test/fix: OSX Test Repair (#1843)

* test with gguf instead of ggml. Updates testPrompt to match? Adds debugging line to Dockerfile that I've found helpful recently.

* fix testPrompt slightly

* Sad Experiment: Test GH runner without metal?

* break apart CGO_LDFLAGS

* switch runner

* upstream llama.cpp disables Metal on Github CI!

* missed a dir from clean-tests

* CGO_LDFLAGS

* tmate failure + NO_ACCELERATE

* whisper.cpp has a metal fix

* do the exact opposite of the name of this branch, but keep it around for unrelated fixes?

* add back newlines

* add tmate to linux for testing

* update fixtures

* timeout for tmate

.dockerignore

@@ -3,4 +3,4 @@ models
 examples/chatbot-ui/models
 examples/rwkv/models
 examples/**/models
-Dockerfile
+Dockerfile*

.github/workflows/image-pr.yml

@@ -22,6 +22,7 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      makeflags: "-j3"
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
@@ -80,6 +81,7 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      makeflags: "-j3"
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}

.github/workflows/image.yml

@@ -26,6 +26,7 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      makeflags: "-j3"
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}
@@ -199,6 +200,7 @@ jobs:
       platforms: ${{ matrix.platforms }}
       runs-on: ${{ matrix.runs-on }}
       base-image: ${{ matrix.base-image }}
+      makeflags: "-j3"
     secrets:
       dockerUsername: ${{ secrets.DOCKERHUB_USERNAME }}
       dockerPassword: ${{ secrets.DOCKERHUB_PASSWORD }}

.github/workflows/image_build.yml

@@ -46,6 +46,11 @@ on:
         required: true
         default: ''
         type: string
+      makeflags:
+        description: 'Make Flags'
+        required: false
+        default: ''
+        type: string
     secrets:
       dockerUsername:
         required: true
@@ -160,6 +165,7 @@ jobs:
             FFMPEG=${{ inputs.ffmpeg }}
             IMAGE_TYPE=${{ inputs.image-type }}
             BASE_IMAGE=${{ inputs.base-image }}
+            MAKEFLAGS=${{ inputs.makeflags }}
           context: .
           file: ./Dockerfile
           platforms: ${{ inputs.platforms }}

.github/workflows/test.yml

@@ -105,9 +105,13 @@ jobs:
       - name: Test
         run: |
           GO_TAGS="stablediffusion tts" make test
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3
+        timeout-minutes: 5
 
   tests-apple:
-    runs-on: macOS-latest
+    runs-on: macOS-14
     strategy:
       matrix:
         go-version: ['1.21.x']
@@ -130,4 +134,8 @@ jobs:
         run: |
           export C_INCLUDE_PATH=/usr/local/include
           export CPLUS_INCLUDE_PATH=/usr/local/include
-          CMAKE_ARGS="-DLLAMA_F16C=OFF -DLLAMA_AVX512=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF" make test
+          BUILD_TYPE="GITHUB_CI_HAS_BROKEN_METAL" CMAKE_ARGS="-DLLAMA_F16C=OFF -DLLAMA_AVX512=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF" make test
+      - name: Setup tmate session if tests fail
+        if: ${{ failure() }}
+        uses: mxschmitt/action-tmate@v3
+        timeout-minutes: 5

Dockerfile

@@ -63,7 +63,9 @@ WORKDIR /build
 RUN test -n "$TARGETARCH" \
     || (echo 'warn: missing $TARGETARCH, either set this `ARG` manually, or run using `docker buildkit`')
 
-# Extras requirements
+###################################
+###################################
+
 FROM requirements-core as requirements-extras
 
 RUN curl https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | gpg --dearmor > conda.gpg && \
@@ -93,8 +95,11 @@ FROM requirements-${IMAGE_TYPE} as builder
 ARG GO_TAGS="stablediffusion tts"
 ARG GRPC_BACKENDS
 ARG BUILD_GRPC=true
+ARG MAKEFLAGS
+
 ENV GRPC_BACKENDS=${GRPC_BACKENDS}
 ENV GO_TAGS=${GO_TAGS}
+ENV MAKEFLAGS=${MAKEFLAGS}
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
 ENV NVIDIA_REQUIRE_CUDA="cuda>=${CUDA_MAJOR_VERSION}.0"
 ENV NVIDIA_VISIBLE_DEVICES=all
@@ -103,6 +108,7 @@ WORKDIR /build
 
 COPY . .
 COPY .git .
+RUN echo "GO_TAGS: $GO_TAGS"
 RUN make prepare
 
 # If we are building with clblas support, we need the libraries for the builds
@@ -116,10 +122,10 @@ RUN if [ "${BUILD_TYPE}" = "clblas" ]; then \
 RUN GRPC_BACKENDS=backend-assets/grpc/stablediffusion make build
 
 RUN if [ "${BUILD_GRPC}" = "true" ]; then \
-    git clone --recurse-submodules -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
+    git clone --recurse-submodules --jobs 4 -b v1.58.0 --depth 1 --shallow-submodules https://github.com/grpc/grpc && \
     cd grpc && mkdir -p cmake/build && cd cmake/build && cmake -DgRPC_INSTALL=ON \
       -DgRPC_BUILD_TESTS=OFF \
-       ../.. && make -j12 install \
+       ../.. && make install \
     ; fi
 
 # Rebuild with defaults backends
@@ -139,10 +145,12 @@ ARG FFMPEG
 ARG BUILD_TYPE
 ARG TARGETARCH
 ARG IMAGE_TYPE=extras
+ARG MAKEFLAGS
 
 ENV BUILD_TYPE=${BUILD_TYPE}
 ENV REBUILD=false
 ENV HEALTHCHECK_ENDPOINT=http://localhost:8080/readyz
+ENV MAKEFLAGS=${MAKEFLAGS}
 
 ARG CUDA_MAJOR_VERSION=11
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
@@ -186,43 +194,43 @@ COPY --from=builder /build/backend-assets/grpc/stablediffusion ./backend-assets/
 
 ## Duplicated from Makefile to avoid having a big layer that's hard to push
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/autogptq \
+    make -C backend/python/autogptq \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/bark \
+    make -C backend/python/bark \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/diffusers \
+    make -C backend/python/diffusers \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/vllm \
+    make -C backend/python/vllm \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/mamba \
+    make -C backend/python/mamba \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/sentencetransformers \
+    make -C backend/python/sentencetransformers \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/transformers \
+    make -C backend/python/transformers \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/vall-e-x \
+    make -C backend/python/vall-e-x \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/exllama \
+    make -C backend/python/exllama \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-     make -C backend/python/exllama2 \
+    make -C backend/python/exllama2 \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/petals \
+    make -C backend/python/petals \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/transformers-musicgen \
+    make -C backend/python/transformers-musicgen \
     ; fi
 RUN if [ "${IMAGE_TYPE}" = "extras" ]; then \
-	 make -C backend/python/coqui \
+    make -C backend/python/coqui \
     ; fi
 
 # Make sure the models directory exists

Makefile

@@ -4,11 +4,8 @@ GOVET=$(GOCMD) vet
 BINARY_NAME=local-ai
 
 # llama.cpp versions
-GOLLAMA_VERSION?=aeba71ee842819da681ea537e78846dc75949ac0
-
-GOLLAMA_STABLE_VERSION?=50cee7712066d9e38306eccadcfbb44ea87df4b7
-
-CPPLLAMA_VERSION?=19885d205e768579ab090d1e99281cae58c21b54
+GOLLAMA_STABLE_VERSION?=2b57a8ae43e4699d3dc5d1496a1ccd42922993be
+CPPLLAMA_VERSION?=d01b3c4c32357567f3531d4e6ceffc5d23e87583
 
 # gpt4all version
 GPT4ALL_REPO?=https://github.com/nomic-ai/gpt4all
@@ -19,13 +16,13 @@ RWKV_REPO?=https://github.com/donomii/go-rwkv.cpp
 RWKV_VERSION?=661e7ae26d442f5cfebd2a0881b44e8c55949ec6
 
 # whisper.cpp version
-WHISPER_CPP_VERSION?=37a709f6558c6d9783199e2b8cbb136e1c41d346
+WHISPER_CPP_VERSION?=a56f435fd475afd7edf02bfbf9f8c77f527198c2
 
