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

class.py changes

Browse source
This commit is contained in:
Concedo 2023-09-20 21:27:49 +08:00
parent b63cf223c9
commit 712b8423f6
1 changed files with 5 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

10
class.py
View file

@ -57,7 +57,7 @@ class model_backend(InferenceModel):
self.kcpp_ropescale = 0.0 self.kcpp_ropescale = 0.0
self.kcpp_ropebase = 10000.0 self.kcpp_ropebase = 10000.0
self.kcpp_useclblast = None self.kcpp_useclblast = None
self.kcpp_useclublas = None self.kcpp_usecublas = None
self.kcpp_noblas = False self.kcpp_noblas = False
self.kcpp_noavx2 = False self.kcpp_noavx2 = False
self.kcpp_nommap = False self.kcpp_nommap = False
@ -85,7 +85,7 @@ class model_backend(InferenceModel):
}) })
requested_parameters.append({ requested_parameters.append({
"uitype": "dropdown", "uitype": "dropdown",
"unit": "text", "unit": "int",
"label": "KoboldCpp Accelerator", "label": "KoboldCpp Accelerator",
"id": "kcpp_accelerator", "id": "kcpp_accelerator",
"default": 0, "default": 0,
@ -174,7 +174,7 @@ class model_backend(InferenceModel):
}) })
requested_parameters.append({ requested_parameters.append({
"uitype": "dropdown", "uitype": "dropdown",
"unit": "text", "unit": "int",
"label": "Smart Context", "label": "Smart Context",
"id": "kcpp_smartcontext", "id": "kcpp_smartcontext",
"default": self.kcpp_smartcontext, "default": self.kcpp_smartcontext,
@ -218,7 +218,7 @@ class model_backend(InferenceModel):
elif accel==1: elif accel==1:
pass pass
elif accel==2: elif accel==2:
self.kcpp_useclublas = ["normal"] self.kcpp_usecublas = ["normal"]
elif accel==3: elif accel==3:
self.kcpp_useclblast = [0,0] self.kcpp_useclblast = [0,0]
elif accel==4: elif accel==4:
@ -250,7 +250,7 @@ class model_backend(InferenceModel):
blasbatchsize=self.kcpp_blasbatchsize, ropeconfig=[self.kcpp_ropescale, self.kcpp_ropebase], stream=False, smartcontext=self.kcpp_smartcontext, blasbatchsize=self.kcpp_blasbatchsize, ropeconfig=[self.kcpp_ropescale, self.kcpp_ropebase], stream=False, smartcontext=self.kcpp_smartcontext,
unbantokens=False, bantokens=None, usemirostat=None, forceversion=0, nommap=self.kcpp_nommap, unbantokens=False, bantokens=None, usemirostat=None, forceversion=0, nommap=self.kcpp_nommap,
usemlock=False, noavx2=self.kcpp_noavx2, debugmode=self.kcpp_debugmode, skiplauncher=True, hordeconfig=None, noblas=self.kcpp_noblas, usemlock=False, noavx2=self.kcpp_noavx2, debugmode=self.kcpp_debugmode, skiplauncher=True, hordeconfig=None, noblas=self.kcpp_noblas,
useclblast=self.kcpp_useclblast, usecublas=self.kcpp_useclublas, gpulayers=self.kcpp_gpulayers, tensor_split=None, config=None, onready='', multiuser=False) useclblast=self.kcpp_useclblast, usecublas=self.kcpp_usecublas, gpulayers=self.kcpp_gpulayers, tensor_split=None, config=None, onready='', multiuser=False)
koboldcpp.main(kcppargs,False) #initialize library without enabling Lite http server koboldcpp.main(kcppargs,False) #initialize library without enabling Lite http server
kcpp_backend_loaded = True kcpp_backend_loaded = True