BLUERT evaluation metric consumed too much RAM
The BLEURT codes almost used up all the 24GB RAM of NVIDIA GeForce RTX 4090 to evaluate just 1 set of sentences.
COMET evaluation metric consumed too much RAM
Here is my code to check the result of evaluation metrics of the given hypothesis, reference and source sentence. Without COMET’s codes, the script runs successfully without Out-of-memory errors. However, when I uncomment the COMET codes (marked below), it used up all the 24GB RAM of NVIDIA GeForce RTX 4090 to evaluate just 1 set of sentences.