removed openblas backend, merged into CPU (with llamafile for BLAS). …

…GPU backend is now automatically selected when running from CLI unless noblas is specified.

.gitignore

@@ -107,7 +107,6 @@ tests/test-tokenizer-1-bpe
 
 /koboldcpp_default.so
 /koboldcpp_failsafe.so
-/koboldcpp_openblas.so
 /koboldcpp_noavx2.so
 /koboldcpp_clblast.so
 /koboldcpp_clblast_noavx2.so
@@ -116,7 +115,6 @@ tests/test-tokenizer-1-bpe
 /koboldcpp_vulkan_noavx2.so
 /koboldcpp_default.dll
 /koboldcpp_failsafe.dll
-/koboldcpp_openblas.dll
 /koboldcpp_noavx2.dll
 /koboldcpp_clblast.dll
 /koboldcpp_clblast_noavx2.dll

Makefile

@@ -1,9 +1,9 @@
 # Add custom options to Makefile.local rather than editing this file.
 -include $(abspath $(lastword ${MAKEFILE_LIST})).local
 
-default: koboldcpp_default koboldcpp_failsafe koboldcpp_openblas koboldcpp_noavx2 koboldcpp_clblast koboldcpp_clblast_noavx2 koboldcpp_cublas koboldcpp_hipblas koboldcpp_vulkan koboldcpp_vulkan_noavx2
+default: koboldcpp_default koboldcpp_failsafe koboldcpp_noavx2 koboldcpp_clblast koboldcpp_clblast_noavx2 koboldcpp_cublas koboldcpp_hipblas koboldcpp_vulkan koboldcpp_vulkan_noavx2
 tools: quantize_gpt2 quantize_gptj quantize_gguf quantize_neox quantize_mpt quantize_clip whispermain sdmain gguf-split
-dev: koboldcpp_openblas
+dev: koboldcpp_default
 dev2: koboldcpp_clblast
 dev3: koboldcpp_vulkan
 
@@ -75,7 +75,6 @@ SIMPLECFLAGS =
 FULLCFLAGS =
 NONECFLAGS =
 
-OPENBLAS_FLAGS = -DGGML_USE_OPENBLAS -DGGML_USE_BLAS -I/usr/local/include/openblas
 CLBLAST_FLAGS = -DGGML_USE_CLBLAST
 FAILSAFE_FLAGS = -DUSE_FAILSAFE
 VULKAN_FLAGS = -DGGML_USE_VULKAN -DSD_USE_VULKAN
@@ -336,7 +335,6 @@ endif
 
 DEFAULT_BUILD =
 FAILSAFE_BUILD =
-OPENBLAS_BUILD =
 NOAVX2_BUILD =
 CLBLAST_BUILD =
 CUBLAS_BUILD =
@@ -346,7 +344,6 @@ VULKAN_BUILD =
 ifeq ($(OS),Windows_NT)
 	DEFAULT_BUILD = $(CXX) $(CXXFLAGS)  $^ -shared -o $@.dll $(LDFLAGS)
 	FAILSAFE_BUILD = $(CXX) $(CXXFLAGS) $^ -shared -o $@.dll $(LDFLAGS)
-	OPENBLAS_BUILD = $(CXX) $(CXXFLAGS) $^ lib/libopenblas.lib -shared -o $@.dll $(LDFLAGS)
 	NOAVX2_BUILD = $(CXX) $(CXXFLAGS) $^ -shared -o $@.dll $(LDFLAGS)
 	CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ lib/OpenCL.lib lib/clblast.lib -shared -o $@.dll $(LDFLAGS)
 	VULKAN_BUILD = $(CXX) $(CXXFLAGS) $^ lib/vulkan-1.lib -shared -o $@.dll $(LDFLAGS)
@@ -364,14 +361,11 @@ else
 	NOAVX2_BUILD = $(CXX) $(CXXFLAGS)  $^ -shared -o $@.so $(LDFLAGS)
 	endif
 
