initial commit

Adding hugging-face

initial commit

Use the RAG DB to store the data and finally functional poc. Next make different tables for different videos

Workaround Gradio UI

Wip=> Fix code erros regarding filepath, metadatas

Build basic mvp that brings in all the things we learned so far

Confirm DataStorage in LancdDB

Create table in multi model LanceDB

add llava and whisper support

Update README.md

Update README.md

Update README.md

Update ReadMe: with learning

Fix embed_query to check dict and convert tensor to list

Use transformer to get embeddings from hugging and use text embeddings

Remove Prediction Gurad for document embeddings

Changing embedding from PredictionGuard to Local

Add extra file for storing multimodel data in rag

Remove Prediction Gurad - in progress

Remove Prediction Gurad - in progress

Option code to use display

Prediction Guard hit again with LLava - closing this here

Video Captionaing and Frame Generation

Add utility files that helps in downloading videos, extracting webvvt, and other video and caption related helper methods

Add utility files that helps in downloading videos, extracting webvvt, and other video and caption related helper methods

Plot shows, however, I am not certain if it is showing the right data. The main confusion is that to see vector data of 512 dimenstion, you need to reduce it to 2 dimension on a scaler plot. The function given here does not work. First np.concatenate does not like lists that has embeddings and has grad init. It want me to detach numpy. The second problem is with MinMaxScaler method that has issue with dimention, it expects 2, but one is given. Not very clear on this

Remove Hugging Face download

Must Refactor Code :)

Remove ununsed files

Add addition requirements

File Rename

Calculate Cosine and Euclidean Distance

Fix tensor to numphy convert

Load Tensor and calculate distance

Rename file, remove unused imports

Save Embeddings in a file to save process every time

Add an optional file to show multiple ways to use BridgeTower

Add an optional file to show multiple ways to use BridgeTower

Remove Prediction Guard and correct the code so model can return 512 vector

Remove PredictionGuard and use Transform to download model

Write Step 2 to download data

Update ReadMe: Understand Vector DB

Update Readme with chatGPT recommended text

Update Readme: Learn how to prepare data (image, text, video) for multimodel, without getting much into ML

Update Readme: Finish Basic use of Gradio

Add Basic components with GR Blocks, GR Chat Interface

First Demo with Gradio