WikiCLIP_annotations / annotation /README_MERGE.md
Artanic30's picture
Upload annotation/README_MERGE.md with huggingface_hub
38293ff verified
# Large File Merge Instructions
## Overview
Some large files have been split into 1GB chunks for easier upload/download. The split files are stored in dedicated `*_files/` directories alongside the original file location.
## Files that need merging
- `kb_index.faiss` (25.3 GB)
- Split into ~26 parts in `2M_kb_eval/echosight_evqa_2M_kb/kb_index_files/`
## How to Merge
After downloading this dataset, navigate to the specific `*_files/` directory and run the merge script:
```bash
cd annotation/2M_kb_eval/echosight_evqa_2M_kb/kb_index_files/
python merge_chunks.py
```
This will automatically:
1. Detect all `.part*` files in current directory
2. Merge them back into the complete file (e.g., `kb_index.faiss`)
3. Place the merged file in the parent directory (`../`)
4. Optionally delete the `.part` files after successful merge
## Requirements
- Python 3.6+
- Sufficient disk space (at least 2x the size of the largest file during merge)
## Verification
After merging, verify the file exists in the parent directory:
```bash
ls -lh ../kb_index.faiss
```
Expected size: ~25.3 GB