 # bert.cpp version
 BERT_VERSION?=6abe312cded14042f6b7c3cd8edf082713334a4d
 
 # go-piper version
-PIPER_VERSION?=d6b6275ba037dabdba4a8b65dfdf6b2a73a67f07
+PIPER_VERSION?=9d0100873a7dbb0824dfea40e8cec70a1b110759
 
 # stablediffusion version
 STABLEDIFFUSION_VERSION?=362df9da29f882dbf09ade61972d16a1f53c3485
@@ -38,6 +35,7 @@ export STABLE_BUILD_TYPE?=$(BUILD_TYPE)
 export CMAKE_ARGS?=
 
 CGO_LDFLAGS?=
+CGO_LDFLAGS_WHISPER?=
 CUDA_LIBPATH?=/usr/local/cuda/lib64/
 GO_TAGS?=
 BUILD_ID?=git
@@ -72,7 +70,7 @@ UNAME_S := $(shell uname -s)
 endif
 
 ifeq ($(OS),Darwin)
-	CGO_LDFLAGS += -lcblas -framework Accelerate
+	
 	ifeq ($(OSX_SIGNING_IDENTITY),)
 		OSX_SIGNING_IDENTITY := $(shell security find-identity -v -p codesigning | grep '"' | head -n 1 | sed -E 's/.*"(.*)"/\1/')
 	endif
@@ -83,6 +81,12 @@ ifeq ($(OS),Darwin)
 	# disable metal if on Darwin and any other value is explicitly passed.
 	else ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DLLAMA_METAL=OFF
+		export LLAMA_NO_ACCELERATE=1
+	endif
+
+	ifeq ($(BUILD_TYPE),metal)
+#			-lcblas 	removed: it seems to always be listed as a duplicate flag.
+		CGO_LDFLAGS += -framework Accelerate
 	endif
 endif
 
@@ -91,10 +95,12 @@ ifeq ($(BUILD_TYPE),openblas)
 	export WHISPER_OPENBLAS=1
 endif
 
+
 ifeq ($(BUILD_TYPE),cublas)
 	CGO_LDFLAGS+=-lcublas -lcudart -L$(CUDA_LIBPATH)
 	export LLAMA_CUBLAS=1
 	export WHISPER_CUBLAS=1
+	CGO_LDFLAGS_WHISPER+=-L$(CUDA_LIBPATH)/stubs/ -lcuda
 endif
 
 ifeq ($(BUILD_TYPE),hipblas)
@@ -148,7 +154,6 @@ endif
 
 ALL_GRPC_BACKENDS=backend-assets/grpc/langchain-huggingface
 ALL_GRPC_BACKENDS+=backend-assets/grpc/bert-embeddings
-ALL_GRPC_BACKENDS+=backend-assets/grpc/llama
 ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-cpp
 ALL_GRPC_BACKENDS+=backend-assets/grpc/llama-ggml
 ALL_GRPC_BACKENDS+=backend-assets/grpc/gpt4all
@@ -168,40 +173,41 @@ ifeq ($(BUILD_API_ONLY),true)
 	GRPC_BACKENDS=
 endif
 
-.PHONY: all test build vendor
+.PHONY: all test build vendor get-sources prepare-sources prepare
 
 all: help
 
-## GPT4ALL
-sources/gpt4all:
-	git clone --recurse-submodules $(GPT4ALL_REPO) sources/gpt4all
-	cd sources/gpt4all && git checkout -b build $(GPT4ALL_VERSION) && git submodule update --init --recursive --depth 1
-
-## go-piper
-sources/go-piper:
-	git clone --recurse-submodules https://github.com/mudler/go-piper sources/go-piper
-	cd sources/go-piper && git checkout -b build $(PIPER_VERSION) && git submodule update --init --recursive --depth 1
-
 ## BERT embeddings
 sources/go-bert:
 	git clone --recurse-submodules https://github.com/go-skynet/go-bert.cpp sources/go-bert
 	cd sources/go-bert && git checkout -b build $(BERT_VERSION) && git submodule update --init --recursive --depth 1
 
-## stable diffusion
-sources/go-stable-diffusion:
-	git clone --recurse-submodules https://github.com/mudler/go-stable-diffusion sources/go-stable-diffusion
-	cd sources/go-stable-diffusion && git checkout -b build $(STABLEDIFFUSION_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-bert/libgobert.a: sources/go-bert
+	$(MAKE) -C sources/go-bert libgobert.a
 
-sources/go-stable-diffusion/libstablediffusion.a:
-	$(MAKE) -C sources/go-stable-diffusion libstablediffusion.a
+## go-llama-ggml
+sources/go-llama-ggml:
+	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama-ggml
+	cd sources/go-llama-ggml && git checkout -b build $(GOLLAMA_STABLE_VERSION) && git submodule update --init --recursive --depth 1
 
-## tiny-dream
-sources/go-tiny-dream:
-	git clone --recurse-submodules https://github.com/M0Rf30/go-tiny-dream sources/go-tiny-dream
-	cd sources/go-tiny-dream && git checkout -b build $(TINYDREAM_VERSION) && git submodule update --init --recursive --depth 1
+sources/go-llama-ggml/libbinding.a: sources/go-llama-ggml
+	$(MAKE) -C sources/go-llama-ggml BUILD_TYPE=$(STABLE_BUILD_TYPE) libbinding.a
 
-sources/go-tiny-dream/libtinydream.a:
-	$(MAKE) -C sources/go-tiny-dream libtinydream.a
+## go-piper
+sources/go-piper:
+	git clone --recurse-submodules https://github.com/mudler/go-piper sources/go-piper
+	cd sources/go-piper && git checkout -b build $(PIPER_VERSION) && git submodule update --init --recursive --depth 1
+
+sources/go-piper/libpiper_binding.a: sources/go-piper
+	$(MAKE) -C sources/go-piper libpiper_binding.a example/main piper.o
+
+## GPT4ALL
+sources/gpt4all:
+	git clone --recurse-submodules $(GPT4ALL_REPO) sources/gpt4all
+	cd sources/gpt4all && git checkout -b build $(GPT4ALL_VERSION) && git submodule update --init --recursive --depth 1
+
+sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a: sources/gpt4all
+	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ libgpt4all.a
 
 ## RWKV
 sources/go-rwkv:
@@ -211,23 +217,23 @@ sources/go-rwkv:
 sources/go-rwkv/librwkv.a: sources/go-rwkv
 	cd sources/go-rwkv && cd rwkv.cpp &&	cmake . -DRWKV_BUILD_SHARED_LIBRARY=OFF &&	cmake --build . && 	cp librwkv.a ..
 
-sources/go-bert/libgobert.a: sources/go-bert
-	$(MAKE) -C sources/go-bert libgobert.a
+## stable diffusion
+sources/go-stable-diffusion:
+	git clone --recurse-submodules https://github.com/mudler/go-stable-diffusion sources/go-stable-diffusion
+	cd sources/go-stable-diffusion && git checkout -b build $(STABLEDIFFUSION_VERSION) && git submodule update --init --recursive --depth 1
 
-backend-assets/gpt4all: sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
-	mkdir -p backend-assets/gpt4all
-	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.so backend-assets/gpt4all/ || true
-	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
-	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
+sources/go-stable-diffusion/libstablediffusion.a: sources/go-stable-diffusion
+	$(MAKE) -C sources/go-stable-diffusion libstablediffusion.a
 
-backend-assets/espeak-ng-data: sources/go-piper
-	mkdir -p backend-assets/espeak-ng-data
-	$(MAKE) -C sources/go-piper piper.o
-	@cp -rf sources/go-piper/piper-phonemize/pi/share/espeak-ng-data/. backend-assets/espeak-ng-data
+## tiny-dream
+sources/go-tiny-dream:
+	git clone --recurse-submodules https://github.com/M0Rf30/go-tiny-dream sources/go-tiny-dream
+	cd sources/go-tiny-dream && git checkout -b build $(TINYDREAM_VERSION) && git submodule update --init --recursive --depth 1
 
-sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a: sources/gpt4all
-	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ libgpt4all.a
+sources/go-tiny-dream/libtinydream.a: sources/go-tiny-dream
+	$(MAKE) -C sources/go-tiny-dream libtinydream.a
 
