Open mandal4 opened 4 years ago
I think FT performance could be dependent on samples of novel class at Fine-tuning phase. Did you select the samples totally randomly to produce the paper's result?
I have the same issue. This may depend heavily on sampling,are results in Tab1 well selected or random sampled?
I think FT performance could be dependent on samples of novel class at Fine-tuning phase. Did you select the samples totally randomly to produce the paper's result?