From ccd622a9cfc182c0ac2612d84743358e7f7a959d Mon Sep 17 00:00:00 2001
From: Frederik Arnold <frederik.arnold@hu-berlin.de>
Date: Tue, 23 Jan 2024 07:06:15 +0100
Subject: [PATCH] Update similarity training
---
.../training/similarity/TrainSimilarity.py | 23 +++----------------
1 file changed, 3 insertions(+), 20 deletions(-)
diff --git a/indiquo/training/similarity/TrainSimilarity.py b/indiquo/training/similarity/TrainSimilarity.py
index cd11fc4..d7981bd 100644
--- a/indiquo/training/similarity/TrainSimilarity.py
+++ b/indiquo/training/similarity/TrainSimilarity.py
@@ -36,9 +36,6 @@ def main():
with open(join(input_path, 'train_set.tsv'), 'r') as train_file:
reader = csv.reader(train_file, delimiter='\t')
- # skip first row (header)
- # next(reader, None)
-
for row in reader:
ie = InputExample(texts=[row[0], row[1], row[2]])
train_examples.append(ie)
@@ -49,35 +46,21 @@ def main():
with open(join(input_path, 'val_set.tsv'), 'r') as train_file:
reader = csv.reader(train_file, delimiter='\t')
- # skip first row (header)
- # next(reader, None)
-
for row in reader:
val_anchor.append(row[0])
val_positive.append(row[1])
val_negative.append(row[2])
- # model = SentenceTransformer('deutsche-telekom/gbert-large-paraphrase-cosine')
model = SentenceTransformer(model_name)
-
- # Define your train examples. You need more than just two examples...
- # train_examples = [InputExample(texts=['My first sentence', 'My second sentence'], label=0.8),
- # InputExample(texts=['Another pair', 'Unrelated sentence'], label=0.3)]
-
train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=8)
- # train_loss = losses.CosineSimilarityLoss(model)
train_loss = losses.TripletLoss(model=model)
- # train_loss = losses.BatchHardSoftMarginTripletLoss(
- # model=model,
- # distance_metric=BatchHardTripletLossDistanceFunction.eucledian_distance,
- # )
- # evaluator = evaluation.EmbeddingSimilarityEvaluator(val_anchor, sen, scores)
- # evaluator = evaluation.TripletEvaluator(val_anchor, val_positive, val_negative)
+ evaluator = evaluation.TripletEvaluator(val_anchor, val_positive, val_negative)
# Tune the model
- model.fit(train_objectives=[(train_dataloader, train_loss)], epochs=3, warmup_steps=100, output_path=output_path)
+ model.fit(train_objectives=[(train_dataloader, train_loss)], epochs=3, warmup_steps=100,
+ evaluator=evaluator, evaluation_steps=10000, output_path=output_path)
if __name__ == '__main__':
--
GitLab