+## whisper
 sources/whisper.cpp:
 	git clone https://github.com/ggerganov/whisper.cpp.git sources/whisper.cpp
 	cd sources/whisper.cpp && git checkout -b build $(WHISPER_CPP_VERSION) && git submodule update --init --recursive --depth 1
@@ -235,47 +241,34 @@ sources/whisper.cpp:
 sources/whisper.cpp/libwhisper.a: sources/whisper.cpp
 	cd sources/whisper.cpp && make libwhisper.a
 
-sources/go-llama:
-	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama
-	cd sources/go-llama && git checkout -b build $(GOLLAMA_VERSION) && git submodule update --init --recursive --depth 1
-
-sources/go-llama-ggml:
-	git clone --recurse-submodules https://github.com/go-skynet/go-llama.cpp sources/go-llama-ggml
-	cd sources/go-llama-ggml && git checkout -b build $(GOLLAMA_STABLE_VERSION) && git submodule update --init --recursive --depth 1
-
-sources/go-llama/libbinding.a: sources/go-llama
-	$(MAKE) -C sources/go-llama BUILD_TYPE=$(BUILD_TYPE) libbinding.a
-
-sources/go-llama-ggml/libbinding.a: sources/go-llama-ggml
-	$(MAKE) -C sources/go-llama-ggml BUILD_TYPE=$(STABLE_BUILD_TYPE) libbinding.a
-
-sources/go-piper/libpiper_binding.a: sources/go-piper
-	$(MAKE) -C sources/go-piper libpiper_binding.a example/main
-
-backend/cpp/llama/llama.cpp:
-	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama llama.cpp
-
-get-sources: backend/cpp/llama/llama.cpp sources/go-llama sources/go-llama-ggml sources/gpt4all sources/go-piper sources/go-rwkv sources/whisper.cpp sources/go-bert sources/go-stable-diffusion sources/go-tiny-dream
-	touch $@
+get-sources: sources/go-llama-ggml sources/gpt4all sources/go-piper sources/go-rwkv sources/whisper.cpp sources/go-bert sources/go-stable-diffusion sources/go-tiny-dream
 
 replace:
-	$(GOCMD) mod edit -replace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang
 	$(GOCMD) mod edit -replace github.com/donomii/go-rwkv.cpp=$(CURDIR)/sources/go-rwkv
 	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp=$(CURDIR)/sources/whisper.cpp
 	$(GOCMD) mod edit -replace github.com/ggerganov/whisper.cpp/bindings/go=$(CURDIR)/sources/whisper.cpp/bindings/go
 	$(GOCMD) mod edit -replace github.com/go-skynet/go-bert.cpp=$(CURDIR)/sources/go-bert
-	$(GOCMD) mod edit -replace github.com/mudler/go-stable-diffusion=$(CURDIR)/sources/go-stable-diffusion
 	$(GOCMD) mod edit -replace github.com/M0Rf30/go-tiny-dream=$(CURDIR)/sources/go-tiny-dream
 	$(GOCMD) mod edit -replace github.com/mudler/go-piper=$(CURDIR)/sources/go-piper
+	$(GOCMD) mod edit -replace github.com/mudler/go-stable-diffusion=$(CURDIR)/sources/go-stable-diffusion
+	$(GOCMD) mod edit -replace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang
+
+dropreplace:
+	$(GOCMD) mod edit -dropreplace github.com/donomii/go-rwkv.cpp
+	$(GOCMD) mod edit -dropreplace github.com/ggerganov/whisper.cpp
+	$(GOCMD) mod edit -dropreplace github.com/ggerganov/whisper.cpp/bindings/go
+	$(GOCMD) mod edit -dropreplace github.com/go-skynet/go-bert.cpp
+	$(GOCMD) mod edit -dropreplace github.com/M0Rf30/go-tiny-dream
+	$(GOCMD) mod edit -dropreplace github.com/mudler/go-piper
+	$(GOCMD) mod edit -dropreplace github.com/mudler/go-stable-diffusion
+	$(GOCMD) mod edit -dropreplace github.com/nomic-ai/gpt4all/gpt4all-bindings/golang
 
 prepare-sources: get-sources replace
 	$(GOCMD) mod download
-	touch $@
 
 ## GENERIC
 rebuild: ## Rebuilds the project
 	$(GOCMD) clean -cache
-	$(MAKE) -C sources/go-llama clean
 	$(MAKE) -C sources/go-llama-ggml clean
 	$(MAKE) -C sources/gpt4all/gpt4all-bindings/golang/ clean
 	$(MAKE) -C sources/go-rwkv clean
@@ -287,7 +280,6 @@ rebuild: ## Rebuilds the project
 	$(MAKE) build
 
 prepare: prepare-sources $(OPTIONAL_TARGETS)
-	touch $@
 
 clean: ## Remove build related file
 	$(GOCMD) clean -cache
@@ -298,10 +290,15 @@ clean: ## Remove build related file
 	rm -rf backend-assets
 	$(MAKE) -C backend/cpp/grpc clean
 	$(MAKE) -C backend/cpp/llama clean
+	$(MAKE) dropreplace
+
+clean-tests:
+	rm -rf test-models
+	rm -rf test-dir
+	rm -rf core/http/backend-assets
 
 ## Build:
-
-build: backend-assets grpcs prepare ## Build the project
+build: prepare backend-assets grpcs ## Build the project
 	$(info ${GREEN}I local-ai build info:${RESET})
 	$(info ${GREEN}I BUILD_TYPE: ${YELLOW}$(BUILD_TYPE)${RESET})
 	$(info ${GREEN}I GO_TAGS: ${YELLOW}$(GO_TAGS)${RESET})
@@ -319,10 +316,10 @@ osx-signed: build
 run: prepare ## run local-ai
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" $(GOCMD) run ./
 
-test-models/testmodel:
+test-models/testmodel.ggml:
 	mkdir test-models
 	mkdir test-dir
-	wget -q https://huggingface.co/TheBloke/orca_mini_3B-GGML/resolve/main/orca-mini-3b.ggmlv3.q4_0.bin -O test-models/testmodel
+	wget -q https://huggingface.co/TheBloke/orca_mini_3B-GGML/resolve/main/orca-mini-3b.ggmlv3.q4_0.bin -O test-models/testmodel.ggml
 	wget -q https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.en.bin -O test-models/whisper-en
 	wget -q https://huggingface.co/mudler/all-MiniLM-L6-v2/resolve/main/ggml-model-q4_0.bin -O test-models/bert
 	wget -q https://cdn.openai.com/whisper/draft-20220913a/micro-machines.wav -O test-dir/audio.wav
@@ -334,7 +331,7 @@ prepare-test: grpcs
 	cp -rf backend-assets core/http
 	cp tests/models_fixtures/* test-models
 
-test: prepare test-models/testmodel grpcs
+test: prepare test-models/testmodel.ggml grpcs
 	@echo 'Running tests'
 	export GO_TAGS="tts stablediffusion"
 	$(MAKE) prepare-test
@@ -454,87 +451,85 @@ ifeq ($(BUILD_API_ONLY),true)
 	touch backend-assets/keep
 endif
 
-backend-assets/grpc:
+backend-assets/espeak-ng-data: sources/go-piper sources/go-piper/libpiper_binding.a
+	mkdir -p backend-assets/espeak-ng-data
+	@cp -rf sources/go-piper/piper-phonemize/pi/share/espeak-ng-data/. backend-assets/espeak-ng-data
+
+backend-assets/gpt4all: sources/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
+	mkdir -p backend-assets/gpt4all
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.so backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dylib backend-assets/gpt4all/ || true
+	@cp sources/gpt4all/gpt4all-bindings/golang/buildllm/*.dll backend-assets/gpt4all/ || true
+
+backend-assets/grpc: replace
 	mkdir -p backend-assets/grpc
 
