Upload 2 files

app.py

@@ -14,13 +14,13 @@ is_hf_space = "SPACE_ID" in os.environ
 
 if is_hf_space:
     # In a Space, load model from the Hub and data from the repo root
-    model_path = os.environ.get("MODEL_REPO_ID", "philtoms/minilm-alice-base-rsft-v1")
-    data_path = "alice_pairs.jsonl"
+    model_path = os.environ.get("MODEL_REPO_ID", "philtoms/minilm-alice-base-rsft-v2")
+    data_path = "training_triplets.jsonl"
     print(f"Running on HF Spaces. Using model from Hub: {model_path}")
 else:
     # Locally, construct absolute paths based on the script's location
-    model_path = os.path.join(script_dir, "..", "models", "minilm-alice-base-rsft-v1", "final")
-    data_path = os.path.join(script_dir, "..", "data", "alice_pairs.jsonl")
+    model_path = os.path.join(script_dir, "..", "models", "minilm-alice-base-rsft-v2", "final")
+    data_path = os.path.join(script_dir, "..", "data", "training_triplets.jsonl")
     print(f"Running locally. Using local model at: {model_path}")
 
 # --- Model and Tokenizer Loading ---
@@ -39,7 +39,7 @@ with open(data_path, "r") as f:
     for line in f:
         dataset.append(json.loads(line))
 
-corpus = [item["passage"] for item in dataset]
+corpus = [item["positive"] for item in dataset]
 
 # Pre-compute corpus embeddings
 with torch.no_grad():