vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

cherrypicked the Hipblas fixed from PR #571

Browse source
This commit is contained in:
Concedo 2023-12-22 21:29:20 +08:00
parent 77463e0e9c
commit 852ca780c9
1 changed files with 11 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

16
koboldcpp.py
View file

@ -408,6 +408,7 @@ totalgens = 0
currentusergenkey = "" #store a special key so polled streaming works even in multiuser currentusergenkey = "" #store a special key so polled streaming works even in multiuser
args = None #global args args = None #global args
gui_layers_untouched = True gui_layers_untouched = True
runmode_untouched = True
preloaded_story = None preloaded_story = None
sslvalid = False sslvalid = False
@ -1288,12 +1289,13 @@ def show_new_gui():
MaxMemory[0] = max(int(FetchedCUdeviceMem[idx])*1024*1024,MaxMemory[0]) MaxMemory[0] = max(int(FetchedCUdeviceMem[idx])*1024*1024,MaxMemory[0])
#autopick cublas if suitable, requires at least 3.5GB VRAM to auto pick #autopick cublas if suitable, requires at least 3.5GB VRAM to auto pick
global exitcounter global exitcounter, runmode_untouched
if exitcounter < 100 and MaxMemory[0]>3500000000 and CUDevicesNames[0]!="" and ("Use CuBLAS" in runopts or "Use hipBLAS (ROCM)" in runopts) and runopts_var.get()=="Use OpenBLAS": #we do not want to autoselect hip/cublas if the user has already changed their desired backend!
if exitcounter < 100 and MaxMemory[0]>3500000000 and (("Use CuBLAS" in runopts and CUDevicesNames[0]!="") or "Use hipBLAS (ROCm)" in runopts) and (any(CUDevicesNames) or any(CLDevicesNames)) and runmode_untouched:
if "Use CuBLAS" in runopts: if "Use CuBLAS" in runopts:
runopts_var.set("Use CuBLAS") runopts_var.set("Use CuBLAS")
elif "Use hipBLAS (ROCM)" in runopts: elif "Use hipBLAS (ROCm)" in runopts:
runopts_var.set("Use hipBLAS (ROCM)") runopts_var.set("Use hipBLAS (ROCm)")
changed_gpu_choice_var() changed_gpu_choice_var()
return return
@ -1394,6 +1396,8 @@ def show_new_gui():
gpulayers_var.trace("w", changed_gpulayers) gpulayers_var.trace("w", changed_gpulayers)
def changerunmode(a,b,c): def changerunmode(a,b,c):
global runmode_untouched
runmode_untouched = False
index = runopts_var.get() index = runopts_var.get()
if index == "Use CLBlast" or index == "CLBlast NoAVX2 (Old CPU)" or index == "Use CuBLAS" or index == "Use hipBLAS (ROCm)": if index == "Use CLBlast" or index == "CLBlast NoAVX2 (Old CPU)" or index == "Use CuBLAS" or index == "Use hipBLAS (ROCm)":
quick_gpuname_label.grid(row=3, column=1, padx=75, sticky="W") quick_gpuname_label.grid(row=3, column=1, padx=75, sticky="W")
@ -1525,6 +1529,8 @@ def show_new_gui():
runopts_var.trace('w', changerunmode) runopts_var.trace('w', changerunmode)
changerunmode(1,1,1) changerunmode(1,1,1)
global runmode_untouched
runmode_untouched = True
# Tokens Tab # Tokens Tab
tokens_tab = tabcontent["Tokens"] tokens_tab = tabcontent["Tokens"]
@ -1704,7 +1710,7 @@ def show_new_gui():
if cublas_option: if cublas_option:
runopts_var.set(cublas_option) runopts_var.set(cublas_option)
elif hipblas_option: elif hipblas_option:
runopts_var.set(cublas_option) runopts_var.set(hipblas_option)
lowvram_var.set(1 if "lowvram" in dict["usecublas"] else 0) lowvram_var.set(1 if "lowvram" in dict["usecublas"] else 0)
mmq_var.set(1 if "mmq" in dict["usecublas"] else 0) mmq_var.set(1 if "mmq" in dict["usecublas"] else 0)
gpu_choice_var.set("All") gpu_choice_var.set("All")