-backend-assets/grpc/llama: backend-assets/grpc sources/go-llama/libbinding.a
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(CURDIR)/sources/go-llama
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama LIBRARY_PATH=$(CURDIR)/sources/go-llama \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama ./backend/go/llm/llama/
-# TODO: every binary should have its own folder instead, so can have different  implementations
-ifeq ($(BUILD_TYPE),metal)
-	cp backend/cpp/llama/llama.cpp/ggml-metal.metal backend-assets/grpc/
-endif
-
-## BACKEND CPP LLAMA START
-# Sets the variables in case it has to build the gRPC locally.
-INSTALLED_PACKAGES=$(CURDIR)/backend/cpp/grpc/installed_packages
-INSTALLED_LIB_CMAKE=$(INSTALLED_PACKAGES)/lib/cmake
-ADDED_CMAKE_ARGS=-Dabsl_DIR=${INSTALLED_LIB_CMAKE}/absl \
-                 -DProtobuf_DIR=${INSTALLED_LIB_CMAKE}/protobuf \
-                 -Dutf8_range_DIR=${INSTALLED_LIB_CMAKE}/utf8_range \
-                 -DgRPC_DIR=${INSTALLED_LIB_CMAKE}/grpc \
-                 -DCMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES=${INSTALLED_PACKAGES}/include
-
-backend/cpp/llama/grpc-server:
-ifdef BUILD_GRPC_FOR_BACKEND_LLAMA
-	$(MAKE) -C backend/cpp/grpc build
-	export _PROTOBUF_PROTOC=${INSTALLED_PACKAGES}/bin/proto && \
-	export _GRPC_CPP_PLUGIN_EXECUTABLE=${INSTALLED_PACKAGES}/bin/grpc_cpp_plugin && \
-	export PATH="${INSTALLED_PACKAGES}/bin:${PATH}" && \
-	CMAKE_ARGS="${CMAKE_ARGS} ${ADDED_CMAKE_ARGS}" LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server
-else
-	echo "BUILD_GRPC_FOR_BACKEND_LLAMA is not defined."
-	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server
-endif
-## BACKEND CPP LLAMA END
-
-##
-backend-assets/grpc/llama-cpp: backend-assets/grpc backend/cpp/llama/grpc-server
-	cp -rfv backend/cpp/llama/grpc-server backend-assets/grpc/llama-cpp
-# TODO: every binary should have its own folder instead, so can have different metal implementations
-ifeq ($(BUILD_TYPE),metal)
-	cp backend/cpp/llama/llama.cpp/build/bin/ggml-metal.metal backend-assets/grpc/
-endif
-
-backend-assets/grpc/llama-ggml: backend-assets/grpc sources/go-llama-ggml/libbinding.a
-	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(CURDIR)/sources/go-llama-ggml
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama-ggml LIBRARY_PATH=$(CURDIR)/sources/go-llama-ggml \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-ggml ./backend/go/llm/llama-ggml/
-
-backend-assets/grpc/gpt4all: backend-assets/grpc backend-assets/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ LIBRARY_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt4all ./backend/go/llm/gpt4all/
-
-backend-assets/grpc/rwkv: backend-assets/grpc sources/go-rwkv/librwkv.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-rwkv LIBRARY_PATH=$(CURDIR)/sources/go-rwkv \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./backend/go/llm/rwkv
-
-backend-assets/grpc/bert-embeddings: backend-assets/grpc sources/go-bert/libgobert.a
+backend-assets/grpc/bert-embeddings: sources/go-bert sources/go-bert/libgobert.a backend-assets/grpc
 	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-bert LIBRARY_PATH=$(CURDIR)/sources/go-bert \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/bert-embeddings ./backend/go/llm/bert/
 
+backend-assets/grpc/gpt4all: sources/gpt4all sources/gpt4all/gpt4all-bindings/golang/libgpt4all.a backend-assets/gpt4all backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ LIBRARY_PATH=$(CURDIR)/sources/gpt4all/gpt4all-bindings/golang/ \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/gpt4all ./backend/go/llm/gpt4all/
+
 backend-assets/grpc/langchain-huggingface: backend-assets/grpc
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/langchain-huggingface ./backend/go/llm/langchain/
 
-backend-assets/grpc/stablediffusion: backend-assets/grpc
-	if [ ! -f backend-assets/grpc/stablediffusion ]; then \
-		$(MAKE) sources/go-stable-diffusion; \
-		$(MAKE) sources/go-stable-diffusion/libstablediffusion.a; \
-		CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-stable-diffusion/ LIBRARY_PATH=$(CURDIR)/sources/go-stable-diffusion/ \
-		$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./backend/go/image/stablediffusion; \
-	fi
+backend/cpp/llama/llama.cpp:
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama llama.cpp
 
-backend-assets/grpc/tinydream: backend-assets/grpc sources/go-tiny-dream/libtinydream.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/go-tiny-dream \
-	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/tinydream ./backend/go/image/tinydream
+INSTALLED_PACKAGES=$(CURDIR)/backend/cpp/grpc/installed_packages
+INSTALLED_LIB_CMAKE=$(INSTALLED_PACKAGES)/lib/cmake
+ADDED_CMAKE_ARGS=-Dabsl_DIR=${INSTALLED_LIB_CMAKE}/absl \
+				 -DProtobuf_DIR=${INSTALLED_LIB_CMAKE}/protobuf \
+				 -Dutf8_range_DIR=${INSTALLED_LIB_CMAKE}/utf8_range \
+				 -DgRPC_DIR=${INSTALLED_LIB_CMAKE}/grpc \
+				 -DCMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES=${INSTALLED_PACKAGES}/include
+backend/cpp/llama/grpc-server:
+# Conditionally build grpc for the llama backend to use if needed
+ifdef BUILD_GRPC_FOR_BACKEND_LLAMA
+	$(MAKE) -C backend/cpp/grpc build
+	_PROTOBUF_PROTOC=${INSTALLED_PACKAGES}/bin/proto \
+	_GRPC_CPP_PLUGIN_EXECUTABLE=${INSTALLED_PACKAGES}/bin/grpc_cpp_plugin \
+	PATH="${INSTALLED_PACKAGES}/bin:${PATH}" \
+	CMAKE_ARGS="${CMAKE_ARGS} ${ADDED_CMAKE_ARGS}" \
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) \
+	$(MAKE) -C backend/cpp/llama grpc-server
+else
+	echo "BUILD_GRPC_FOR_BACKEND_LLAMA is not defined."
+	LLAMA_VERSION=$(CPPLLAMA_VERSION) $(MAKE) -C backend/cpp/llama grpc-server
+endif
 
-backend-assets/grpc/piper: backend-assets/grpc backend-assets/espeak-ng-data sources/go-piper/libpiper_binding.a
+backend-assets/grpc/llama-cpp: backend-assets/grpc backend/cpp/llama/grpc-server
+	cp -rfv backend/cpp/llama/grpc-server backend-assets/grpc/llama-cpp
+# TODO: every binary should have its own folder instead, so can have different metal implementations
+ifeq ($(BUILD_TYPE),metal)
+	cp backend/cpp/llama/llama.cpp/build/bin/default.metallib backend-assets/grpc/
+endif
+
+backend-assets/grpc/llama-ggml: sources/go-llama-ggml sources/go-llama-ggml/libbinding.a backend-assets/grpc
+	$(GOCMD) mod edit -replace github.com/go-skynet/go-llama.cpp=$(CURDIR)/sources/go-llama-ggml
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-llama-ggml LIBRARY_PATH=$(CURDIR)/sources/go-llama-ggml \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/llama-ggml ./backend/go/llm/llama-ggml/
+
+backend-assets/grpc/piper: sources/go-piper sources/go-piper/libpiper_binding.a backend-assets/grpc backend-assets/espeak-ng-data
 	CGO_CXXFLAGS="$(PIPER_CGO_CXXFLAGS)" CGO_LDFLAGS="$(PIPER_CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/sources/go-piper \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/piper ./backend/go/tts/
 
