facebookresearch/ijepa
Official codebase for I-JEPA, the Image-based Joint-Embedding Predictive Architecture. First outlined in the CVPR paper, "Self-supervised learning from images with a joint-embedding predictive architecture."
PythonNOASSERTION
Issues
- 1
- 4
Please Open Source ViT-B/16
#15 opened by sybapp - 2
Question about dinov2 vs ijepa
#66 opened by lanalex - 13
Training from scratch
#55 opened by Ugenteraan - 0
Why normalize twice since the target encoder also normalize the feature at last?
#67 opened by jianghaojun - 5
Struggling to Train Downstream Classifier
#58 opened by NolanGC - 0
- 5
Suggestions to train on low-end GPU
#62 opened by ravioliasb - 0
generative model decoder for target-encoder
#56 opened by sls619 - 4
Why is there no class token?
#61 opened by swarajnanda2021 - 2
- 4
Training loss increases
#41 opened by ChristianEschen - 0
Why there is no **unscale_** when you use amp?
#64 opened by jsrdcht - 5
Downstream task
#51 opened by ankan8145 - 0
Did you ever train with the ViTC?
#63 opened by cvg25 - 1
- 0
Bias in Multiblock Mask Collator
#60 opened by joeagriffith - 14
Struggling to replicate evaluation results
#47 opened by finbarrtimbers - 1
- 6
How to load ijepa checkpoints?
#50 opened by namrahrehman - 3
config files about vit-small and vit-base
#39 opened by stonewjf - 2
- 1
- 1
imagenet1k Huggingface extraction
#54 opened by AasherH - 7
Why doesn't the model collapse?
#25 opened by Ber666 - 0
- 0
Is there any visual code?
#45 opened by xiangtaowong - 3
I Just download the image1k dataset from huggingface, what should i do next to process the tar.gz file and train this model
#44 opened by james2v - 2
Linear probing
#36 opened by zankerx - 2
RuntimeError: Default process group has not been initialized, please make sure to call init_process_group.
#43 opened by alkdischen - 0
- 2
All blocks have same size within a batch
#38 opened by aRI0U - 4
Using Pretrained Model
#31 opened by Giridhar9652 - 1
Torch version
#37 opened by KyeonghaRho - 4
is batch size per gpu important to re-implement the accuracy reported in the paper?
#35 opened by Gus-Guo - 1
Some questions about context and target
#32 opened by lezhang7 - 2
- 1
Making the code a bit more installable
#33 opened by lanalex - 1
- 4
Difference between I-JEPA and data2vec2.0?
#14 opened by rayleizhu - 1
Is the VIT predictor important?
#29 opened by ItakeLs - 1
minimum ijepa, 100+ lines (except vit(<100 lines) and mask(can be <100 lines))
#26 opened by yuedajiong - 3
Automatic1111 Integration
#23 opened by Ashdevz - 1
Typo in Paper
#27 opened by hu-po - 2
- 2
- 2
Longer training schedule?
#17 opened by JUGGHM - 1
Capitalization correction in README.md
#19 opened by aroramrinaal - 1
Will The sketche decoder release?
#13 opened by wangxiaochun520 - 2