"RuntimeWarning: invalid value encountered in arccos" in Cosine similarity

The new added code has some issue:

# Distance based on cosine similarity dot = np.sum(np.multiply(embeddings1, embeddings2), axis=1) norm = np.linalg.norm(embeddings1, axis=1) * np.linalg.norm(embeddings2, axis=1) similarity = dot / norm dist = np.arccos(similarity) / math.pi

The valid range for arccos is [-1,1], I guess we need to clip the value before np.arccos.

[supermanhuyu]

after add np.minimum , the warning disappeared.

 dist = np.arccos(np.minimum(1, similarity)) / math.pi

[davidsandberg]

The similarity should already be in the range [-1, 1] (https://en.wikipedia.org/wiki/Cosine_similarity) so I’m not sure clipping would fix the issue. Could you print out the two embeddings for the case where you see the problem? It could be that they are the problem, e.g. containing NaNs, Infs etc. But if it turns out that the similarity goes out of the range due to some rounding effects clipping would make perfect sense. Thanks!