Add a comment describing how generate works for Bloom models

[AMR-KELEG]

https://github.com/AMR-KELEG/DLAMA/blob/2d22d09121006a4210654bfd1a50d5a4bf6535f3/mlama/scripts/evaluate_bloom.py#L133C25-L133C33

generate uses greedy search decoding by default so you don’t have to pass any parameters to enable it. This means the parameters num_beams is set to 1 and do_sample=False.

Source: https://huggingface.co/docs/transformers/main/generation_strategies#greedy-search