films.csv

[dharanict]

I have a quick question. This is regarding milvus-sdk-go.

I see this file:
https://github.com/milvus-io/milvus-sdk-go/blob/master/examples/films.csv

I am wondering what these 8 vectors represent, for example:
[0.1292651, 0.9912045, 0.3094408, 0.046774946, 0.38359597, 0.7688596, 0.7061159, 0.5430034]

Where can I get more details? Also, is it possible to search only few vectors, say vector 0 to 2? How can I achieve it?

I am new to Milvus and also AI. Sorry if the question is basic and/or incorrect. Request you to correct as needed and respond.

Thanks
Dharani

[xiaofan-luan]

I have a quick question. This is regarding milvus-sdk-go.

I see this file: https://github.com/milvus-io/milvus-sdk-go/blob/master/examples/films.csv

I am wondering what these 8 vectors represent, for example: [0.1292651, 0.9912045, 0.3094408, 0.046774946, 0.38359597, 0.7688596, 0.7061159, 0.5430034]

Where can I get more details? Also, is it possible to search only few vectors, say vector 0 to 2? How can I achieve it?

I am new to Milvus and also AI. Sorry if the question is basic and/or incorrect. Request you to correct as needed and respond.

Thanks Dharani

Vector is usually generate by embedding models. you can train your own model(especially for recommendation system) or use pretrained. models, like openAI or cohere.

unfortunately, you can not explain any of its dimension since it's all learned from deep learning models

[dharanict]

Thank you for the quick response. Makes sense to me now.

Can you or someone tell me what was used to derive these vectors? I mean what was used to arrive at these vectors. I tried to search as much as possible but I could not find what is the source of this file for milvus-sdk-go.

[yhmo]

The films.csv is used here: https://github.com/milvus-io/milvus-sdk-go/blob/9e6be2a942a671422587ee34352eabbf563e3686/examples/insert/insert.go#L59
It is to demonstrate how to insert data into milvus, The example reads the data from films.csv, tells users how to call the Insert() interface.
The vectors in the films.csv are meaningless, just for Insert() demonstration.

In real practice, you might have a description(maybe hundreds of words) for each film. You can use an embedding model(OpenAI, etc) to generate a vector for each description. Then each record is a film, which contains the name of the film, a description of the film, and a vector of the description. Assume you have thousands of films inserted into milvus, then you can use the same embedding model(OpenAI, etc) to generate a query vector from some questions, such as "Which movie did Arnold Schwarzenegger star in", "What was the Best Picture at the 2002 Oscars", and use the query vectors to call Search() interface, milvus will return the most relevent movies for you. Basically, that is how a vector database works.

[dharanict]

Okay, got it. Thanks!