vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-09 16:44:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

experimental swa flag

Browse source
This commit is contained in:
Concedo 2025-05-23 21:33:26 +08:00
parent 499283c63a
commit c4df151298
3 changed files with 11 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

1
expose.h
View file

@ -70,6 +70,7 @@ struct load_model_inputs
const int quant_k = 0;
const int quant_v = 0;
const bool check_slowness = false;
const bool swa_support = false;
const bool quiet = false;
const int debugmode = 0;
};

6
gpttype_adapter.cpp
View file

@ -1927,7 +1927,11 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
kcpp_data->use_smartcontext = inputs.use_smartcontext;
kcpp_data->use_contextshift = inputs.use_contextshift;
kcpp_data->use_fastforward = inputs.use_fastforward;
kcpp_data->swa_full = (inputs.use_fastforward || inputs.use_contextshift)?true:false;
kcpp_data->swa_full = !inputs.swa_support;//(inputs.use_fastforward || inputs.use_contextshift)?true:false;
if(!kcpp_data->swa_full)
{
printf("\n!!!!!!!!!!!!!!!!!!!\nExperimental FLAG - SWA SUPPORT IS ENABLED!\n!!!!!!!!!!!!!!!!!!!\n");
}
debugmode = inputs.debugmode;
draft_ctx = nullptr;
guidance_ctx = nullptr;

5
koboldcpp.py
View file

@ -192,6 +192,7 @@ class load_model_inputs(ctypes.Structure):
("quant_k", ctypes.c_int),
("quant_v", ctypes.c_int),
("check_slowness", ctypes.c_bool),
("swa_support", ctypes.c_bool),
("quiet", ctypes.c_bool),
("debugmode", ctypes.c_int)]
@ -1248,6 +1249,7 @@ def load_model(model_filename):
inputs.override_kv = args.overridekv.encode("UTF-8") if args.overridekv else "".encode("UTF-8")
inputs.override_tensors = args.overridetensors.encode("UTF-8") if args.overridetensors else "".encode("UTF-8")
inputs.check_slowness = (not args.highpriority and os.name == 'nt' and 'Intel' in platform.processor())
inputs.swa_support = args.experiment_swa
inputs = set_backend_props(inputs)
ret = handle.load_model(inputs)
return ret
@ -6907,6 +6909,9 @@ if __name__ == '__main__':
admingroup.add_argument("--adminpassword", metavar=('[password]'), help="Require a password to access admin functions. You are strongly advised to use one for publically accessible instances!", default=None)
admingroup.add_argument("--admindir", metavar=('[directory]'), help="Specify a directory to look for .kcpps configs in, which can be used to swap models.", default="")
experimentgroup = parser.add_argument_group('Experimental Commands, can change or break any time!')
experimentgroup.add_argument("--experiment_swa", help="Enables SWA mode. There are no safety checks.", action='store_true')
deprecatedgroup = parser.add_argument_group('Deprecated Commands, DO NOT USE!')
deprecatedgroup.add_argument("--hordeconfig", help=argparse.SUPPRESS, nargs='+')
deprecatedgroup.add_argument("--sdconfig", help=argparse.SUPPRESS, nargs='+')