dandelin/ViLT issues and pull requests

#95 - RuntimeError: Invalid pretrained config, cannot load weights. Use `pretrained=False` for random init.

Issue - State: open - Opened by nameyun 2 months ago

#94 - missing file

Issue - State: open - Opened by hussainafroz 3 months ago

#93 - Initial commit for 2024 VizWiz challenge

Pull Request - State: closed - Opened by harrychien1311 7 months ago

#92 - When distributed training was performed, the program remained unresponsive

Issue - State: open - Opened by mumu029 10 months ago

#91 - ValueError: Connection error, and we cannot find the requested files in the cached path. Please try again or make sure your Internet connection is on.

Issue - State: open - Opened by Spring24ch 10 months ago - 1 comment

#90 - requests.exceptions.MissingSchema: Invalid URL 'None': No scheme supplied. Perhaps you meant https://None?

Issue - State: open - Opened by lingshen233 11 months ago

#89 - Which python could I use

Issue - State: open - Opened by Thealove 11 months ago

#88 - 更改输入

Issue - State: open - Opened by wzh226 12 months ago - 1 comment

#87 - error: subprocess-exited-with-error

Issue - State: open - Opened by My12123 about 1 year ago - 1 comment

#86 - KeyError: 'false_image_0'

Issue - State: open - Opened by Chendaqiang01 over 1 year ago

#85 - ViLT on GQA

Pull Request - State: open - Opened by keshavshivkumar over 1 year ago

#84 - AttributeError: 'TracebackException' object has no attribute 'exc_traceback'

Issue - State: open - Opened by liuliAI over 1 year ago - 3 comments

#83 - cannot import name 'Final' from 'typing'

Issue - State: open - Opened by lijiabad over 1 year ago - 2 comments

#82 - What could be the reason that the model weights are not updating while finetuning?

Issue - State: open - Opened by DDXDaniel over 1 year ago - 2 comments

#81 - RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

Issue - State: open - Opened by jiajia95Murphy over 1 year ago - 2 comments

#80 - Can't the weight folder be opened before the pre-training is over?

Issue - State: open - Opened by yang178908 over 1 year ago

#79 - fine-tuning ViLT for MLM task with a new dataset

Issue - State: open - Opened by Ellyuca over 1 year ago

#78 - pyarrow.lib.ArrowInvalid: Not an Arrow file

Issue - State: closed - Opened by psrimanreddy almost 2 years ago - 2 comments

#77 - Mistakes in vqa_dict.json ？

Issue - State: closed - Opened by boqian-li almost 2 years ago

#76 - What is the image resolution during VQA finetuning and pretraining?

Issue - State: open - Opened by sanyalsunny111 about 2 years ago

#75 - The problem of fine-flickr30k

Issue - State: open - Opened by wuqiang12345 about 2 years ago

#74 - pretrain datasets

Issue - State: open - Opened by mactavish91 about 2 years ago

#73 - Question about train on coco dataset

Issue - State: open - Opened by Bmilab22 about 2 years ago - 1 comment

#72 - RuntimeError: CUDA error: invalid device function

Issue - State: open - Opened by lonestar234028 about 2 years ago - 3 comments

#71 - train on coco dataset

Issue - State: open - Opened by weiyutao886 about 2 years ago - 7 comments

#70 - How to set the config to create a stand_alone commandline demo ?

Issue - State: closed - Opened by Ngheissari about 2 years ago - 1 comment

#69 - About MS-COCO pre-training dataset

Issue - State: open - Opened by 4fee8fea about 2 years ago - 1 comment

#68 - About SBU Caption dataset

Issue - State: open - Opened by 4fee8fea about 2 years ago - 1 comment

#67 - How to use the modal-type embedding in the output of encoder?

Issue - State: open - Opened by rginjapan over 2 years ago - 1 comment

#66 - train customer data

Issue - State: open - Opened by guanhdrmq over 2 years ago - 1 comment

#65 - utils/write_<>.py: Is there any way to write to disk on the fly instead of loading the entire dataFrame into memory?

Issue - State: closed - Opened by zdxdsw over 2 years ago - 1 comment

#64 - ViLT training time

Issue - State: open - Opened by xii-rao over 2 years ago - 1 comment

#63 - Why is answers set to 0 for irtr even for the positive case?

