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

reverted the failsafe removal, since they dropped support for dll check

Browse source
This commit is contained in:
Concedo 2023-08-31 15:39:32 +08:00
parent b6914ebd04
commit e2fd30b5d1
4 changed files with 32 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

10
Makefile
View file

@ -341,15 +341,15 @@ $(info )
ggml.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) -c $< -o $@
ggml_openblas.o: ggml.c ggml.h
ggml_openblas.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
ggml_failsafe.o: ggml.c ggml.h
ggml_failsafe.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(NONECFLAGS) -c $< -o $@
ggml_noavx2.o: ggml.c ggml.h
ggml_noavx2.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(SIMPLECFLAGS) -c $< -o $@
ggml_clblast.o: ggml.c ggml.h
ggml_clblast.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
ggml_cublas.o: ggml.c ggml.h
ggml_cublas.o: ggml.c ggml.h ggml-cuda.h k_quants.h
$(CC) $(CFLAGS) $(FULLCFLAGS) $(CUBLAS_FLAGS) $(HIPFLAGS) -c $< -o $@
#quants K

15
klite.embd
View file

@ -7846,13 +7846,26 @@ Kobold Lite is under the AGPL v3.0 License for the purposes of koboldcpp. Please
}
if(foundself!=null && foundself.length>0)
{
//exception: check to see if it's actually opponent naming us and not our turn
if(localsettings.chatopponent!="" && tempfullsearchable.startsWith(localsettings.chatopponent+": "))
{
myturnchat = false;
chatunits.push({
name:localsettings.chatopponent,
msg:tempfullsearchable.split(localsettings.chatopponent+": ")[1],
myturn:myturnchat});
}
else
{
myturnchat = true;
chatunits.push({
name:foundself[0].substring(0,foundself[0].length-2),
msg:tempfullsearchable.split(foundself[0])[1],
myturn:myturnchat});
}else if(foundopponent != null && foundopponent.length > 0)
}
}
else if(foundopponent != null && foundopponent.length > 0)
{
myturnchat = false;
chatunits.push({

4
llama.cpp
View file

@ -612,6 +612,7 @@ struct llama_mmap {
throw std::runtime_error(format("MapViewOfFile failed: %s", llama_format_win_err(error).c_str()));
}
#ifndef USE_FAILSAFE
#if _WIN32_WINNT >= _WIN32_WINNT_WIN8
if (prefetch) {
// Advise the kernel to preload the mapped memory
@ -626,6 +627,9 @@ struct llama_mmap {
#else
#pragma message("warning: You are building for pre-Windows 8; prefetch not supported")
#endif // _WIN32_WINNT >= _WIN32_WINNT_WIN8
#else
printf("\nPrefetchVirtualMemory skipped in compatibility mode.\n");
#endif
}
~llama_mmap() {

4
otherarch/llama-util.h
View file

@ -271,6 +271,7 @@ struct llama_v3_mmap {
throw std::runtime_error(format_old("MapViewOfFile failed: %s", llama_v3_format_win_err(error).c_str()));
}
#ifndef USE_FAILSAFE
if (prefetch) {
// The PrefetchVirtualMemory API is only present on Windows 8 and above, so we
// will dynamically load it using GetProcAddress.
@ -294,6 +295,9 @@ struct llama_v3_mmap {
}
}
}
#else
printf("\nPrefetchVirtualMemory skipped in compatibility mode.\n");
#endif
}
~llama_v3_mmap() {