vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 09:04:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

try to use GPU for whisper

Browse source
This commit is contained in:
Concedo 2024-06-03 23:07:26 +08:00
parent a2e304ed4d
commit 967b6572a2
2 changed files with 5 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
Makefile
View file

@ -522,6 +522,8 @@ sdcpp_cublas.o: otherarch/sdcpp/sdtype_adapter.cpp otherarch/sdcpp/stable-diffus
#whisper objects #whisper objects
whispercpp_default.o: otherarch/whispercpp/whisper_adapter.cpp whispercpp_default.o: otherarch/whispercpp/whisper_adapter.cpp
$(CXX) $(CXXFLAGS) -c $< -o $@ $(CXX) $(CXXFLAGS) -c $< -o $@
whispercpp_cublas.o: otherarch/whispercpp/whisper_adapter.cpp
$(CXX) $(CXXFLAGS) $(CUBLAS_FLAGS) $(HIPFLAGS) -c $< -o $@
# idiotic "for easier compilation" # idiotic "for easier compilation"
GPTTYPE_ADAPTER = gpttype_adapter.cpp otherarch/llama_v2.cpp otherarch/llama_v3.cpp llama.cpp otherarch/utils.cpp otherarch/gptj_v1.cpp otherarch/gptj_v2.cpp otherarch/gptj_v3.cpp otherarch/gpt2_v1.cpp otherarch/gpt2_v2.cpp otherarch/gpt2_v3.cpp otherarch/rwkv_v2.cpp otherarch/rwkv_v3.cpp otherarch/neox_v2.cpp otherarch/neox_v3.cpp otherarch/mpt_v3.cpp ggml.h ggml-cuda.h llama.h otherarch/llama-util.h GPTTYPE_ADAPTER = gpttype_adapter.cpp otherarch/llama_v2.cpp otherarch/llama_v3.cpp llama.cpp otherarch/utils.cpp otherarch/gptj_v1.cpp otherarch/gptj_v2.cpp otherarch/gptj_v3.cpp otherarch/gpt2_v1.cpp otherarch/gpt2_v2.cpp otherarch/gpt2_v3.cpp otherarch/rwkv_v2.cpp otherarch/rwkv_v3.cpp otherarch/neox_v2.cpp otherarch/neox_v3.cpp otherarch/mpt_v3.cpp ggml.h ggml-cuda.h llama.h otherarch/llama-util.h
@ -607,7 +609,7 @@ koboldcpp_clblast_noavx2:
endif endif
ifdef CUBLAS_BUILD ifdef CUBLAS_BUILD
koboldcpp_cublas: ggml_v4_cublas.o ggml_v3_cublas.o ggml_v2_cublas.o ggml_v1.o expose.o gpttype_adapter_cublas.o sdcpp_cublas.o whispercpp_default.o llavaclip_cublas.o llava.o ggml-backend_cublas.o $(CUBLAS_OBJS) $(OBJS_FULL) $(OBJS) koboldcpp_cublas: ggml_v4_cublas.o ggml_v3_cublas.o ggml_v2_cublas.o ggml_v1.o expose.o gpttype_adapter_cublas.o sdcpp_cublas.o whispercpp_cublas.o llavaclip_cublas.o llava.o ggml-backend_cublas.o $(CUBLAS_OBJS) $(OBJS_FULL) $(OBJS)
$(CUBLAS_BUILD) $(CUBLAS_BUILD)
else else
koboldcpp_cublas: koboldcpp_cublas:
@ -615,7 +617,7 @@ koboldcpp_cublas:
endif endif
ifdef HIPBLAS_BUILD ifdef HIPBLAS_BUILD
koboldcpp_hipblas: ggml_v4_cublas.o ggml_v3_cublas.o ggml_v2_cublas.o ggml_v1.o expose.o gpttype_adapter_cublas.o sdcpp_cublas.o whispercpp_default.o llavaclip_cublas.o llava.o ggml-backend_cublas.o $(HIP_OBJS) $(OBJS_FULL) $(OBJS) koboldcpp_hipblas: ggml_v4_cublas.o ggml_v3_cublas.o ggml_v2_cublas.o ggml_v1.o expose.o gpttype_adapter_cublas.o sdcpp_cublas.o whispercpp_cublas.o llavaclip_cublas.o llava.o ggml-backend_cublas.o $(HIP_OBJS) $(OBJS_FULL) $(OBJS)
$(HIPBLAS_BUILD) $(HIPBLAS_BUILD)
else else
koboldcpp_hipblas: koboldcpp_hipblas:

2
colab.ipynb
View file

@ -91,7 +91,7 @@
"!echo Finding prebuilt binary for {kvers}\r\n", "!echo Finding prebuilt binary for {kvers}\r\n",
"!wget -O dlfile.tmp https://kcppcolab.concedo.workers.dev/?{kvers} && mv dlfile.tmp koboldcpp_cublas.so\r\n", "!wget -O dlfile.tmp https://kcppcolab.concedo.workers.dev/?{kvers} && mv dlfile.tmp koboldcpp_cublas.so\r\n",
"!test -f koboldcpp_cublas.so && echo Prebuilt Binary Exists || echo Prebuilt Binary Does Not Exist\r\n", "!test -f koboldcpp_cublas.so && echo Prebuilt Binary Exists || echo Prebuilt Binary Does Not Exist\r\n",
"!test -f koboldcpp_cublas.so && echo Build Skipped || make koboldcpp_cublas LLAMA_CUBLAS=1 LLAMA_COLAB=1 LLAMA_PORTABLE=1\r\n", "!test -f koboldcpp_cublas.so && echo Build Skipped || make -j2 koboldcpp_cublas LLAMA_CUBLAS=1 LLAMA_COLAB=1 LLAMA_PORTABLE=1\r\n",
"!cp koboldcpp_cublas.so koboldcpp_cublas.dat\r\n", "!cp koboldcpp_cublas.so koboldcpp_cublas.dat\r\n",
"!apt update\r\n", "!apt update\r\n",
"!apt install aria2 -y\r\n", "!apt install aria2 -y\r\n",