wip

app.py

@@ -15,7 +15,8 @@ from io import BytesIO
 
 # For baseline 'sentence-transformers/paraphrase-albert-base-v2'
 model_name = 'output/training_OnlineConstrativeLoss-2023-03-14_01-24-44'
-model_name = 'output/training_OnlineConstrativeLoss-2023-03-17_16-10-39'
+model_name = 'output/training_OnlineConstrativeLoss-2023-03-17_23-15-52'
+model_name = 'output/training_OnlineConstrativeLoss-2023-03-17_23-50-15'
 
 similarity_threshold = 0.9
 

quick_evaluate.py

@@ -13,6 +13,7 @@ model_name = 'output/training_OnlineConstrativeLoss-2023-03-10_11-17-15'
 model_name = 'output/training_OnlineConstrativeLoss-2023-03-11_00-24-35'
 model_name = 'output/training_OnlineConstrativeLoss-2023-03-11_01-00-19'
 model_name='output/training_OnlineConstrativeLoss-2023-03-17_16-10-39'
+model_name='output/training_OnlineConstrativeLoss-2023-03-17_23-15-52'
 model_sbert = SentenceTransformer(model_name)
 
 

train.py

@@ -24,10 +24,12 @@ logger = logging.getLogger(__name__)
 
 
 #As base model, we use DistilBERT-base that was pre-trained on NLI and STSb data
-model = SentenceTransformer('sentence-transformers/paraphrase-albert-base-v2')
+model_name ='sentence-transformers/paraphrase-albert-base-v2'
+model_name = 'sentence-transformers/all-mpnet-base-v1'
+model = SentenceTransformer(model_name)
 num_epochs = 12
 # Smaller is generally better more accurate results.
-train_batch_size = 5
+train_batch_size = 10
 
 #As distance metric, we use cosine distance (cosine_distance = 1-cosine_similarity)
 distance_metric = losses.SiameseDistanceMetric.COSINE_DISTANCE