Sentence Similarity
sentence-transformers
PyTorch
Transformers
English
t5
text-embedding
embeddings
information-retrieval
beir
text-classification
language-model
text-clustering
text-semantic-similarity
text-evaluation
prompt-retrieval
text-reranking
feature-extraction
English
Sentence Similarity
natural_questions
ms_marco
fever
hotpot_qa
mteb
Eval Results
Normalizing embedding vectors
#17
by
hiranya911
- opened
Are Instructor embeddings normalized by default? I see a normalize_embeddings boolean parameter in the encode API. But with or without this parameter, encode seems to produce the same result, and it does indeed looks normalized.
from InstructorEmbedding import INSTRUCTOR
model = INSTRUCTOR('hkunlp/instructor-large')
a = model.encode([['Embed for retrieval:', 'Hello world']])
b = model.encode([['Embed for retrieval:', 'Hello world']], normalize_embeddings=True)
print(sum(item * item for item in a[0])) # Prints 1.0000
print(sum(item * item for item in b[0])) # Also prints 1.0000
nice.