-
Notifications
You must be signed in to change notification settings - Fork 1.4k
Description
Version
v0.2.2
Model
UI-TARS-7B-DPO
Deployment Method
Local
Issue Description
Network is stable and the model service is running properly. However, after executing the first action (opening the browser), when waiting to perform the second action, this connection issue occurs. The first action executes quickly, but the second action keeps waiting until it errors out. No request information was received by the vLLM backend.
Error Logs
INVOKE_RETRY_ERROR: Too many model invoke failures: Connection error.
Error: Connection error.
at OpenAI2.makeRequest (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:72643:13)
at async UITarsModel.invokeModelProvider (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:73361:20)
at async UITarsModel.invoke (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:73383:20)
at async async_retry.retries._retry_model (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:73597:28)
at async GUIAgent.run (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:73595:85)
at async runAgent (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:156716:3)
at async Object.handle (D:\models\UI-TARS-desktop-main(UI-TARS-1.5-7B-dpo)\apps\ui-tars\dist\main\main.js:156761:5)
at async WebContents. (node:electron/js2c/browser_init:2:87023)