-	ifdef LLAMA_OPENBLAS
-	OPENBLAS_BUILD = $(CXX) $(CXXFLAGS) $^ $(ARCH_ADD) -lopenblas -shared -o $@.so $(LDFLAGS)
-	endif
 	ifdef LLAMA_CLBLAST
 		ifeq ($(UNAME_S),Darwin)
-			CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ -lclblast -framework OpenCL $(ARCH_ADD) -lopenblas -shared -o $@.so $(LDFLAGS)
+			CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ -lclblast -framework OpenCL $(ARCH_ADD) -shared -o $@.so $(LDFLAGS)
 		else
-			CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ -lclblast -lOpenCL $(ARCH_ADD) -lopenblas -shared -o $@.so $(LDFLAGS)
+			CLBLAST_BUILD = $(CXX) $(CXXFLAGS) $^ -lclblast -lOpenCL $(ARCH_ADD) -shared -o $@.so $(LDFLAGS)
 		endif
 	endif
 	ifdef LLAMA_CUBLAS
@@ -384,13 +378,11 @@ else
 		VULKAN_BUILD = $(CXX) $(CXXFLAGS) $^ -lvulkan -shared -o $@.so $(LDFLAGS)
 	endif
 
-	ifndef LLAMA_OPENBLAS
 	ifndef LLAMA_CLBLAST
 	ifndef LLAMA_CUBLAS
 	ifndef LLAMA_HIPBLAS
 	ifndef LLAMA_VULKAN
-	OPENBLAS_BUILD = @echo 'Your OS $(OS) does not appear to be Windows. For faster speeds, install and link a BLAS library. Set LLAMA_OPENBLAS=1 to compile with OpenBLAS support or LLAMA_CLBLAST=1 to compile with ClBlast support. This is just a reminder, not an error.'
-	endif
+	VULKAN_BUILD = @echo 'Your OS $(OS) does not appear to be Windows. For faster speeds, install and link a BLAS library. Set LLAMA_VULKAN=1 to compile with Vulkan support. This is just a reminder, not an error.'
 	endif
 	endif
 	endif
@@ -421,8 +413,6 @@ $(info )
 
 ggml.o: ggml/src/ggml.c ggml/include/ggml.h
 	$(CC)  $(FASTCFLAGS) $(FULLCFLAGS) -c $< -o $@
-ggml_v4_openblas.o: ggml/src/ggml.c ggml/include/ggml.h
-	$(CC)  $(FASTCFLAGS) $(FULLCFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
 ggml_v4_failsafe.o: ggml/src/ggml.c ggml/include/ggml.h
 	$(CC)  $(FASTCFLAGS) $(NONECFLAGS) -c $< -o $@
 ggml_v4_noavx2.o: ggml/src/ggml.c ggml/include/ggml.h
@@ -480,15 +470,13 @@ llavaclip_cublas.o: examples/llava/clip.cpp examples/llava/clip.h
 llavaclip_vulkan.o: examples/llava/clip.cpp examples/llava/clip.h
 	$(CXX) $(CXXFLAGS) $(VULKAN_FLAGS) -c $< -o $@
 
-#this is only used for openblas and accelerate
+#this is only used for accelerate
 ggml-blas.o: ggml/src/ggml-blas.cpp ggml/include/ggml-blas.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@
 
 #version 3 libs
 ggml_v3.o: otherarch/ggml_v3.c otherarch/ggml_v3.h
 	$(CC)  $(FASTCFLAGS) $(FULLCFLAGS) -c $< -o $@
-ggml_v3_openblas.o: otherarch/ggml_v3.c otherarch/ggml_v3.h
-	$(CC)  $(FASTCFLAGS) $(FULLCFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
 ggml_v3_failsafe.o: otherarch/ggml_v3.c otherarch/ggml_v3.h
 	$(CC)  $(FASTCFLAGS) $(NONECFLAGS) -c $< -o $@
 ggml_v3_noavx2.o: otherarch/ggml_v3.c otherarch/ggml_v3.h
@@ -503,8 +491,6 @@ ggml_v3_clblast_noavx2.o: otherarch/ggml_v3.c otherarch/ggml_v3.h
 #version 2 libs
 ggml_v2.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
 	$(CC)  $(FASTCFLAGS) $(FULLCFLAGS) -c $< -o $@
-ggml_v2_openblas.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
-	$(CC)  $(FASTCFLAGS) $(FULLCFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
 ggml_v2_failsafe.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
 	$(CC)  $(FASTCFLAGS) $(NONECFLAGS) -c $< -o $@
 ggml_v2_noavx2.o: otherarch/ggml_v2.c otherarch/ggml_v2.h
@@ -569,8 +555,6 @@ gpttype_adapter_failsafe.o: $(GPTTYPE_ADAPTER)
 	$(CXX) $(CXXFLAGS) $(FAILSAFE_FLAGS) -c $< -o $@
 gpttype_adapter.o: $(GPTTYPE_ADAPTER)
 	$(CXX) $(CXXFLAGS) -c $< -o $@
-gpttype_adapter_openblas.o: $(GPTTYPE_ADAPTER)
-	$(CXX) $(CXXFLAGS) $(OPENBLAS_FLAGS) -c $< -o $@
 gpttype_adapter_clblast.o: $(GPTTYPE_ADAPTER)
 	$(CXX) $(CXXFLAGS) $(CLBLAST_FLAGS) -c $< -o $@
 gpttype_adapter_cublas.o: $(GPTTYPE_ADAPTER)
@@ -583,7 +567,7 @@ gpttype_adapter_vulkan_noavx2.o: $(GPTTYPE_ADAPTER)
 	$(CXX) $(CXXFLAGS) $(FAILSAFE_FLAGS) $(VULKAN_FLAGS) -c $< -o $@
 
