Gradio Interface #34
Draft
Gradio Interface #34
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…d example queries in JSON.
Image handling - parse uploaded image to PIL - embed PIL object to vector (placeholder dummy function) - retrieve images based on provided UUID to (local file path, remote URI) mapping Python functions to interact with DB server - check server health -> Boolean - search by providing vector, top_n, nprobe -> tabular result (UUID, distance)
- Removed db interface, replaced it with client - Added scripts to handle image retrieval for HDF5 - Implemented a Gradio interface to search using the vector db server - Updated the interface documentation with instructions to initialize the app - Updated dependency for this web interface
Contributor
Author
|
Here's the updated Gradio interface with export disabled, and modified parameter names and short description. 
|
- `spark_stratified_sampler.py`: stratify sample data for leader index training - `train_leader.py`: train a FAISS leader index with cuvs enabled - `create_lookup.py`: create partition manifest and ID-mapped lookup tables from source Parquets - `train_index.py`: add data to leader index to build the complete FAISS index for the vector dataset - `convert_to_sqlite.py`: ingest parquet lookup tables into SQLite db and build index to improve search performance
added script to merge FAISS index partition into a monolithic object
added conda env config file as FAISS installation is best supported in conda channels
Mocking the server from @smenon8 but using a monolithic approach to load the merged index in-memory to enable fast search. Use SQLite as the FAISS integer ID to metadata matching.
added script that start a image retrieval server that takes the UUID from client and return search result from h5 files on disk. Used SQLite with pre-built index on `uuid` to speed-up h5 files identification. Enabled multi-threading to parallelize the h5 I/O process.
- added a new front-end application server that only communicate to the FAISS index search server and the image retreival server. - added css to temp fix the scrollbar [issue](gradio-app/gradio#10033) in Gradio - fixed a bug in `image.py` to ensure embedding vector and model is on the same device (both CPU or both GPU)
- added slurm job script for quick deployment - added guide on how to setup each server
TOL-200M FAISS Search
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Gradio web interface for the BioCLIP Vector Database image search.
See
docs/interface.mdfor documentation.