vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 01:24:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

multigpu warning only once

Browse source
This commit is contained in:
Concedo 2025-05-08 00:55:09 +08:00
parent 38b3bffcef
commit 7c5d47f688
1 changed files with 5 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

5
koboldcpp.py
View file

@ -78,6 +78,7 @@ defaultport = 5001
showsamplerwarning = True showsamplerwarning = True
showmaxctxwarning = True showmaxctxwarning = True
showusedmemwarning = True showusedmemwarning = True
showmultigpuwarning = True
session_kudos_earned = 0 session_kudos_earned = 0
session_jobs = 0 session_jobs = 0
session_starttime = None session_starttime = None
@ -950,7 +951,7 @@ def extract_modelfile_params(filepath,sdfilepath,whisperfilepath,mmprojfilepath,
modelfile_extracted_meta = None modelfile_extracted_meta = None
def autoset_gpu_layers(ctxsize, sdquanted, bbs, qkv_level): #shitty algo to determine how many layers to use def autoset_gpu_layers(ctxsize, sdquanted, bbs, qkv_level): #shitty algo to determine how many layers to use
global showusedmemwarning, modelfile_extracted_meta # reference cached values instead global showusedmemwarning, showmultigpuwarning, modelfile_extracted_meta # reference cached values instead
gpumem = MaxMemory[0] gpumem = MaxMemory[0]
usedmem = 0 usedmem = 0
if MaxFreeMemory[0]>0: if MaxFreeMemory[0]>0:
@ -973,6 +974,8 @@ def autoset_gpu_layers(ctxsize, sdquanted, bbs, qkv_level): #shitty algo to dete
if match: if match:
total_parts = int(match.group(2)) total_parts = int(match.group(2))
if total_parts > 1 and total_parts <= 9: if total_parts > 1 and total_parts <= 9:
if showmultigpuwarning:
showmultigpuwarning = False
print("Multi-Part GGUF detected. Layer estimates may not be very accurate - recommend setting layers manually.") print("Multi-Part GGUF detected. Layer estimates may not be very accurate - recommend setting layers manually.")
fsize *= total_parts fsize *= total_parts
if modelfile_extracted_meta[3] > 1024*1024*1024*5: #sdxl tax if modelfile_extracted_meta[3] > 1024*1024*1024*5: #sdxl tax