 clean:
-	rm -vf *.o main sdmain whispermain quantize_gguf quantize_clip quantize_gpt2 quantize_gptj quantize_neox quantize_mpt quantize-stats perplexity embedding benchmark-matmult save-load-state gguf imatrix vulkan-shaders-gen gguf-split gguf-split.exe vulkan-shaders-gen.exe imatrix.exe gguf.exe main.exe sdmain.exe whispermain.exe quantize_clip.exe quantize_gguf.exe quantize_gptj.exe quantize_gpt2.exe quantize_neox.exe quantize_mpt.exe koboldcpp_default.dll koboldcpp_openblas.dll koboldcpp_failsafe.dll koboldcpp_noavx2.dll koboldcpp_clblast.dll koboldcpp_clblast_noavx2.dll koboldcpp_cublas.dll koboldcpp_hipblas.dll koboldcpp_vulkan.dll koboldcpp_vulkan_noavx2.dll koboldcpp_default.so koboldcpp_openblas.so koboldcpp_failsafe.so koboldcpp_noavx2.so koboldcpp_clblast.so koboldcpp_clblast_noavx2.so koboldcpp_cublas.so koboldcpp_hipblas.so koboldcpp_vulkan.so koboldcpp_vulkan_noavx2.so
+	rm -vf *.o main sdmain whispermain quantize_gguf quantize_clip quantize_gpt2 quantize_gptj quantize_neox quantize_mpt quantize-stats perplexity embedding benchmark-matmult save-load-state gguf imatrix vulkan-shaders-gen gguf-split gguf-split.exe vulkan-shaders-gen.exe imatrix.exe gguf.exe main.exe sdmain.exe whispermain.exe quantize_clip.exe quantize_gguf.exe quantize_gptj.exe quantize_gpt2.exe quantize_neox.exe quantize_mpt.exe koboldcpp_default.dll koboldcpp_failsafe.dll koboldcpp_noavx2.dll koboldcpp_clblast.dll koboldcpp_clblast_noavx2.dll koboldcpp_cublas.dll koboldcpp_hipblas.dll koboldcpp_vulkan.dll koboldcpp_vulkan_noavx2.dll koboldcpp_default.so koboldcpp_failsafe.so koboldcpp_noavx2.so koboldcpp_clblast.so koboldcpp_clblast_noavx2.so koboldcpp_cublas.so koboldcpp_hipblas.so koboldcpp_vulkan.so koboldcpp_vulkan_noavx2.so
 	rm -vrf ggml/src/ggml-cuda/*.o
 	rm -vrf ggml/src/ggml-cuda/template-instances/*.o
 
@@ -612,14 +596,6 @@ vulkan-shaders-gen: ggml/src/vulkan-shaders/vulkan-shaders-gen.cpp
 koboldcpp_default: ggml.o ggml_v3.o ggml_v2.o ggml_v1.o expose.o gpttype_adapter.o sdcpp_default.o whispercpp_default.o llavaclip_default.o llava.o ggml-backend_default.o $(OBJS_FULL) $(OBJS)
 	$(DEFAULT_BUILD)
 
