search

tobran / GALIP

[CVPR2023] A faster, smaller, and better text-to-image model for large-scale training
MIT License
225 stars 25 forks source link

Configuration Modification！ #2

Open nevoliu opened 1 year ago

nevoliu commented 1 year ago

Hi MingTao, Great of your work！My server configuration has only two 3090 GPUs, how do I modify the script/train.sh file? thanks.

nevoliu commented 1 year ago

When I was troubleshooting the problem, I found that the dataset was not loading. Further I found out that the reason was that the dataset did not have ../data/coco/text/COCO_train2014_000000106379.txt file, can you provide some information about this file? I look forward to your reply. 111

tobran commented 1 year ago

Hi, you can modify the set CUDA_VISIBLE_DEVICES=0,1 and nodes=2 in train.sh for your training. The data error looks like there is no prepared dataset under ./data. You can download it according to the readme.

nevoliu commented 1 year ago

Hi, you can modify the set CUDA_VISIBLE_DEVICES=0,1 and nodes=2 in train.sh for your training. The data error looks like there is no prepared dataset under ./data. You can download it according to the readme.

Thank you very much for your reply. Based on your answer, I solved the current problem. I am a new graduate student and I just came across the text to image task. I couldn't find . /cfg/coco.yml inside . /data/coco/npz/coco_val256_FIDK0.npz file, can you tell me where I can download this file and the information about it?

tobran commented 1 year ago

Hello, the coco.yml is under ./code/cfg/ folder and the npz file has been uploaded to Google Drive, you can find the link (https://drive.google.com/file/d/15Fw-gErCEArOFykW3YTnLKpRcPgI_3AB/view?usp=sharing) in the readme.

dipth11 commented 1 year ago

Hi, you can modify the set CUDA_VISIBLE_DEVICES=0,1 and nodes=2 in train.

Hi, you can modify the set CUDA_VISIBLE_DEVICES=0,1 and nodes=2 in train.sh for your training. The data error looks like there is no prepared dataset under ./data. You can download it according to the readme.

Hello, I encountered this error: RuntimeError: Default process group has not been initialized, please make sure to call init_process_group. I dont know how to solve it.