#Distance Embeddings

1 messages · Page 1 of 1 (latest)

crimson ginkgo
#

Can anyone help me with the distance calculation? I've tried different ways and I'm lost

`def distances_from_embeddings(q_embeddings, embeddings, distance_metric='cosine'):

try:
    q_embeddings = np.array(q_embeddings)
    embeddings = np.array(embeddings)
    print("Shape of q_embedding:", q_embeddings.shape)
    print("Shape of embeddings:", embeddings.shape)

    q_embeddings_reshaped = q_embeddings.reshape(1, -1)
    embeddings_reshaped = embeddings.reshape(1, -1)
    if distance_metric == 'cosine':
        distances = cdist(q_embeddings_reshaped, embeddings_reshaped, metric='cosine')[0]
    elif distance_metric == 'euclidean':
        distances = cdist(q_embeddings_reshaped, embeddings_reshaped, metric='euclidean')[0]
    else:
        raise ValueError("Invalid distance metric.")
    embeddings['distances'] = distances
except openai.APIStatusError as e:
    print(e.status_code)
except ValueError as e:
    print("invalid metric.", e)
except KeyError as e:
    print("no key or dataframe:", e)
except IndexError as e:
    print("invalid consult.", e)
except Exception as e:
    print("Erro :", e)
return embeddings`

result:
invalid metric. XA and XB must have the same number of columns (i.e. feature dimension.)
'Completion' object is not subscriptable
Shape of q_embedding: (1536,)
Shape of embeddings: (1,)

minor plaza
#

Are you trying to compare the cos similarity?

crimson ginkgo
#

The objective is to calculate the embedding distance and q_embedding and put it in the distances variable

#

I wanted to send more of the code to give more context but there was a character limit