🔧 Refactor: Eliminate CSS Duplications & Enhance Security/Accessibility

scripts/deployment/bake_emotion_model.py

@@ -3,6 +3,7 @@
 import sys
 
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
+from src.constants import DEFAULT_EMOTION_MODEL_ID
 
 try:
     from huggingface_hub import login  # type: ignore
@@ -11,7 +12,7 @@
 
 
 def main() -> int:
-    model_id = os.environ.get("EMOTION_MODEL_ID", "0xmnrv/samo")
+    model_id = os.environ.get("EMOTION_MODEL_ID", DEFAULT_EMOTION_MODEL_ID)
     token = os.environ.get("HF_TOKEN")
 
     if token and login is not None:
@@ -34,4 +35,4 @@ def main() -> int:
 
 
 if __name__ == "__main__":
-    raise SystemExit(main())
+    raise SystemExit(main())

scripts/deployment/patch_config_and_upload.py

@@ -1,11 +1,17 @@
 # patch_config_and_upload.py
 # pip install -U transformers huggingface_hub
+
+# Add src to path to import constants
 import os
+import sys
+sys.path.append(os.path.join(os.path.dirname(__file__), '../..'))
 import tempfile
-from transformers import AutoConfig
 from huggingface_hub import HfApi, HfFolder
+from transformers import AutoConfig
+
+from src.constants import DEFAULT_EMOTION_MODEL_ID
 
-MODEL_ID = os.getenv("MODEL_ID", "0xmnrv/samo")
+MODEL_ID = os.getenv("EMOTION_MODEL_ID", DEFAULT_EMOTION_MODEL_ID)
 
 # Get token from environment or local storage
 TOKEN = os.getenv("HF_TOKEN")

scripts/maintenance/infer_mapping_and_eval.py

@@ -1,13 +1,18 @@
+# Add src to path to import constants
 import os
+import sys
+sys.path.append(os.path.join(os.path.dirname(__file__), '../..'))
 import numpy as np
 import torch
-from tqdm import tqdm
 from datasets import load_dataset
-from transformers import AutoTokenizer, AutoModelForSequenceClassification
-from sklearn.metrics import f1_score, accuracy_score
 from scipy.optimize import linear_sum_assignment
+from sklearn.metrics import f1_score, accuracy_score
+from tqdm import tqdm
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
+from src.constants import DEFAULT_EMOTION_MODEL_ID
 
-MODEL_ID = os.getenv("MODEL_ID", "0xmnrv/samo")
+MODEL_ID = os.getenv("EMOTION_MODEL_ID", DEFAULT_EMOTION_MODEL_ID)
 TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_HUB_TOKEN")
 DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 BATCH = int(os.getenv("BATCH_SIZE", "32"))

scripts/maintenance/metrics_test.py

@@ -1,15 +1,20 @@
 # metrics_test.py
 # pip install -U transformers datasets scikit-learn torch tqdm huggingface_hub
 
+# Add src to path to import constants
 import os
+import sys
+sys.path.append(os.path.join(os.path.dirname(__file__), '../..'))
 import numpy as np
 import torch
-from tqdm import tqdm
 from datasets import load_dataset
-from transformers import AutoTokenizer, AutoModelForSequenceClassification
 from sklearn.metrics import f1_score, accuracy_score
+from tqdm import tqdm
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
+from src.constants import DEFAULT_EMOTION_MODEL_ID
 
-MODEL_ID = os.getenv("MODEL_ID", "0xmnrv/samo")
+MODEL_ID = os.getenv("EMOTION_MODEL_ID", DEFAULT_EMOTION_MODEL_ID)
 TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_HUB_TOKEN")
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 BATCH = int(os.getenv("BATCH_SIZE", "32"))

scripts/testing/hf_serverless_smoke.py

@@ -1,13 +1,16 @@
 #!/usr/bin/env python3
+# Add src to path to import constants
 import os
+import sys
+sys.path.append(os.path.join(os.path.dirname(__file__), '../..'))
 import json
 import time
-import sys
 from typing import List, Tuple
 
+from src.constants import DEFAULT_EMOTION_MODEL_ID
 import requests
 
-HF_REPO = os.getenv("HF_REPO", "0xmnrv/samo")
+HF_REPO = os.getenv("EMOTION_MODEL_ID", DEFAULT_EMOTION_MODEL_ID)
 HF_TOKEN = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_TOKEN")
 
 API_URL = f"https://api-inference.huggingface.co/models/{HF_REPO}"

src/constants.py

@@ -8,3 +8,6 @@
 # Emotion model configuration
 DEFAULT_EMOTION_MODEL_DIR = "/app/models/emotion-english-distilroberta-base"
 EMOTION_MODEL_DIR = os.getenv("EMOTION_MODEL_DIR", DEFAULT_EMOTION_MODEL_DIR)
+
+# Default emotion model ID for HuggingFace Hub
+DEFAULT_EMOTION_MODEL_ID = "duelker/samo-goemotions-deberta-v3-large"

src/unified_ai_api.py

@@ -399,10 +399,10 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
         try:
             # Prefer loading our HF Hub model; fallback to local BERT if unavailable
             try:
-                from src.models.emotion_detection.hf_loader import (
+                from src.models.emotion_detection.hf_loader import \
                     load_emotion_model_multi_source
-                )
-                hf_model_id = os.getenv("EMOTION_MODEL_ID", "0xmnrv/samo")
+                from .constants import DEFAULT_EMOTION_MODEL_ID
+                hf_model_id = os.getenv("EMOTION_MODEL_ID", DEFAULT_EMOTION_MODEL_ID)
                 hf_token = os.getenv("HF_TOKEN")
                 local_dir = os.getenv("EMOTION_MODEL_LOCAL_DIR")
                 archive_url = os.getenv("EMOTION_MODEL_ARCHIVE_URL")