Fixed backend issues: issue #212

[Dasux]

This PR addresses issue #212
Issues in script.sh and server.py have been resolved

Changes made:

• Replaced all instances of the deprecated jors_winkler module with jaro_winkler_similarity in server.py and main.py to resolve the import error.

• Implemented a fallback mechanism to force the script to use the CPU instead of the GPU if insufficient VRAM is detected, addressing the CUDA out-of-memory error on devices with limited GPU memory.

• Set the environment variable os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True" to avoid memory fragmentation and improve GPU memory allocation handling.

NOTE:
Two of the files I've added in this PR.. jaro_wrinkler.py and convert_score_and_distance_functions.py should be replaced from their respective directories... namely:

1. EduAid/backend/venv/lib/python3.12/site-packages/similarity/jaro_winkler.py
2. EduAid/backend/venv/lib/python3.12/site-packages/similarity/convert_score_and_distance_functions.py

These changes should fix the reported issues with the server script and improve the overall reliability on systems with limited GPU resources.

[Aditya062003]

Hey, would be great if you could provide an screen capture for the issue resolved!

[Dasux]

sure,,,

The script.sh update was already handled in #188 by someone else, so I focused on fixing the jellyfish function issues.

a little more context of the issue

The function jaro_winkler no longer exists in jellyfish—it has been renamed to jaro_winkler_similarity.

This caused import issues in multiple files, such as:

• EduAid/backend/venv/lib/python3.12/site-packages/similarity/jaro_winkler.py

• EduAid/backend/venv/lib/python3.12/site-packages/similarity/convert_score_and_distance_functions.py

I updated these files to use the correct method name:

Regarding server.py
I believe I have fixed the issue, but I haven't been able to fully test it due to hardware limitations.

I’m running into a CUDA out-of-memory error—my iGPU has 1.69GB VRAM, with only 41MB free. The script tries to load 96MB, which causes it to fail.

I’ve looked into alternatives, but so far, I haven’t found a workaround.

I think it'd be better if someone else could verify server.py fix since i cant test it properly.