skygpu
/
skynet
mirror of https://github.com/skygpu/skynet.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
120 Commits
5 Branches
10 Tags
4.5 MiB
Commit Graph

120 Commits (7e6e25465f394debd87fc4ffdab4fff9951b3913)

Author SHA1 Message Date
Guillermo Rodriguez 70ca9b11cd Need to add more cmd line params to be able to deploy 2022-12-24 11:20:25 -03:00
Guillermo Rodriguez 5652ee985a Missing dep Pillow 2022-12-24 10:55:21 -03:00
Guillermo Rodriguez 1ec908b244 Disable torch import on cuda-less runs 2022-12-24 10:54:33 -03:00
Guillermo Rodriguez b4289ba04f Add upscaler 2022-12-24 10:39:40 -03:00
Guillermo Rodriguez e5fe70d610 Drop old txt2img.py script 2022-12-21 11:59:20 -03:00
Guillermo Rodriguez b41c2babd9 Finish telegram frontend integration 
Correctly update user stats after a txt2img request
Swap tg_id to BIGINT on database
Refactor help topic system to be more generic
Fix skynet brain cli entry point
Fix multi request rpc session, we werent creating a context on the req side
Fix redo method
DGPU now returns image metadata
Improve error messages brain to frontend
Add version as constant
Add telegram dep to requirements
 2022-12-21 11:53:50 -03:00
Guillermo Rodriguez b255e03d65 Minor fixes to upscaler cli tool 2022-12-19 12:42:31 -03:00
Guillermo Rodriguez 6874282edf Add authenticated messaging, also cmd line utils txt2img and upscale 2022-12-19 12:36:02 -03:00
Guillermo Rodriguez bed5268f08 Rework dgpu client to be single task 
Add a lot of dgpu real image gen tests
Modified docker files and environment to allow for quick test relaunch without image rebuild
Rename package from skynet_bot to skynet
Drop tractor usage cause cuda is oriented to just a single proc managing gpu resources
Add ackwnoledge phase to image request for quick gpu disconnected type scenarios
Add click entry point for dgpu
Add posibility to reuse postgres_db fixture on same session by checking if schema init has been already done
 2022-12-17 11:39:42 -03:00
Guillermo Rodriguez e7c7f97218 Add TLS for encrypted comms, next HMAC 2022-12-11 13:06:07 -03:00
Guillermo Rodriguez 4224e1645c Remove unused import from tg module and unecesary docker command on db startup 2022-12-11 11:10:17 -03:00
Guillermo Rodriguez f8373f681b Started making roboust testing fixtures to init fresh db and skynet 
Add simple dgpu worker connection test
Make db connection handler manage schema and table init logic
Keep tweaking dgpu main handler attemtping to fix subactor hangs
Change frontend open rpc logic to return a wrapped rpc_call fn referencing the new socket
Decupled user config request validation from telegram module
Fix next_worker logic, now takes in account multiple tasks per dgpu
Add dgpu_workers and dgpu_next calls
Fixed readme, moved db init code into db module
 2022-12-11 11:02:55 -03:00
Guillermo Rodriguez 6247970dc8 Move gpu lifecycle tests to its own file to make torch import optional 2022-12-11 08:38:42 -03:00
Guillermo Rodriguez 20b377dd32 Add gpu worker tests to debug tractor lifecycle stuff also logging 
Minor tweak to cuda docker image
Pin tractor branch
Add triton to cuda reqs
 2022-12-11 08:32:25 -03:00
Guillermo Rodriguez 66d997c039 First decupled architecture, still working on integrating tractor gpu workers 2022-12-10 18:18:03 -03:00
Guillermo Rodriguez f06a09b5bd Add more ais, and start upscaler config and pipeline 2022-12-06 17:50:04 -03:00
Guillermo Rodriguez 7c1816b86e Add 4 new ais, donation command, unknown cmd handler 2022-12-06 09:29:41 -03:00
Guillermo Rodriguez 72149ae1b4 Add roles and start working on algo selection, clean up rr code 2022-12-05 19:36:21 -03:00
Guillermo Rodriguez 31058c116a Add round robin 2022-12-04 21:31:57 -03:00
Guillermo Rodriguez 66f21ff9de Initial commit 2022-12-04 20:39:43 -03:00