-backend-assets/grpc/whisper: backend-assets/grpc sources/whisper.cpp/libwhisper.a
-	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/whisper.cpp LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
+backend-assets/grpc/rwkv: sources/go-rwkv sources/go-rwkv/librwkv.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-rwkv LIBRARY_PATH=$(CURDIR)/sources/go-rwkv \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/rwkv ./backend/go/llm/rwkv
+
+backend-assets/grpc/stablediffusion: sources/go-stable-diffusion sources/go-stable-diffusion/libstablediffusion.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" C_INCLUDE_PATH=$(CURDIR)/sources/go-stable-diffusion/ LIBRARY_PATH=$(CURDIR)/sources/go-stable-diffusion/ \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/stablediffusion ./backend/go/image/stablediffusion
+
+backend-assets/grpc/tinydream: sources/go-tiny-dream sources/go-tiny-dream/libtinydream.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS)" LIBRARY_PATH=$(CURDIR)/go-tiny-dream \
+	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/tinydream ./backend/go/image/tinydream
+
+backend-assets/grpc/whisper: sources/whisper.cpp sources/whisper.cpp/libwhisper.a backend-assets/grpc
+	CGO_LDFLAGS="$(CGO_LDFLAGS) $(CGO_LDFLAGS_WHISPER)" C_INCLUDE_PATH=$(CURDIR)/sources/whisper.cpp LIBRARY_PATH=$(CURDIR)/sources/whisper.cpp \
 	$(GOCMD) build -ldflags "$(LD_FLAGS)" -tags "$(GO_TAGS)" -o backend-assets/grpc/whisper ./backend/go/transcribe/
 
 grpcs: prepare $(GRPC_BACKENDS)

backend/backend.proto

@@ -191,6 +191,7 @@ message TTSRequest {
   string text = 1;
   string model = 2;
   string dst = 3;
+  string voice = 4;
 }
 
 message TokenizationResponse {

backend/cpp/grpc/Makefile

@@ -48,7 +48,7 @@ $(INSTALLED_PACKAGES): grpc_build
 
 $(GRPC_REPO):
 	git clone --depth $(GIT_CLONE_DEPTH) -b $(TAG_LIB_GRPC) $(GIT_REPO_LIB_GRPC) $(GRPC_REPO)/grpc
-	cd $(GRPC_REPO)/grpc && git submodule update --init --recursive --depth $(GIT_CLONE_DEPTH)
+	cd $(GRPC_REPO)/grpc && git submodule update --jobs 2 --init --recursive --depth $(GIT_CLONE_DEPTH)
 
 $(GRPC_BUILD): $(GRPC_REPO)
 	mkdir -p $(GRPC_BUILD)

backend/cpp/llama/Makefile

@@ -18,6 +18,12 @@ else ifeq ($(BUILD_TYPE),clblas)
 # If it's hipblas we do have also to set CC=/opt/rocm/llvm/bin/clang CXX=/opt/rocm/llvm/bin/clang++ 
 else ifeq ($(BUILD_TYPE),hipblas)
 	CMAKE_ARGS+=-DLLAMA_HIPBLAS=ON
+# If it's OSX, DO NOT embed the metal library - -DLLAMA_METAL_EMBED_LIBRARY=ON requires further investigation
+# But if it's OSX without metal, disable it here
+else ifeq ($(OS),darwin)
+	ifneq ($(BUILD_TYPE),metal)
+		CMAKE_ARGS+=-DLLAMA_METAL=OFF
+	endif
 endif
 
 ifeq ($(BUILD_TYPE),sycl_f16)
@@ -35,7 +41,7 @@ llama.cpp:
 	fi
 	cd llama.cpp && git checkout -b build $(LLAMA_VERSION) && git submodule update --init --recursive --depth 1
 
-llama.cpp/examples/grpc-server:
+llama.cpp/examples/grpc-server: llama.cpp
 	mkdir -p llama.cpp/examples/grpc-server
 	cp -r $(abspath ./)/CMakeLists.txt llama.cpp/examples/grpc-server/
 	cp -r $(abspath ./)/grpc-server.cpp llama.cpp/examples/grpc-server/

backend/cpp/llama/grpc-server.cpp

@@ -1084,7 +1084,7 @@ struct llama_server_context
             slot.has_next_token = false;
         }
 
-        if (!slot.cache_tokens.empty() && result.tok == llama_token_eos(model))
+        if (result.tok == llama_token_eos(model))
         {
             slot.stopped_eos = true;
             slot.has_next_token = false;

backend/python/common-env/transformers/transformers-nvidia.yml

@@ -30,6 +30,7 @@ dependencies:
       - async-timeout==4.0.3
       - attrs==23.1.0
       - bark==0.1.5
+      - bitsandbytes==0.43.0
       - boto3==1.28.61
       - botocore==1.31.61
       - certifi==2023.7.22

backend/python/transformers/transformers_server.py

@@ -23,7 +23,7 @@ if XPU:
     from intel_extension_for_transformers.transformers.modeling import AutoModelForCausalLM
     from transformers import AutoTokenizer, AutoModel, set_seed
 else:
-    from transformers import AutoTokenizer, AutoModel, AutoModelForCausalLM, set_seed
+    from transformers import AutoTokenizer, AutoModel, AutoModelForCausalLM, set_seed, BitsAndBytesConfig
 
 
 _ONE_DAY_IN_SECONDS = 60 * 60 * 24
@@ -75,18 +75,50 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
             A Result object that contains the result of the LoadModel operation.
         """
         model_name = request.Model
+
+        compute = "auto"
+        if request.F16Memory == True:
+            compute=torch.bfloat16
+
+        self.CUDA = request.CUDA
+
+        device_map="cpu"
+
+        quantization = None
+
+        if self.CUDA:
+            if request.Device:
+                device_map=request.Device
+            else:
+                device_map="cuda:0"
+            if request.Quantization == "bnb_4bit":
+                quantization = BitsAndBytesConfig(
+                    load_in_4bit = True,
+                    bnb_4bit_compute_dtype = compute,
+                    bnb_4bit_quant_type = "nf4",
+                    bnb_4bit_use_double_quant = True,
+                    load_in_8bit = False,
+                )
+            elif request.Quantization == "bnb_8bit":
+                quantization = BitsAndBytesConfig(
+                    load_in_4bit=False,
+                    bnb_4bit_compute_dtype = None,
+                    load_in_8bit=True,                                   
+                )
+                                                   
+    
         try:
             if request.Type == "AutoModelForCausalLM":
                 if XPU:
+                    if quantization == "xpu_4bit":
+                        xpu_4bit = True
                     self.model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=request.TrustRemoteCode,
-                                              device_map="xpu", load_in_4bit=True)
+                                              device_map="xpu", load_in_4bit=xpu_4bit)
                 else:
-                    self.model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=request.TrustRemoteCode)
+                    self.model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=request.TrustRemoteCode, use_safetensors=True, quantization_config=quantization, device_map=device_map, torch_dtype=compute)
             else:
-                self.model = AutoModel.from_pretrained(model_name, trust_remote_code=request.TrustRemoteCode)
-
-            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-            self.CUDA = False
+                self.model = AutoModel.from_pretrained(model_name, trust_remote_code=request.TrustRemoteCode,  use_safetensors=True,  quantization_config=quantization, device_map=device_map, torch_dtype=compute)
+            self.tokenizer = AutoTokenizer.from_pretrained(model_name, use_safetensors=True)
             self.XPU = False
 
             if XPU:
@@ -97,13 +129,6 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                 except Exception as err:
                     print("Not using XPU:", err, file=sys.stderr)
 
-            if request.CUDA or torch.cuda.is_available():
-                try:
-                    print("Loading model", model_name, "to CUDA.", file=sys.stderr)
-                    self.model = self.model.to("cuda")
-                    self.CUDA = True
-                except Exception as err:
-                    print("Not using CUDA:", err, file=sys.stderr)
         except Exception as err:
             return backend_pb2.Result(success=False, message=f"Unexpected {err=}, {type(err)=}")
         # Implement your logic here for the LoadModel service
@@ -130,13 +155,17 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         encoded_input = self.tokenizer(request.Embeddings, padding=True, truncation=True, max_length=max_length, return_tensors="pt")    
 
         # Create word embeddings
-        model_output = self.model(**encoded_input)
+        if self.CUDA:
+            encoded_input = encoded_input.to("cuda")
+
+        with torch.no_grad():    
+            model_output = self.model(**encoded_input)
 
         # Pool to get sentence embeddings; i.e. generate one 1024 vector for the entire sentence
-        sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask']).detach().numpy()
+        sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
         print("Calculated embeddings for: " + request.Embeddings, file=sys.stderr)
         print("Embeddings:", sentence_embeddings, file=sys.stderr)
-        return backend_pb2.EmbeddingResult(embeddings=sentence_embeddings)
+        return backend_pb2.EmbeddingResult(embeddings=sentence_embeddings[0])
 
     def Predict(self, request, context):
         """
@@ -163,12 +192,8 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
         if XPU:
             inputs = inputs.to("xpu")
 