-ifdef OPENBLAS_BUILD
-koboldcpp_openblas: ggml_v4_openblas.o ggml_v3_openblas.o ggml_v2_openblas.o ggml_v1.o expose.o gpttype_adapter_openblas.o sdcpp_default.o whispercpp_default.o llavaclip_default.o llava.o ggml-backend_default.o ggml-blas.o $(OBJS_FULL) $(OBJS)
-	$(OPENBLAS_BUILD)
-else
-koboldcpp_openblas:
-	$(DONOTHING)
-endif
-
 ifdef FAILSAFE_BUILD
 koboldcpp_failsafe: ggml_v4_failsafe.o ggml_v3_failsafe.o ggml_v2_failsafe.o ggml_v1_failsafe.o expose.o gpttype_adapter_failsafe.o sdcpp_default.o whispercpp_default.o llavaclip_default.o llava.o ggml-backend_default.o $(OBJS_FAILSAFE) $(OBJS)
 	$(FAILSAFE_BUILD)

README.md

@@ -15,7 +15,7 @@ KoboldCpp is an easy-to-use AI text-generation software for GGML and GGUF models
 - You can also run it using the command line. For info, please check `koboldcpp.exe --help`
 
 ## Linux Usage (Precompiled Binary, Recommended)
-On modern Linux systems, you should download the `koboldcpp-linux-x64-cuda1150` prebuilt PyInstaller binary on the **[releases page](https://github.com/LostRuins/koboldcpp/releases/latest)**. Simply download and run the binary.
+On modern Linux systems, you should download the `koboldcpp-linux-x64-cuda1150` prebuilt PyInstaller binary on the **[releases page](https://github.com/LostRuins/koboldcpp/releases/latest)**. Simply download and run the binary (You may have to `chmod +x` it first).
 
