4 Commits

Author	SHA1	Message	Date
Enixile	817f4169ce	BLIP2 was loading into system RAM rather than VRAM for me. I found that adding argument "device_map=device" forces it to load the entire model onto the specified device. I tested it for cuda and cpu successfully. ... GIT was failing because its processor returns a BatchEncoding and its method to() only accepts one device parameter, so I removed the second dtype parameter and it worked. I don't see how BLIP2 is different, but it works with both parameters, so I left it to avoid breaking any functionality.	2024-01-06 15:16:59 -05:00
Victor Hall	002b51f2b2	caption thing	2023-03-25 20:17:56 -04:00
Victor Hall	3744bc0dc9	minor update to caption.py	2023-03-25 12:28:49 -04:00
Victor Hall	352f5d2341	early alpha of caption script	2023-03-20 09:21:13 -04:00