-        outputs = self.model.generate(inputs,max_new_tokens=max_tokens, temperature=request.Temperature, top_p=request.TopP)
-
-        generated_text = self.tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
-        # Remove prompt from response if present
-        if request.Prompt in generated_text:
-            generated_text = generated_text.replace(request.Prompt, "")
+        outputs = self.model.generate(inputs,max_new_tokens=max_tokens, temperature=request.Temperature, top_p=request.TopP, do_sample=True, pad_token_id=self.tokenizer.eos_token_id)
+        generated_text = self.tokenizer.batch_decode(outputs[:, inputs.shape[1]:], skip_special_tokens=True)[0]
 
         return backend_pb2.Reply(message=bytes(generated_text, encoding='utf-8'))
 

core/backend/embeddings.go

@@ -10,10 +10,6 @@ import (
 )
 
 func ModelEmbedding(s string, tokens []int, loader *model.ModelLoader, backendConfig config.BackendConfig, appConfig *config.ApplicationConfig) (func() ([]float32, error), error) {
-	if !backendConfig.Embeddings {
-		return nil, fmt.Errorf("endpoint disabled for this model by API configuration")
-	}
-
 	modelFile := backendConfig.Model
 
 	grpcOpts := gRPCModelOpts(backendConfig)

core/backend/tts.go

@@ -29,7 +29,7 @@ func generateUniqueFileName(dir, baseName, ext string) string {
 	}
 }
 
-func ModelTTS(backend, text, modelFile string, loader *model.ModelLoader, appConfig *config.ApplicationConfig, backendConfig config.BackendConfig) (string, *proto.Result, error) {
+func ModelTTS(backend, text, modelFile, voice string, loader *model.ModelLoader, appConfig *config.ApplicationConfig, backendConfig config.BackendConfig) (string, *proto.Result, error) {
 	bb := backend
 	if bb == "" {
 		bb = model.PiperBackend
@@ -44,12 +44,12 @@ func ModelTTS(backend, text, modelFile string, loader *model.ModelLoader, appCon
 		model.WithAssetDir(appConfig.AssetsDestination),
 		model.WithLoadGRPCLoadModelOpts(grpcOpts),
 	})
-	piperModel, err := loader.BackendLoader(opts...)
+	ttsModel, err := loader.BackendLoader(opts...)
 	if err != nil {
 		return "", nil, err
 	}
 
-	if piperModel == nil {
+	if ttsModel == nil {
 		return "", nil, fmt.Errorf("could not load piper model")
 	}
 
@@ -57,25 +57,31 @@ func ModelTTS(backend, text, modelFile string, loader *model.ModelLoader, appCon
 		return "", nil, fmt.Errorf("failed creating audio directory: %s", err)
 	}
 
-	fileName := generateUniqueFileName(appConfig.AudioDir, "piper", ".wav")
+	fileName := generateUniqueFileName(appConfig.AudioDir, "tts", ".wav")
 	filePath := filepath.Join(appConfig.AudioDir, fileName)
 
 	// If the model file is not empty, we pass it joined with the model path
 	modelPath := ""
 	if modelFile != "" {
-		if bb != model.TransformersMusicGen {
-			modelPath = filepath.Join(loader.ModelPath, modelFile)
-			if err := utils.VerifyPath(modelPath, appConfig.ModelPath); err != nil {
+		// If the model file is not empty, we pass it joined with the model path
+		// Checking first that it exists and is not outside ModelPath
+		// TODO: we should actually first check if the modelFile is looking like
+		// a FS path
+		mp := filepath.Join(loader.ModelPath, modelFile)
+		if _, err := os.Stat(mp); err == nil {
+			if err := utils.VerifyPath(mp, appConfig.ModelPath); err != nil {
 				return "", nil, err
 			}
+			modelPath = mp
 		} else {
 			modelPath = modelFile
 		}
 	}
 
-	res, err := piperModel.TTS(context.Background(), &proto.TTSRequest{
+	res, err := ttsModel.TTS(context.Background(), &proto.TTSRequest{
 		Text:  text,
 		Model: modelPath,
+		Voice: voice,
 		Dst:   filePath,
 	})
 

core/config/backend_config.go

@@ -276,8 +276,12 @@ func (cfg *BackendConfig) SetDefaults(debug bool, threads, ctx int, f16 bool) {
 		cfg.F16 = &f16
 	}
 
+	if cfg.Debug == nil {
+		cfg.Debug = &falseV
+	}
+
 	if debug {
-		cfg.Debug = &debug
+		cfg.Debug = &trueV
 	}
 }
 

core/http/api.go

@@ -6,6 +6,7 @@ import (
 	"os"
 	"strings"
 
+	"github.com/go-skynet/LocalAI/core/http/endpoints/elevenlabs"
 	"github.com/go-skynet/LocalAI/core/http/endpoints/localai"
 	"github.com/go-skynet/LocalAI/core/http/endpoints/openai"
 
@@ -21,6 +22,24 @@ import (
 	"github.com/gofiber/fiber/v2/middleware/recover"
 )
 
+func readAuthHeader(c *fiber.Ctx) string {
+	authHeader := c.Get("Authorization")
+
+	// elevenlabs
+	xApiKey := c.Get("xi-api-key")
+	if xApiKey != "" {
+		authHeader = "Bearer " + xApiKey
+	}
+
+	// anthropic
+	xApiKey = c.Get("x-api-key")
+	if xApiKey != "" {
+		authHeader = "Bearer " + xApiKey
+	}
+
+	return authHeader
+}
+
 func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) (*fiber.App, error) {
 	// Return errors as JSON responses
 	app := fiber.New(fiber.Config{
@@ -94,10 +113,12 @@ func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *confi
 			return c.Next()
 		}
 
-		authHeader := c.Get("Authorization")
+		authHeader := readAuthHeader(c)
 		if authHeader == "" {
 			return c.Status(fiber.StatusUnauthorized).JSON(fiber.Map{"message": "Authorization header missing"})
 		}
+
+		// If it's a bearer token
 		authHeaderParts := strings.Split(authHeader, " ")
 		if len(authHeaderParts) != 2 || authHeaderParts[0] != "Bearer" {
 			return c.Status(fiber.StatusUnauthorized).JSON(fiber.Map{"message": "Invalid Authorization header format"})
@@ -111,7 +132,6 @@ func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *confi
 		}
 
 		return c.Status(fiber.StatusUnauthorized).JSON(fiber.Map{"message": "Invalid API key"})
-
 	}
 
 	if appConfig.CORS {
@@ -147,6 +167,11 @@ func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *confi
 	app.Get("/models/jobs/:uuid", auth, modelGalleryEndpointService.GetOpStatusEndpoint())
 	app.Get("/models/jobs", auth, modelGalleryEndpointService.GetAllStatusEndpoint())
 
+	app.Post("/tts", auth, localai.TTSEndpoint(cl, ml, appConfig))
+
+	// Elevenlabs
+	app.Post("/v1/text-to-speech/:voice-id", auth, elevenlabs.TTSEndpoint(cl, ml, appConfig))
+
 	// openAI compatible API endpoint
 
 	// chat
@@ -181,7 +206,7 @@ func App(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *confi
 
 	// audio
 	app.Post("/v1/audio/transcriptions", auth, openai.TranscriptEndpoint(cl, ml, appConfig))
-	app.Post("/tts", auth, localai.TTSEndpoint(cl, ml, appConfig))
+	app.Post("/v1/audio/speech", auth, localai.TTSEndpoint(cl, ml, appConfig))
 
 	// images
 	app.Post("/v1/images/generations", auth, openai.ImageEndpoint(cl, ml, appConfig))

core/http/api_test.go

@@ -666,15 +666,15 @@ var _ = Describe("API test", func() {
 			Expect(err).ToNot(HaveOccurred())
 			Expect(len(models.Models)).To(Equal(6)) // If "config.yaml" should be included, this should be 8?
 		})
-		It("can generate completions", func() {
-			resp, err := client.CreateCompletion(context.TODO(), openai.CompletionRequest{Model: "testmodel", Prompt: testPrompt})
+		It("can generate completions via ggml", func() {
+			resp, err := client.CreateCompletion(context.TODO(), openai.CompletionRequest{Model: "testmodel.ggml", Prompt: testPrompt})
 			Expect(err).ToNot(HaveOccurred())
 			Expect(len(resp.Choices)).To(Equal(1))
 			Expect(resp.Choices[0].Text).ToNot(BeEmpty())
 		})
 