Issue - State: closed - Opened by TheShadow29 over 2 years ago - 3 comments

#62 - How to use ViLT model for Spanish Text ?

Issue - State: open - Opened by karndeepsingh over 2 years ago - 1 comment

#61 - Finetune is failing ValueError: operands could not be broadcast together with shapes (384,576) (3,)

Issue - State: open - Opened by amitkayal over 2 years ago - 2 comments

#60 - Flickr30k Image and Text Retrieval - Query regarding training

Issue - State: open - Opened by gchhablani over 2 years ago - 2 comments

#59 - Waiting for localhost

Issue - State: open - Opened by seifmaged31 over 2 years ago

#58 - Question about GCC

Issue - State: open - Opened by Sry2016 over 2 years ago - 1 comment

#57 - AttributeError: 'LightningDistributedDataParallel' object has no attribute '_sync_params'

Issue - State: open - Opened by KimSoybean over 2 years ago - 1 comment

#56 - Checkpoint file for VQA might be wrong

Issue - State: open - Opened by jia2lin3yuan1 over 2 years ago

#55 - while read file idx 2740206 in conceptual_caption_train_0 -> image file is truncated

Issue - State: open - Opened by campper over 2 years ago - 1 comment

#54 - what is the meaning of "split" in /vilt/utils/write_conceptual_caption.py

Issue - State: open - Opened by campper over 2 years ago

#53 - Question about pad_choice

Issue - State: open - Opened by Richar-Du over 2 years ago

#52 - Question about ITM pretraining

Issue - State: open - Opened by EagleW almost 3 years ago

#51 - Got better results than in the paper:

Issue - State: open - Opened by JoanFM almost 3 years ago - 2 comments

#50 - integrate with Lightning ecosystem CI

Issue - State: open - Opened by pl-ghost almost 3 years ago

#49 - COCO split for pre-training

Issue - State: open - Opened by sanjayss34 almost 3 years ago

#48 - ITM Objectives task will not be enabled？

Issue - State: open - Opened by CQUTWangHong almost 3 years ago - 2 comments

#47 - self.mask_token at 553 line in vision_transformer.py #35 I find the self.mask_token has not been defined.

Issue - State: open - Opened by yr666666 almost 3 years ago

#46 - How to run programs on multiple machines, such as 4 machines with 8 gpus(4*8=32 in total)?

Issue - State: closed - Opened by raojay7 almost 3 years ago

#45 - Question about GCC dataset download

Issue - State: open - Opened by yr666666 almost 3 years ago - 1 comment

#44 - python run.py with data_root=content/datasets num_gpus=2 num_nodes=1 task_mlm_itm whole_word_masking=True step100k per_gpu_batchsize=64

Issue - State: open - Opened by F-Yuan303 almost 3 years ago - 6 comments

#43 - Reproduce Flickr30k Evaluation results - DataSet problem

Issue - State: closed - Opened by JoanFM almost 3 years ago - 2 comments

#42 - Incorrect dataset link.

Issue - State: closed - Opened by sunanhe almost 3 years ago - 1 comment

#41 - Pre-training Time

Issue - State: closed - Opened by haoshuai714 almost 3 years ago - 1 comment

#40 - ViltTransformerSS usage clarification

Issue - State: open - Opened by JoanFM almost 3 years ago - 10 comments

#39 - missing values in vqa_dict.json

Issue - State: closed - Opened by csarron almost 3 years ago - 1 comment

#38 - Visual Grounding

Issue - State: open - Opened by prashan almost 3 years ago - 1 comment

#37 - small difference between paper and code about token type embedding

Issue - State: closed - Opened by AAbathur almost 3 years ago - 2 comments

#36 - Add ViLT to HuggingFace Transformers

Issue - State: closed - Opened by NielsRogge almost 3 years ago - 13 comments

#35 - `self.mask_token` at 553 line in `vision_transformer.py`

Issue - State: open - Opened by hspark1212 almost 3 years ago - 6 comments

#34 - Is it necessary to manually add the sampler in the dataloader?

Issue - State: closed - Opened by zhiqiangdon almost 3 years ago - 2 comments

#33 - About finetune on f30k.

