Shoufa Chen

It seems that the building process used the cached things. You need to delete your local `gradio-box` folder and clone from https://github.com/ShoufaChen/gradio-box again. Then do the building process.

It seems your building process is OK. Can you try ``` import gradio as gr def predict(inp): image = inp['image'] boxes = inp['boxes'] sub_images = [] for box in boxes:...

May I ask why there Chinese characters? is it automatically translate or you modify something?

How about this code demo: ``` import gradio as gr def predict(inp): print("hello, world") demo = gr.Interface(fn=predict, inputs=gr.Image(tool="boxes", type="pil"), outputs=gr.Gallery()) demo.launch() ``` Can you get any log in the terminal...

Hi @niiickZ , Thanks for your interest. We can reproduce this error. The current workaround is to click the `clear` (`清除`) button before uploading new images. Sorry for this inconvenience.

In case it might be helpful, provide a bash script here: ``` for IDX in $(seq -w 00 42); do echo $IDX curl "a_long_prefix_path$IDX%2Ezip" \ -H 'authority: cuhko365-my.sharepoint.com' \ .........................

@Ir1d You can get this url prefix from [here](https://docs.google.com/document/d/1krVb4B3rZw-0FaBBPS7c3SJKfqq5AVYTs2HN2LnlBPQ/edit#heading=h.2ukfzxh5c9pq). It should start with `https://cuhko365-my.sharepoint.com/` while end with `mvi%5F00%2Ezip` for `mvi_00.zip`. Then just change the `00` part in `mvi%5F00%2Ezip` with the...

Hi @gokulavasan @andrewkho , I found that current StatefulDataloader works well with `DistributedSampler` without any modifications. Would you mind please explaining why it might be inefficient? Thanks in advance.

> This currently isn't broken right? ie fast-forwarding the sampler will work, but may be inefficient. I'm OK either way for before/after release branch cut Hi @andrewkho , Does fast-forwarding...

Hello, @Con6924 This issue has been fixed in https://github.com/FoundationVision/LlamaGen/pull/23. Please have a try with main branch.