fix for empty string input to openai/text-embedding-3-large

[ayush1298]

fix for #1650.
I dont have OpenAI API key to test whether its working correctly or not. @KennethEnevoldsen Can you please test it, when you get time.

Code Quality

• Code Formatted: Format the code using make lint to maintain consistent style.

Documentation

• Updated Documentation: Add or update documentation to reflect the changes introduced in this PR.

Testing

• New Tests Added: Write tests to cover new functionality. Validate with make test-with-coverage.
• Tests Passed: Run tests locally using make test or make test-with-coverage to ensure no existing functionality is broken.

[ayush1298]

@KennethEnevoldsen I have added fix that should based on what you explained. Also, to test it as we are calling model.encode so we should require OPENAI API KEY.
1 more thing is, I tested this using OPENAI API KEY(got it from friend), but still I am getting error for empty strings. I tried checking it but error was occuring even before calling of encode method. I was not able to find out what was cause of the error though. Can you please have a look at it?

[KennethEnevoldsen]

I addded some suggestions. Do give it another try and try to give me a debug trace if it doesn't work.

[KennethEnevoldsen]

this already fill out the empty texts.

You can then simply fill in the text_embeddings using:

mask = [i for i, t in enumerate(text) if t.strip()]
final_emb[mask, :] = text_emb

Here is a small sample to show the idea:

import numpy as np

matrix1 = np.zeros((5, 4))
matrix2 = np.random.rand(3, 4)
matrix1[[0, 2, 4],:] = matrix2

[ayush1298]

Didn't get these one. Current implementation was 1st creating whole vector of zeros, and then only setting text embeddings, at positions where non-empty text is present. I think you are suggesting same only.

[ayush1298]

from __future__ import annotations
import os
import numpy as np
from mteb.models.openai_models import OpenAIWrapper

os.environ["OPENAI_API_KEY"] = "API-KEY"
wrapper = OpenAIWrapper(
    model_name="text-embedding-3-small",
    max_tokens=8191,
    embed_dim=1536,
)

# 3) call encode on empty, non‑empty, mixed
cases = [
    # ["hello world"],   
    ["", "foo", ""], 
]
print("Testing OpenAIWrapper with different cases:")

for case in cases:
    print(f"\nTesting case: {case!r}")
    try:
        out = wrapper.encode(case)
        print("output:", out)
        print("Shape:", out.shape)

    except Exception as e:
        print(f"Error: {e}")

@KennethEnevoldsen Getting the following output when running the above testing code:

Testing OpenAIWrapper with different cases:

Testing case: ['hello world']
output: [[-0.00676333 -0.03919632  0.03417581 ... -0.01964353 -0.01937133
  -0.02247135]]
Shape: (1, 1536)

Testing case: ['', 'foo', '']
Error: Error code: 400 - {'error': {'message': "'$.input' is invalid. Please check the API reference: https://platform.openai.com/docs/api-reference.", 'type': 'invalid_request_error', 'param': None, 'code': None}}

So, code is running but still not able to handle empty strings. I have tried to put some logging statements at start of encode method, but when input is empty, then I am getting error even before calling this encode method.
Can you have a look at it now?

[KennethEnevoldsen]

Thanks for taking the time on this @ayush1298. I built upon you earlier pr. and made a fix: #2676

[ayush1298]

Thanks for taking the time on this @ayush1298. I built upon you earlier pr. and made a fix: #2676

Thanks @KennethEnevoldsen