Issue - State: open - Opened by GuoBruce almost 3 years ago - 2 comments

#32 - inference image captioning

Issue - State: open - Opened by trucvip123 almost 3 years ago - 2 comments

#31 - Possible out-of-memory issue of dataloader

Issue - State: closed - Opened by zhiqiangdon almost 3 years ago - 4 comments

#30 - Download and process GCC and SBU

Issue - State: open - Opened by wanng-ide almost 3 years ago - 10 comments

#29 - Time cost when fine-tuning on COCO-irtr dataset

Issue - State: closed - Opened by simplelifetime about 3 years ago - 2 comments

#28 - License

Issue - State: closed - Opened by zdou0830 about 3 years ago - 2 comments

#27 - Visualizations for VQA dataset

Issue - State: closed - Opened by aurooj about 3 years ago

#26 - Finetune on F30K IR/TR，why use itm

Issue - State: closed - Opened by DRACOyu about 3 years ago

#25 - Masked Patch Prediction

Issue - State: closed - Opened by Alloooshe about 3 years ago - 1 comment

#24 - itm and mlm training accuracy

Issue - State: open - Opened by junchen14 about 3 years ago - 2 comments

#23 - FLOPS calculation

Issue - State: open - Opened by junchen14 about 3 years ago - 1 comment

#22 - Do I have to download the pre-training weights of VILT or the pre-training weights of VIT

Issue - State: open - Opened by CQUTWangHong about 3 years ago - 3 comments

#21 - Time Calculation

Issue - State: open - Opened by menatallh about 3 years ago - 6 comments

#20 - AttributeError: module 'vilt' has no attribute 'modules'

Issue - State: closed - Opened by leonodelee about 3 years ago - 4 comments

#19 - Nearly 3 points below the provided VQAv2 fine-tuning result

Issue - State: closed - Opened by 4444xhc about 3 years ago - 2 comments

#18 - got error when I turn on mpp (Masked Patch Prediction)

Issue - State: closed - Opened by GewelsJI about 3 years ago - 8 comments

#17 - The time consumption for pretraining

Issue - State: closed - Opened by junchen14 over 3 years ago - 1 comment

#16 - Regarding pretraining time

Issue - State: closed - Opened by juhongm999 over 3 years ago - 2 comments

#15 - Debug model

Issue - State: open - Opened by hongzhenwang over 3 years ago - 1 comment

#14 - Question about image transformation: short edge is still 384 for the fine-tuning task?

Issue - State: closed - Opened by Jxu-Thu over 3 years ago - 2 comments

#13 - Flickr30k Finetune results does not match the provided checkpoint

Issue - State: open - Opened by JACKHAHA363 over 3 years ago - 10 comments

#12 - Unable to reproduce the 100k results

Issue - State: closed - Opened by JACKHAHA363 over 3 years ago - 8 comments

#11 - Bump pillow from 8.1.0 to 8.2.0

Pull Request - State: closed - Opened by dependabot[bot] over 3 years ago
Labels: dependencies

#10 - A very large batchsize requires 64 GPUs

Issue - State: open - Opened by Jxu-Thu over 3 years ago - 14 comments

#9 - I reproduced the code of pytorch version, but get different result

Issue - State: closed - Opened by NostalgiaOfTime over 3 years ago - 4 comments

#8 - Finetuning VQA from checkpoint has unmatched keys

Issue - State: closed - Opened by JACKHAHA363 over 3 years ago - 1 comment

#7 - Pretraining unstable

Issue - State: closed - Opened by liuhl-source over 3 years ago - 2 comments

#6 - python run.py with data_root="/arrows_flickr30k" num_gpus=1 num_nodes=1 task_finetune_irtr_f30k_randaug per_gpu_batchsize=4 load_path="vilt_200k_mlm_itm.ckpt"

Issue - State: open - Opened by jkkishore1999 over 3 years ago - 15 comments

#5 - python3 run.py with data_root=../../data/COCO num_gpus=2 num_nodes=1 task_mlm_itm whole_word_masking=True step200k per_gpu_batchsize=32

Issue - State: closed - Opened by jaeseokbyun over 3 years ago - 2 comments

#4 - python run.py with data_root=/data/workspace/dataset num_gpus=4 num_nodes=1 task_finetune_irtr_f30k_randaug per_gpu_batchsize=4 load_path="weights/vilt_200k_mlm_itm.ckpt"

GitHub / dandelin/ViLT issues and pull requests