock Unisex Arizona Essentials EVA Sandals", "Birkenstock Boston Clogs"),
+        ("Crocs Classic Clog", "Crocs Unisex-Adult Classic Clogs without Jibbitz", "Crocs Baya Clog"),
+        ("Oral-B iO Series 9", "Oral-B iO Series 9 Electric Toothbrush with 4 Brush Heads", "Oral-B Pro 1000"),
+        ("Philips Sonicare DiamondClean", "Philips Sonicare DiamondClean Smart 9500 Rechargeable", "Philips Sonicare 4100"),
+        ("Keurig K-Elite Coffee Maker", "Keurig K-Elite Single Serve K-Cup Pod Coffee Brewer", "Keurig K-Express"),
+        ("Nespresso Vertuo Next", "Nespresso Vertuo Next Coffee and Espresso Machine by Breville", "Nespresso Essenza Mini"),
+        ("KitchenAid Artisan Stand Mixer", "KitchenAid KSM150PSER Artisan Series 5-Qt. Stand Mixer", "KitchenAid Classic Series"),
+        ("Cuisinart Food Processor 14 Cup", "Cuisinart DFP-14BCNY 14-Cup Food Processor", "Cuisinart Mini Prep Plus"),
+        ("Weber Spirit II E-310", "Weber Spirit II E-310 3-Burner Liquid Propane Grill", "Weber Traveler Portable Grill"),
+        ("Traeger Pro 575 Pellet Grill", "Traeger Grills Pro Series 575 Wood Pellet Grill and Smoker", "Traeger Ironwood 650"),
+        ("Dewalt 20V Max Cordless Drill", "DEWALT 20V MAX Cordless Drill / Driver Kit, Compact", "Dewalt 20V Max Impact Driver"),
+        ("Milwaukee M18 Fuel Impact Driver", "Milwaukee 2853-20 M18 FUEL 1/4 Hex Impact Driver", "Milwaukee M12 Impact Driver"),
+        ("Samsonite Omni PC Hardside", "Samsonite Omni PC Hardside Expandable Luggage 20 inch", "Samsonite Winfield 2"),
+        ("Away The Carry-On", "Away Travel The Carry-On Suitcase - Black", "Away The Bigger Carry-On"),
+        ("Coleman Sundome Camping Tent", "Coleman Camping Tent with WeatherTec Setup", "Coleman Skydome Tent"),
+        ("Yeti Tundra 45 Cooler", "YETI Tundra 45 Cooler Hard Cooler", "YETI Roadie 24 Cooler"),
+        ("Bose SoundLink Flex", "Bose SoundLink Flex Bluetooth Portable Speaker", "Bose SoundLink Micro"),
+        ("JBL Flip 6 Speaker", "JBL Flip 6 - Portable Bluetooth Speaker", "JBL Charge 5"),
     ]
 
     a_list = []