-
Notifications
You must be signed in to change notification settings - Fork 2.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update Queue API, documentation #2026
Conversation
Small nits, but overall LGTM! Very clean |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Documentation and naming for queue function is miles ahead of mine congrats 🎉
The only point I could make is since we change duration history total and count and calculation why don't we take the average or calculate estimations function index by function index, there was an issue around it and could be great if you want to handle it right now which would make our etas more precise.
By the way if we don't use last x number of durations for calculating ETA the calculation could get ugly when the queue size gets very big because when the queue size is Big there's a lot of communication going on which will slow the processes of events. |
Let's bring in function specific ETA's later. |
I don't understand why things would be any different with "last x" calculation? |
* queue-refactor-backend (#1489) * queue-refactor-backend - create a template for the new design * queue-refactor-backend - clean after the old queue * queue-refactor-backend - add basic test to websocket endpoint * queue-refactor-backend - small fix * queue-refactor-backend - debugs&fixes&finalizations - test the flow with postman * queue-refactor-backend - tweaks on websocket closing * queue-refactor-backend - cleanup * queue-refactor-backend - cleanup & tweaks * queue-refactor-backend - cleanup & tweaks * queue-refactor-backend - cleanup & tweaks - correct the exception handling * queue-refactor-backend - add websockets dependency * queue-refactor-backend - reformat * queue-refactor-backend - add single event test * queue-refactor-backend - tweaks - remove outdated tests * queue-refactor-backend - reformat * queue-refactor-backend - reformat * queue-refactor-backend - reformat * queue-refactor-backend - add Queue configurations to Blocks.launch() - add live_queue_update to send estimations whenever a job gets fetched from the Queue * queue-refactor-backend - add Queue configurations to Blocks.launch() - add live_queue_update to send estimations whenever a job gets fetched from the Queue * queue-refactor-backend - tweaks * queue-refactor-backend - make SLEEP_WHEN_FREE shorter Co-authored-by: Ali Abid <aabid94@gmail.com> * Add estimation parameters to queue (#1889) * - tweaks on Estimation * version * Revert "version" This reverts commit bd1f4d7. * some fix and tweaks * implement queue frontend (#1950) * implement queue frontend * fix types * fix ws endpoint in build mode * cleanup * Queue tweaks (#1909) * tweaks on estimation payload * Queue keep ws connections open (#1910) * 1. keep ws connections open after the event process is completed 2. do not send estimations periodically if live queue updates is open * fix calculation * 1. tweaks on event_queue * fix issue - create new ws for each request * format * fix * fix tests * fix tests * tets * test * changes * changes * changes * change' * wtf * changes * changes * file perms * Release queue beta v1 (#1971) * - release the new queue * - bypass the issue in the tests - rewrite the lost part in the codebase * - add concurrent queue example (#1978) * rank_eta calc * Queue fixes (#1981) * change * format * - comment out queue tests as they dont work well * - reformat * Update gradio/event_queue.py Co-authored-by: Ömer Faruk Özdemir <farukozderim@gmail.com> * changes * changes * change * weird fix Co-authored-by: Ömer Faruk Özdemir <farukozderim@gmail.com> * release-queue-v3 (#1988) * Fix frontend queuing to target secure WSS (#1996) * change * format * changes * queue-concurrency-tweaks (#2002) 1. make gather_data and broadcast_estimation sequential instead of concurrent because they were deleting elements at the same time and raising expections which was lowering the performance * Update Queue API, documentation (#2026) * changes * changes * fixes * changes * change * fix Co-authored-by: Ömer Faruk Özdemir <farukozderim@gmail.com> Co-authored-by: pngwn <hello@pngwn.io>
Made Queue API a little clearer, added docstrings and documentation to site. Remove print statements. Improved estimation logic (no need to store duration history as a list)