search

keyu-tian / SparK

[ICLR'23 Spotlight🔥] The first successful BERT/MAE-style pretraining on any convolutional network; Pytorch impl. of "Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling"
https://arxiv.org/abs/2301.03580
MIT License
1.41k stars 82 forks source link

error in downstream_imagenet #34

Closed alskdjfasdfsadf closed 1 year ago

alskdjfasdfsadf commented 1 year ago

Thanks for your great work! however, in downstream_imagenet, I found an error. I used the code '~/SparK/downstream_imagenet$ bash ./main.sh exp1 --data_path=/home/users/data --model=resnet50 --resume_from=/home/users/SparK/pretrain/output/resnet50_1kpretrained.pth --bs=32'

and the following error happened: [05-17 12:28:48] (nstream_imagenet/main.py, line 49)=> [FT start] from ep0 [05-17 12:28:48] (nstream_imagenet/main.py, line 58)=> [loader_train.sampler.set_epoch(0)] Traceback (most recent call last): File "/home/users/SparK/downstream_imagenet/main.py", line 189, in main_ft() File "/home/users/SparK/downstream_imagenet/main.py", line 60, in main_ft train_loss, train_acc = fine_tune_one_epoch(ep, args, tb_lg, loader_train, iters_train, criterion, mixup_fn, model, model_ema, optimizer, params_req_grad) File "/home/users/SparK/downstream_imagenet/main.py", line 129, in fine_tune_one_epoch inp, tar = mixup_fn(inp, tar) ValueError: too many values to unpack (expected 2) sys:1: ResourceWarning: unclosed file <_io.TextIOWrapper name=3 encoding='UTF-8'> ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 3006233) of binary: /usr/bin/python3

keyu-tian commented 1 year ago

Thanks and i fixed it in the latest commit.