dandelin
Flickr30k Finetune results does not match the provided checkpoint
Hi authors,
I take the provided pretrained 200k checkpoint and did the finetuning of flickr30k. The IR and TR scores after are 64.5 and 81.7. The TR score lower than the one in the paper. My finetuning command is
$PYTHONBIN run.py with data_root=vilt_dataset/ \
num_gpus=8 num_nodes=1 task_finetune_irtr_f30k \
per_gpu_batchsize=4 load_path="weights/vilt_200k.ckpt" \
exp_name="f30k/finetune_official"
I also test the given vilt_irtr_f30k.ckpt and the results is good, with IR=65.3, TR=83.5. Can I ask what is the process of getting vilt_irtr_f30k.ckpt?
@JACKHAHA363
The fine-tuning results can be unstable due to augmentations. Also, we have only trained the IR/TR fine-tuning models for a single time. You may increase the training epochs (greater than 10 epochs, maybe 20 epochs?) to get more stable and better results.
I tried longer epochs but that end up overfitting with increasing val loss. Would you mind providing the checkpoint for 100k steps also?
@JACKHAHA363 Sure, you can grab it here https://www.dropbox.com/s/lcqmbx587szaox3/vilt_100k_wwm_pretrain.ckpt?dl=0 (will be expired someday)
Are you able to solve this issue? @JACKHAHA363 I have similar issues on both flicker and coco retrieval.
Hi, bro. I found ir/tr evaluation result on flickr is still unstable even using official finetuned checkpoint. Sometimes I got 63.94(ir)/83.6(tr), sometimes it changed to 64.3(ir)/83.7(tr). How do you think it? @dandelin @JACKHAHA363
Hi.Thanks for your reply. But I find shuffle in IR/TR image loader (code) is exactly False. In fact, shuffle is not exexplicitly set to False. However, the default value in torch.utils.data.DataLoader is False. Besides, shuffle in dist_sampler is also False.
I have no idea about the unsteable result.
T_T ...
------------------ 原始邮件 ------------------ 发件人: "Wonjae @.>; 发送时间: 2021年12月30日(星期四) 晚上10:36 收件人: @.>; 抄送: "by @.>; @.>; 主题: Re: [dandelin/ViLT] Flickr30k Finetune results does not match the provided checkpoint (#13)
Hi @byougert
Interpolating the position embedding dynamically for every image batch can cause subtle fluctuation while doing a batch evaluation. During computing IR/TR, we pass batch of images to vit.visual_embed (code), and inside of vit.visual_embed method, we calculate the maximum width and height given the batch of images dynamically and interpolate the pos_embed upto those dimensions (code).
I think setting shuffle=False argument in IR/TR image loader (code) would make the evaluation not fluctuate by fixing the order of image batches.
It seems I somehow deleted the argument while polishing the code for public release. Please set the argument and let me know it fixes the issue.
— Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you were mentioned.Message ID: @.***>
Hi @byougert
Oops, you got the mail. I deleted the comment right after posted it as I noticed I put shuffle=False in DistributedSampler(image_dset, shuffle=False).
Though after quick investigation, I found the true reason.
It was the precision=16, set in https://github.com/dandelin/ViLT/blob/master/run.py#L51.
After setting precision=32 during evaluation I was able to get stable result.
I guess the score from rank_output is very cluttered so they need larger precision.
Thanks for the report and I will revise the EVAL.md. :)
Hi, bro. Yes, i received your message in my mail but couldn't find the reply in github. hhhh.... Thanks for your reply and nice work.
Hi, @dandelin I'm sorry to say that the result seems still puzzled. Last night, when I changed precision to 32 during evaluation, two similar but NOT SAME results appeared, which showed one was 0.6480(ir)/0.8370(tr) but the other was 0.6460(ir)/0.8370(tr). Acatlly, seed is exactly fixed to 0. I have no idea what causes the differece. Y_Y