-		It("can generate chat completions ", func() {
-			resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "testmodel", Messages: []openai.ChatCompletionMessage{openai.ChatCompletionMessage{Role: "user", Content: testPrompt}}})
+		It("can generate chat completions via ggml", func() {
+			resp, err := client.CreateChatCompletion(context.TODO(), openai.ChatCompletionRequest{Model: "testmodel.ggml", Messages: []openai.ChatCompletionMessage{openai.ChatCompletionMessage{Role: "user", Content: testPrompt}}})
 			Expect(err).ToNot(HaveOccurred())
 			Expect(len(resp.Choices)).To(Equal(1))
 			Expect(resp.Choices[0].Message.Content).ToNot(BeEmpty())

core/http/endpoints/elevenlabs/tts.go

@@ -0,0 +1,55 @@
+package elevenlabs
+
+import (
+	"github.com/go-skynet/LocalAI/core/backend"
+	"github.com/go-skynet/LocalAI/core/config"
+	fiberContext "github.com/go-skynet/LocalAI/core/http/ctx"
+	"github.com/go-skynet/LocalAI/pkg/model"
+
+	"github.com/go-skynet/LocalAI/core/schema"
+	"github.com/gofiber/fiber/v2"
+	"github.com/rs/zerolog/log"
+)
+
+func TTSEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfig *config.ApplicationConfig) func(c *fiber.Ctx) error {
+	return func(c *fiber.Ctx) error {
+
+		input := new(schema.ElevenLabsTTSRequest)
+		voiceID := c.Params("voice-id")
+
+		// Get input data from the request body
+		if err := c.BodyParser(input); err != nil {
+			return err
+		}
+
+		modelFile, err := fiberContext.ModelFromContext(c, ml, input.ModelID, false)
+		if err != nil {
+			modelFile = input.ModelID
+			log.Warn().Msgf("Model not found in context: %s", input.ModelID)
+		}
+
+		cfg, err := cl.LoadBackendConfigFileByName(modelFile, appConfig.ModelPath,
+			config.LoadOptionDebug(appConfig.Debug),
+			config.LoadOptionThreads(appConfig.Threads),
+			config.LoadOptionContextSize(appConfig.ContextSize),
+			config.LoadOptionF16(appConfig.F16),
+		)
+		if err != nil {
+			modelFile = input.ModelID
+			log.Warn().Msgf("Model not found in context: %s", input.ModelID)
+		} else {
+			if input.ModelID != "" {
+				modelFile = input.ModelID
+			} else {
+				modelFile = cfg.Model
+			}
+		}
+		log.Debug().Msgf("Request for model: %s", modelFile)
+
+		filePath, _, err := backend.ModelTTS(cfg.Backend, input.Text, modelFile, voiceID, ml, appConfig, *cfg)
+		if err != nil {
+			return err
+		}
+		return c.Download(filePath)
+	}
+}

core/http/endpoints/localai/tts.go

@@ -46,7 +46,7 @@ func TTSEndpoint(cl *config.BackendConfigLoader, ml *model.ModelLoader, appConfi
 			cfg.Backend = input.Backend
 		}
 
-		filePath, _, err := backend.ModelTTS(cfg.Backend, input.Input, modelFile, ml, appConfig, *cfg)
+		filePath, _, err := backend.ModelTTS(cfg.Backend, input.Input, modelFile, input.Voice, ml, appConfig, *cfg)
 		if err != nil {
 			return err
 		}

core/http/endpoints/openai/request.go

@@ -185,6 +185,14 @@ func updateRequestConfig(config *config.BackendConfig, input *schema.OpenAIReque
 		config.RepeatPenalty = input.RepeatPenalty
 	}
 
+	if input.FrequencyPenalty!= 0 {
+		config.FrequencyPenalty = input.FrequencyPenalty
+	}
+
+	if input.PresencePenalty!= 0 {
+		config.PresencePenalty = input.PresencePenalty
+	}
+
 	if input.Keep != 0 {
 		config.Keep = input.Keep
 	}

core/schema/elevenlabs.go

@@ -0,0 +1,6 @@
+package schema
+
+type ElevenLabsTTSRequest struct {
+	Text    string `json:"text" yaml:"text"`
+	ModelID string `json:"model_id" yaml:"model_id"`
+}

core/schema/localai.go

@@ -17,5 +17,6 @@ type BackendMonitorResponse struct {
 type TTSRequest struct {
 	Model   string `json:"model" yaml:"model"`
 	Input   string `json:"input" yaml:"input"`
+	Voice   string `json:"voice" yaml:"voice"`
 	Backend string `json:"backend" yaml:"backend"`
 }

core/schema/openai.go

@@ -108,7 +108,7 @@ type ChatCompletionResponseFormat struct {
 type OpenAIRequest struct {
 	PredictionOptions
 
-	Context context.Context    `json:"-"`
+	Context context.Context  `json:"-"`
 	Cancel  context.CancelFunc `json:"-"`
 
 	// whisper

core/schema/prediction.go

@@ -25,6 +25,7 @@ type PredictionOptions struct {
 	Keep          int     `json:"n_keep" yaml:"n_keep"`
 
 	FrequencyPenalty float64 `json:"frequency_penalty" yaml:"frequency_penalty"`
+	PresencePenalty  float64 `json:"presence_penalty" yaml:"presence_penalty"`
 	TFZ              float64 `json:"tfz" yaml:"tfz"`
 
