vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 09:04:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

enable flash attention for image generation (#1633)

Browse source
This commit is contained in:
Wagner Bruna 2025-07-05 00:20:51 -03:00 • committed by GitHub
parent bc3e4c1197
commit d74c16e6e0
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 8 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

2
koboldcpp.py
View file

@ -273,6 +273,7 @@ class sd_load_model_inputs(ctypes.Structure):
("vulkan_info", ctypes.c_char_p),
("threads", ctypes.c_int),
("quant", ctypes.c_int),
("flash_attention", ctypes.c_bool),
("taesd", ctypes.c_bool),
("tiled_vae_threshold", ctypes.c_int),
("t5xxl_filename", ctypes.c_char_p),
@ -1624,6 +1625,7 @@ def sd_load_model(model_filename,vae_filename,lora_filename,t5xxl_filename,clipl
inputs.threads = thds
inputs.quant = quant
inputs.flash_attention = args.flashattention
inputs.taesd = True if args.sdvaeauto else False
inputs.tiled_vae_threshold = args.sdtiledvae
inputs.vae_filename = vae_filename.encode("UTF-8")