the agent not able to type
hi, i'm running ui-tars on macos latest with 7b huggingface model, no matter what i do (i tried to reinstall it, run on 72b model) the agent is not able to type anything. there is just 'typing [text]' red label but nothing gets typed regardless of the app i'm trying to use. the clicks go through tho.
Hi, can you show the prompt you used and the original model predictions?
Hi, can you show the prompt you used and the original model predictions?
https://github.com/user-attachments/assets/211a0eba-32db-4798-965c-c12fbf0ac8d7
just basic 'hello world' in the google search box from incognito page. it just won't type regardless of the prompt or the app i'm using. i tried to quit all other apps which might cause conflicts with input but it didn't help. pretty weird behavior
Were you able to get past this? Were you able to use the model on MPS or CPU only? VLLM is only supporting CPU usage and it's unusable for that reason. Tried a few times to get it to run strictly with the HF libraries but communicating between the desktop app and the local API server wasn't simple.
Hi, You may want to check out the latest UI-TARS1.5 to see if this issue still exists. https://seed-tars.com/