Add cosine_simirarity_v1 which is more faster than the original cosine_similarity method.

[hwaking]

add cosine_simirarity_v1 method which is 5 times faster than the orginal cosine_similarity method.

add cosine_simirarity_v1 method which is 5 times faster than the orgianl cosine_similarity method.

@vowelparrot update math_utils.py file add cosine_simirarity_v1 which is more faster than the original cosine_similarity method. the time cost test like this: ''' x = [[1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4]] * 2000 y = [[1, 2, 3, 5, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4], [1, 2, 2, 4, 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4]] * 30000 resp = cosine_similarity(x, y) resp1 = cosine_simirarity_v1(x, y) print('resp： ', resp) print('score_list： ', resp1, type(resp1))

output: Function cosine_similarity Took 1.8206 seconds Function cosine_simirarity_v1 Took 0.3706 seconds

resp： [[0.99663517 0.99597014 0.99663517 ... 0.99597014 0.99663517 0.99597014] [0.99663517 0.99597014 0.99663517 ... 0.99597014 0.99663517 0.99597014] [0.99663517 0.99597014 0.99663517 ... 0.99597014 0.99663517 0.99597014] ... '''

[hwaking]

@vowelparrot

[vowelparrot]

Thanks for the PR! The reason we didn't use sklearn is we don't want to add it as a required dependency. We could add an import check at the top and use sklearn if it's available though.

[vowelparrot]

Thanks for the PR! The reason we didn't use sklearn is we don't want to add it as a required dependency. We could add an import check at the top and use sklearn if it's available though.

[dev2049]

if folks wanted to use this could they just import sklearn directly wherever it's actually being used? not sure how much value the wrapper adds

[leo-gan]

@hwaking Hi , could you, please, resolve the merging issues and address the last comments (if needed)? After that, ping me and I push this PR for the review. Thanks! Please, let me know if you need help! The example of the "on-demand import" is here.

[efriis]

Closing because the PR wouldn't line up with the current directory structure of the library (would need to be in /libs/langchain/langchain instead of /langchain). Feel free to reopen against the current head if it's still relevant!