Xiang Li

and it would be nice if you can share your code for the few-shot fine-tuning process.

Dear Qi, in my understanding, the main difference between evaluation setting 1 and setting 2 is that: in the former case, you randomly sample N*K support instances to guide the...

Same issue. The code is incomplete.

@jwwangchn Thanks for your updates. The training part is okay now. Looking forward to your updates on the evaluation code.

Will release the code recently.

@ANLGBOY @qianyewu How long does it take to train the model?

same issue

using pytorch-lightning=1.8 works fine for me.

same issue

line 285: train_xyz, train_label, train_feats = pickle.load(train_f, encoding='bytes')这行其实没啥作用，只有len(train_xyz)有用。一开始我们把train也做了切片，后来发现每次随机采样效果会更好。len(train_xyz)的意思是，假如我们对train切片的话会有多少个小切片，我们把他作为总共训练样本的数量。 line 344: print('Current batch/total batch num: %d/%d'%(batch_idx,num_batches)) 这行是一开始当时debu用的，没啥作用，用来看看跑了多少个step。