 Alternatively, you can also install koboldcpp to the current directory by running the following terminal command:
 ```
@@ -24,7 +24,8 @@ curl -fLo koboldcpp https://github.com/LostRuins/koboldcpp/releases/latest/downl
 After running this command you can launch Koboldcpp from the current directory using `./koboldcpp` in the terminal (for CLI usage, run with `--help`).
 
 ## MacOS (Precompiled Binary)
-- PyInstaller binaries for Modern ARM64 MacOS (M1, M2, M3) are now available! **[Simply download and run the MacOS binary](https://github.com/LostRuins/koboldcpp/releases/latest)**
+- PyInstaller binaries for Modern ARM64 MacOS (M1, M2, M3) are now available! **[Simply download the MacOS binary](https://github.com/LostRuins/koboldcpp/releases/latest)**
+- In a MacOS terminal window, set the file to executable `chmod +x koboldcpp-mac-arm64` and run it with `./koboldcpp-mac-arm64`.
 - Alternatively, or for older x86 MacOS computers, you can clone the repo and compile from source code, see Compiling for MacOS below.
 
 ## Run on Colab
@@ -70,13 +71,13 @@ when you can't use the precompiled binary directly, we provide an automated buil
 ### Compiling on Linux (Manual Method)
 - To compile your binaries from source, clone the repo with `git clone https://github.com/LostRuins/koboldcpp.git`
 - A makefile is provided, simply run `make`.
-- Optional OpenBLAS: Link your own install of OpenBLAS manually with `make LLAMA_OPENBLAS=1`
+- Optional Vulkan: Link your own install of Vulkan SDK manually with `make LLAMA_VULKAN=1`
 - Optional CLBlast: Link your own install of CLBlast manually with `make LLAMA_CLBLAST=1`
 - Note: for these you will need to obtain and link OpenCL and CLBlast libraries.
-  - For Arch Linux: Install `cblas` `openblas` and `clblast`.
-  - For Debian: Install `libclblast-dev` and `libopenblas-dev`.
+  - For Arch Linux: Install `cblas` and `clblast`.
+  - For Debian: Install `libclblast-dev`.
 - You can attempt a CuBLAS build with `LLAMA_CUBLAS=1`, (or `LLAMA_HIPBLAS=1` for AMD). You will need CUDA Toolkit installed. Some have also reported success with the CMake file, though that is more for windows.
-- For a full featured build (all backends), do `make LLAMA_OPENBLAS=1 LLAMA_CLBLAST=1 LLAMA_CUBLAS=1 LLAMA_VULKAN=1`. (Note that `LLAMA_CUBLAS=1` will not work on windows, you need visual studio)
+- For a full featured build (all backends), do `make LLAMA_CLBLAST=1 LLAMA_CUBLAS=1 LLAMA_VULKAN=1`. (Note that `LLAMA_CUBLAS=1` will not work on windows, you need visual studio)
 - After all binaries are built, you can run the python script with the command `koboldcpp.py [ggml_model.gguf] [port]`
 
 ### Compiling on Windows
@@ -87,12 +88,11 @@ when you can't use the precompiled binary directly, we provide an automated buil
   - If you want to generate the .exe file, make sure you have the python module PyInstaller installed with pip (`pip install PyInstaller`). Then run the script `make_pyinstaller.bat`
   - The koboldcpp.exe file will be at your dist folder.
 - **Building with CUDA**: Visual Studio, CMake and CUDA Toolkit is required. Clone the repo, then open the CMake file and compile it in Visual Studio. Copy the `koboldcpp_cublas.dll` generated into the same directory as the `koboldcpp.py` file. If you are bundling executables, you may need to include CUDA dynamic libraries (such as `cublasLt64_11.dll` and `cublas64_11.dll`) in order for the executable to work correctly on a different PC.
-- **Replacing Libraries (Not Recommended)**: If you wish to use your own version of the additional Windows libraries (OpenCL, CLBlast and OpenBLAS), you can do it with:
+- **Replacing Libraries (Not Recommended)**: If you wish to use your own version of the additional Windows libraries (OpenCL, CLBlast, Vulkan), you can do it with:
   - OpenCL - tested with https://github.com/KhronosGroup/OpenCL-SDK . If you wish to compile it, follow the repository instructions. You will need vcpkg.
   - CLBlast - tested with https://github.com/CNugteren/CLBlast . If you wish to compile it you will need to reference the OpenCL files. It will only generate the ".lib" file if you compile using MSVC.
-  - OpenBLAS - tested with https://github.com/xianyi/OpenBLAS .
   - Move the respectives .lib files to the /lib folder of your project, overwriting the older files.
-  - Also, replace the existing versions of the corresponding .dll files located in the project directory root (e.g. libopenblas.dll).
+  - Also, replace the existing versions of the corresponding .dll files located in the project directory root (e.g. clblast.dll).
   - Make the KoboldCpp project using the instructions above.
 
 ### Compiling on MacOS
@@ -127,7 +127,7 @@ when you can't use the precompiled binary directly, we provide an automated buil
     - Metal is enabled by default on macOS, Vulkan support is enabled by default on both Linux and macOS, ROCm support isn't available yet.
     - You can also use `nix3-run` to use KoboldCpp: `nix run --expr ``with import <nixpkgs> { config = { allowUnfree = true; cudaSupport = true; }; }; koboldcpp`` --impure`
     - Or use `nix-shell`: `nix-shell --expr 'with import <nixpkgs> { config = { allowUnfree = true; cudaSupport = true; }; }; koboldcpp' --run "koboldcpp" --impure`
-    - Packages (like OpenBlast, CLBLast, Vulkan, etc.) can be overridden, please refer to the [17th Nix Pill - Nixpkgs Overriding Packages](https://nixos.org/guides/nix-pills/17-nixpkgs-overriding-packages)
