ENTAILMENT_USE_API = False

src/config.py

@@ -17,7 +17,7 @@ ENTAILMENT_BATCH_SIZE = 16
 # When True, run entailment via the HF serverless Inference API (uses HF's
 # compute, no GPU needed on this Space). When False, load DeBERTa locally inside
 # the @spaces.GPU window (requires GPU/ZeroGPU hardware).
-ENTAILMENT_USE_API = True
+ENTAILMENT_USE_API = False
 ENTAILMENT_API_MODEL = "microsoft/deberta-v2-xlarge-mnli"
 ENTAILMENT_API_WORKERS = 8
 LOG_REPO = "cryotud/probabot-calibration-logs"