google-research/big_vision

View on GitHub

Official codebase used to develop Vision Transformer, SigLIP, MLP-Mixer, LiT and more.

Stars: 3.5k
Forks: 227
Open beginner issues: 0
Indexed issues: 53

Dominant language: Jupyter Notebook
License: Apache-2.0
Last GitHub push: May 19, 2025
Latest indexed: Sep 16, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: No beginner labels indexed
PR merge metrics: No merged PRs in 30d

Open All

53 open issues indexed

Colorization uvim model not working Open

google-research/big_vision#19 · 5 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Add other custom optimizers Open

google-research/big_vision#32 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
bfloat16 Training Open

more info needed

google-research/big_vision#41 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Memory Efficient Attention integration Open

google-research/big_vision#59 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
SigLIP and canonicalize Open

google-research/big_vision#78 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Text lowering issue Open

google-research/big_vision#79 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
FlexiVit is also flexible with image resolution? Open

google-research/big_vision#86 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Mixup Per Example? Open

google-research/big_vision#90 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Clarification: SigLIP Image Transform Open

google-research/big_vision#91 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Errors in notebooks Open

google-research/big_vision#93 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Confusion on FlexiViT Open

google-research/big_vision#94 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Error with putting arrays on CPU in cloud TPUs Open

google-research/big_vision#101 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
[QUESTION] How to perform inference on trained model? Open

google-research/big_vision#106 · 8 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Implementation of contrast() seems wrong Open

google-research/big_vision#109 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Behavior of `solarize()` depends on integer overflow Open

google-research/big_vision#110 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
No ChartQA ft weights released on Kaggle Open

google-research/big_vision#111 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
PlaiGemma finetuned model Open

google-research/big_vision#113 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Loss Scale for Training Siglip Open

google-research/big_vision#115 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
[Question] How to inference / captioning short video? Open

google-research/big_vision#117 · 6 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Question about the evaluation on instance segmentation about PaliGemma Open

google-research/big_vision#118 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Could you please give me an instruction on how to use paligemma's evaluator? Open

google-research/big_vision#123 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
What's a "`rezied` method"? Open

google-research/big_vision#124 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
tokenization error when using msiglip Open

google-research/big_vision#126 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Add chunked implementation of siglip sigmoid loss Open

google-research/big_vision#130 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
GIVT (and more generally, new research by Google) on Hugging Face Open

google-research/big_vision#131 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
3B model not working even using the snippet from the Official HuggingFace page. Open

google-research/big_vision#134 · 2 comments · 1 reaction · 0 assignees · Indexed Sep 16, 2026
Fine-tune Paligemma model for segmentation task is failing Open

google-research/big_vision#136 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
What are the supported languages for the multilingual SigLIP models? Open

google-research/big_vision#138 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Pretraining Code for PaliGemma 2 Open

google-research/big_vision#140 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Explanation of “resample_patchemb” function in flexiViT Open

google-research/big_vision#142 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
siglip's [cls token] Open

google-research/big_vision#145 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
V2k Open

google-research/big_vision#146 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
How can edit the paligemma funetuning scripts to plot a validation loss curve and also convert my model to hf model Open

google-research/big_vision#147 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Add a pyproject.toml Open

google-research/big_vision#150 · 0 comments · 2 reactions · 0 assignees · Indexed Sep 16, 2026
max_length of Siglip2 Open

google-research/big_vision#157 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Siglip2 Paper: Why recall is different "text to image retrieval" between "image to text retrieval" in Figure2? Open

google-research/big_vision#158 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
module 'jax.api_util' has no attribute 'debug_info' Open

google-research/big_vision#160 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
could you release the train code of siglip2? Open

google-research/big_vision#161 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Problem with Reproducing PaliGemma Results Open

google-research/big_vision#164 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
cannot achieve the claimed zero-shot retrieval results of SigLIP2 ViT/B16-224 on COCO and Flickr30K in your paper Open

google-research/big_vision#166 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
A simple question about Image Resolution of NaFlex version Open

google-research/big_vision#168 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
siglip2 training code Open

google-research/big_vision#169 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
The ratio of reduced memory with sigmoid loss Open

google-research/big_vision#170 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Open-Vocabulary semantic segmentation Open

google-research/big_vision#171 · 8 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
if I want to use my image-text data to finetune the SigLIP2， where I can get the train code? Open

google-research/big_vision#172 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
siglip or siglip2 image process Open

google-research/big_vision#173 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Installation Open

google-research/big_vision#175 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Have you provide the code for finetuning Siglip2 using all the loss (captioning, dense captioning, self-distillation, masked prediciton) as mentioned in the original paper? Open

google-research/big_vision#178 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 16, 2026
Why the loss is divided by n, not n^2? Open

google-research/big_vision#179 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026
On the effectiveness of the NaFlex variant for high-resolution images Open

google-research/big_vision#181 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 16, 2026

google-research/big_vision

Get new issues in your inbox