+    - Packages (like CLBLast, Vulkan, etc.) can be overridden, please refer to the [17th Nix Pill - Nixpkgs Overriding Packages](https://nixos.org/guides/nix-pills/17-nixpkgs-overriding-packages)
 
 ## Questions and Help Wiki
 - **First, please check out [The KoboldCpp FAQ and Knowledgebase](https://github.com/LostRuins/koboldcpp/wiki) which may already have answers to your questions! Also please search through past issues and discussions.**
@@ -141,13 +141,13 @@ when you can't use the precompiled binary directly, we provide an automated buil
 
 ## Considerations
 - For Windows: No installation, single file executable, (It Just Works)
-- Since v1.0.6, requires libopenblas, the prebuilt windows binaries are included in this repo. If not found, it will fall back to a mode without BLAS.
 - Since v1.15, requires CLBlast if enabled, the prebuilt windows binaries are included in this repo. If not found, it will fall back to a mode without CLBlast.
 - Since v1.33, you can set the context size to be above what the model supports officially. It does increases perplexity but should still work well below 4096 even on untuned models. (For GPT-NeoX, GPT-J, and Llama models) Customize this with `--ropeconfig`.
 - Since v1.42, supports GGUF models for LLAMA and Falcon
 - Since v1.55, lcuda paths on Linux are hardcoded and may require manual changes to the makefile if you do not use koboldcpp.sh for the compilation.
 - Since v1.60, provides native image generation with StableDiffusion.cpp, you can load any SD1.5 or SDXL .safetensors model and it will provide an A1111 compatible API to use.
 - **I try to keep backwards compatibility with ALL past llama.cpp models**. But you are also encouraged to reconvert/update your models if possible for best results.
+- Since v1.75, openblas has been deprecated and removed in favor of the native CPU implementation.
 
 ## License
 - The original GGML library and llama.cpp by ggerganov are licensed under the MIT License

class.py

@@ -97,9 +97,9 @@ def get_requested_parameters(self, model_name, model_path, menu_path, parameters
                                     "menu_path": "",
                                     "refresh_model_inputs": False,
                                     "extra_classes": "",
-                                    'children': [{'text': 'Use No BLAS', 'value': 0}, {'text': 'Use OpenBLAS', 'value': 1}, {'text': 'Use CuBLAS', 'value': 2},
-                                    {'text': 'Use CLBLast GPU #1', 'value': 3},{'text': 'Use CLBLast GPU #2', 'value': 4},{'text': 'Use CLBLast GPU #3', 'value': 5}
-                                    ,{'text': 'NoAVX2 Mode (Old CPU)', 'value': 6},{'text': 'Failsafe Mode (Old CPU)', 'value': 7},{'text': 'Use Vulkan GPU #1', 'value': 8},{'text': 'Use Vulkan GPU #2', 'value': 9}],
+                                    'children': [{'text': 'Use No BLAS', 'value': 0}, {'text': 'Use CuBLAS', 'value': 1},
+                                    {'text': 'Use CLBLast GPU #1', 'value': 2},{'text': 'Use CLBLast GPU #2', 'value': 3},{'text': 'Use CLBLast GPU #3', 'value': 4}
+                                    ,{'text': 'NoAVX2 Mode (Old CPU)', 'value': 5},{'text': 'Failsafe Mode (Old CPU)', 'value': 6},{'text': 'Use Vulkan GPU #1', 'value': 7},{'text': 'Use Vulkan GPU #2', 'value': 8}],
                                     })
         requested_parameters.append({
                                     "uitype": "text",
@@ -238,24 +238,22 @@ def set_input_parameters(self, parameters):
         if accel==0:
             self.kcpp_noblas = True
         elif accel==1:
-           pass
-        elif accel==2:
             self.kcpp_usecublas = ["normal"]
-        elif accel==3:
+        elif accel==2:
             self.kcpp_useclblast = [0,0]
-        elif accel==4:
+        elif accel==3:
             self.kcpp_useclblast = [1,0]
-        elif accel==5:
+        elif accel==4:
             self.kcpp_useclblast = [0,1]
-        elif accel==6:
+        elif accel==5:
             self.kcpp_noavx2 = True
-        elif accel==7:
+        elif accel==6:
             self.kcpp_noavx2 = True
             self.kcpp_noblas = True
             self.kcpp_nommap = True
-        elif accel==8:
+        elif accel==7:
             self.kcpp_usevulkan = [0]
-        elif accel==9:
+        elif accel==8:
             self.kcpp_usevulkan = [1]
         pass
 

environment.yaml

@@ -11,8 +11,6 @@ dependencies:
   - gxx=10
   - pip
   - git=2.35.1
-  - libopenblas
-  - openblas
   - clblast
   - ninja
   - make

gpttype_adapter.cpp

@@ -1624,7 +1624,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in
     gptj_ctx_v3.hparams.rope_freq_scale = neox_ctx_v3.hparams.rope_freq_scale = rope_freq_scale;
     gptj_ctx_v3.hparams.rope_freq_base = neox_ctx_v3.hparams.rope_freq_base = rope_freq_base;
 
-    //this is used for the mem_per_token eval, openblas needs more RAM
+    //this is used for the mem_per_token eval, blas needs more RAM
     bool v3_use_scratch = ggml_v3_cpu_has_gpublas();
 
     int cu_parseinfo_maindevice = inputs.cublas_info<=0?0:inputs.cublas_info;
@@ -2362,11 +2362,11 @@ int GetThreadsToUse(bool blasmode)
     {
         if(!ggml_cpu_has_gpublas())
         {
-            return 1;
+            return std::min(kcpp_data->n_blasthreads, 4);
         }
         else
         {
-             return kcpp_data->n_blasthreads;
+            return kcpp_data->n_blasthreads;
         }
     }
     return kcpp_data->n_threads;