Fix hordeconfig max context setting, and add Makefile flags for cuda F16/KQuants per iter. (LostRuins#252)

* Fix hordeconfig maxcontext setting.

* cuda: Bring DMMV_F16 and KQUANTS_ITER Makefile flags over from llama.

Author: ycros

Makefile

@@ -149,6 +149,14 @@ ifdef LLAMA_CUDA_DMMV_Y
 else
 	NVCCFLAGS += -DGGML_CUDA_DMMV_Y=1
 endif # LLAMA_CUDA_DMMV_Y
+ifdef LLAMA_CUDA_DMMV_F16
+	NVCCFLAGS += -DGGML_CUDA_DMMV_F16
+endif # LLAMA_CUDA_DMMV_F16
+ifdef LLAMA_CUDA_KQUANTS_ITER
+	NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
+else
+	NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2
+endif
 ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
 	$(NVCC) $(NVCCFLAGS) $(CXXFLAGS) $(CUBLAS_CXXFLAGS) -Wno-pedantic -c $< -o $@
 ggml_v2-cuda.o: otherarch/ggml_v2-cuda.cu otherarch/ggml_v2-cuda.h

koboldcpp.py

@@ -724,7 +724,7 @@ def main(args):
             sys.exit(2)
 
     if args.hordeconfig and args.hordeconfig[0]!="":
-        global friendlymodelname, maxhordelen, showdebug
+        global friendlymodelname, maxhordelen, maxhordectx, showdebug
         friendlymodelname = "koboldcpp/"+args.hordeconfig[0]
         if len(args.hordeconfig) > 1:
             maxhordelen = int(args.hordeconfig[1])