 	TypicalP float64 `json:"typical_p" yaml:"typical_p"`

docs/content/docs/features/text-generation.md

@@ -272,3 +272,56 @@ curl http://localhost:8080/v1/completions -H "Content-Type: application/json" -d
    "temperature": 0.1, "top_p": 0.1
  }'
 ```
+
+### Transformers
+
+[Transformers](https://huggingface.co/docs/transformers/index) is a State-of-the-art Machine Learning library for PyTorch, TensorFlow, and JAX.
+
+LocalAI has a built-in integration with Transformers, and it can be used to run models.
+
+This is an extra backend - in the container images (the `extra` images already contains python dependencies for Transformers) is already available and there is nothing to do for the setup.
+
+#### Setup
+
+Create a YAML file for the model you want to use with `transformers`.
+
+To setup a model, you need to just specify the model name in the YAML config file:
+```yaml
+name: transformers
+backend: transformers
+parameters:
+    model: "facebook/opt-125m"
+type: AutoModelForCausalLM
+quantization: bnb_4bit # One of: bnb_8bit, bnb_4bit, xpu_4bit (optional)
+```
+
+The backend will automatically download the required files in order to run the model.
+
+#### Parameters
+
+##### Type
+
+| Type | Description |
+| --- | --- |
+| `AutoModelForCausalLM` | `AutoModelForCausalLM` is a model that can be used to generate sequences. |
+| N/A | Defaults to `AutoModel` |
+
+
+##### Quantization
+
+| Quantization | Description |
+| --- | --- |
+| `bnb_8bit` | 8-bit quantization |
+| `bnb_4bit` | 4-bit quantization |
+| `xpu_4bit` | 4-bit quantization for Intel XPUs |
+
+#### Usage
+
+Use the `completions` endpoint by specifying the `transformers` model:
+```
+curl http://localhost:8080/v1/completions -H "Content-Type: application/json" -d '{   
+   "model": "transformers",
+   "prompt": "Hello, my name is",
+   "temperature": 0.1, "top_p": 0.1
+ }'
+```

docs/content/docs/features/text-to-audio.md

@@ -6,7 +6,13 @@ weight = 11
 url = "/features/text-to-audio/"
 +++
 
-The `/tts` endpoint can be used to generate speech from text.
+## API Compatibility
+
+The LocalAI TTS API is compatible with the [OpenAI TTS API](https://platform.openai.com/docs/guides/text-to-speech) and the [Elevenlabs](https://api.elevenlabs.io/docs) API.
+
+## LocalAI API
+
+The `/tts` endpoint can also be used to generate speech from text.
 
 ## Usage
 

docs/data/version.json

@@ -1,3 +1,3 @@
 {
-  "version": "v2.9.0"
+  "version": "v2.10.0"
 }

main.go

@@ -50,7 +50,7 @@ func main() {
 	app := &cli.App{
 		Name:    "LocalAI",
 		Version: internal.PrintableVersion(),
-		Usage:   "OpenAI compatible API for running LLaMA/GPT models locally on CPU with consumer grade hardware.",
+		Usage:   "OpenAI, OSS alternative. Drop-in compatible API for running LLM, GPT and genAI models locally on CPU, GPUs with consumer grade hardware. Supported server endpoints: OpenAI, Elevenlabs",
 		Flags: []cli.Flag{
 			&cli.BoolFlag{
 				Name:    "f16",
@@ -306,11 +306,16 @@ For a list of compatible model, check out: https://localai.io/model-compatibilit
 				return fmt.Errorf("failed basic startup tasks with error %s", err.Error())
 			}
 
-			closeConfigWatcherFn, err := startup.WatchConfigDirectory(ctx.String("localai-config-dir"), options)
-			defer closeConfigWatcherFn()
+			configdir := ctx.String("localai-config-dir")
+			// Watch the configuration directory
+			// If the directory does not exist, we don't watch it
+			if _, err := os.Stat(configdir); err == nil {
+				closeConfigWatcherFn, err := startup.WatchConfigDirectory(ctx.String("localai-config-dir"), options)
+				defer closeConfigWatcherFn()
 
-			if err != nil {
-				return fmt.Errorf("failed while watching configuration directory %s", ctx.String("localai-config-dir"))
+				if err != nil {
+					return fmt.Errorf("failed while watching configuration directory %s", ctx.String("localai-config-dir"))
+				}
 			}
 
 			appHTTP, err := http.App(cl, ml, options)
@@ -394,6 +399,12 @@ For a list of compatible model, check out: https://localai.io/model-compatibilit
 						Usage:    "Model name to run the TTS",
 						Required: true,
 					},
+					&cli.StringFlag{
+						Name:     "voice",
+						Aliases:  []string{"v"},
+						Usage:    "Voice name to run the TTS (optional)",
+						Required: true,
+					},
 					&cli.StringFlag{
 						Name:    "output-file",
 						Aliases: []string{"o"},
@@ -427,7 +438,7 @@ For a list of compatible model, check out: https://localai.io/model-compatibilit
 
 					defer ml.StopAllGRPC()
 
-					filePath, _, err := backend.ModelTTS(backendOption, text, modelOption, ml, opts, config.BackendConfig{})
+					filePath, _, err := backend.ModelTTS(backendOption, text, modelOption, ctx.String("voice"), ml, opts, config.BackendConfig{})
 					if err != nil {
 						return err
 					}

pkg/grpc/proto/backend.pb.go

@@ -1,7 +1,7 @@
 // Code generated by protoc-gen-go. DO NOT EDIT.
 // versions:
 // 	protoc-gen-go v1.26.0
-// 	protoc        v4.25.3
+// 	protoc        v4.23.4
 // source: backend.proto
 
 package proto
@@ -1436,6 +1436,7 @@ type TTSRequest struct {
 	Text  string `protobuf:"bytes,1,opt,name=text,proto3" json:"text,omitempty"`
 	Model string `protobuf:"bytes,2,opt,name=model,proto3" json:"model,omitempty"`
 	Dst   string `protobuf:"bytes,3,opt,name=dst,proto3" json:"dst,omitempty"`
+	Voice string `protobuf:"bytes,4,opt,name=voice,proto3" json:"voice,omitempty"`
 }
 
 func (x *TTSRequest) Reset() {
@@ -1491,6 +1492,13 @@ func (x *TTSRequest) GetDst() string {
 	return ""
 }
 
+func (x *TTSRequest) GetVoice() string {
+	if x != nil {
+		return x.Voice
+	}
+	return ""
+}
+
 type TokenizationResponse struct {
 	state         protoimpl.MessageState
 	sizeCache     protoimpl.SizeCache
@@ -1896,84 +1904,86 @@ var file_backend_proto_rawDesc = []byte{
 	0x28, 0x09, 0x52, 0x10, 0x45, 0x6e, 0x61, 0x62, 0x6c, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x65,
 	0x74, 0x65, 0x72, 0x73, 0x12, 0x1a, 0x0a, 0x08, 0x43, 0x4c, 0x49, 0x50, 0x53, 0x6b, 0x69, 0x70,
 	0x18, 0x0b, 0x20, 0x01, 0x28, 0x05, 0x52, 0x08, 0x43, 0x4c, 0x49, 0x50, 0x53, 0x6b, 0x69, 0x70,
-	0x22, 0x48, 0x0a, 0x0a, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12,
+	0x22, 0x5e, 0x0a, 0x0a, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x12, 0x12,
 	0x0a, 0x04, 0x74, 0x65, 0x78, 0x74, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x04, 0x74, 0x65,
 	0x78, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x18, 0x02, 0x20, 0x01, 0x28,
 	0x09, 0x52, 0x05, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x10, 0x0a, 0x03, 0x64, 0x73, 0x74, 0x18,
-	0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x22, 0x46, 0x0a, 0x14, 0x54, 0x6f,
-	0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e,
-	0x73, 0x65, 0x12, 0x16, 0x0a, 0x06, 0x6c, 0x65, 0x6e, 0x67, 0x74, 0x68, 0x18, 0x01, 0x20, 0x01,
-	0x28, 0x05, 0x52, 0x06, 0x6c, 0x65, 0x6e, 0x67, 0x74, 0x68, 0x12, 0x16, 0x0a, 0x06, 0x74, 0x6f,
-	0x6b, 0x65, 0x6e, 0x73, 0x18, 0x02, 0x20, 0x03, 0x28, 0x05, 0x52, 0x06, 0x74, 0x6f, 0x6b, 0x65,
-	0x6e, 0x73, 0x22, 0xac, 0x01, 0x0a, 0x0f, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x73, 0x61,
-	0x67, 0x65, 0x44, 0x61, 0x74, 0x61, 0x12, 0x14, 0x0a, 0x05, 0x74, 0x6f, 0x74, 0x61, 0x6c, 0x18,
-	0x01, 0x20, 0x01, 0x28, 0x04, 0x52, 0x05, 0x74, 0x6f, 0x74, 0x61, 0x6c, 0x12, 0x45, 0x0a, 0x09,
-	0x62, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x18, 0x02, 0x20, 0x03, 0x28, 0x0b, 0x32,
-	0x27, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79,
-	0x55, 0x73, 0x61, 0x67, 0x65, 0x44, 0x61, 0x74, 0x61, 0x2e, 0x42, 0x72, 0x65, 0x61, 0x6b, 0x64,
-	0x6f, 0x77, 0x6e, 0x45, 0x6e, 0x74, 0x72, 0x79, 0x52, 0x09, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x64,
-	0x6f, 0x77, 0x6e, 0x1a, 0x3c, 0x0a, 0x0e, 0x42, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e,
-	0x45, 0x6e, 0x74, 0x72, 0x79, 0x12, 0x10, 0x0a, 0x03, 0x6b, 0x65, 0x79, 0x18, 0x01, 0x20, 0x01,
-	0x28, 0x09, 0x52, 0x03, 0x6b, 0x65, 0x79, 0x12, 0x14, 0x0a, 0x05, 0x76, 0x61, 0x6c, 0x75, 0x65,
-	0x18, 0x02, 0x20, 0x01, 0x28, 0x04, 0x52, 0x05, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3a, 0x02, 0x38,
-	0x01, 0x22, 0xbc, 0x01, 0x0a, 0x0e, 0x53, 0x74, 0x61, 0x74, 0x75, 0x73, 0x52, 0x65, 0x73, 0x70,
-	0x6f, 0x6e, 0x73, 0x65, 0x12, 0x33, 0x0a, 0x05, 0x73, 0x74, 0x61, 0x74, 0x65, 0x18, 0x01, 0x20,
-	0x01, 0x28, 0x0e, 0x32, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74,
-	0x61, 0x74, 0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x2e, 0x53, 0x74, 0x61,
-	0x74, 0x65, 0x52, 0x05, 0x73, 0x74, 0x61, 0x74, 0x65, 0x12, 0x30, 0x0a, 0x06, 0x6d, 0x65, 0x6d,
-	0x6f, 0x72, 0x79, 0x18, 0x02, 0x20, 0x01, 0x28, 0x0b, 0x32, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b,
-	0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x73, 0x61, 0x67, 0x65, 0x44,
-	0x61, 0x74, 0x61, 0x52, 0x06, 0x6d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x22, 0x43, 0x0a, 0x05, 0x53,
-	0x74, 0x61, 0x74, 0x65, 0x12, 0x11, 0x0a, 0x0d, 0x55, 0x4e, 0x49, 0x4e, 0x49, 0x54, 0x49, 0x41,
-	0x4c, 0x49, 0x5a, 0x45, 0x44, 0x10, 0x00, 0x12, 0x08, 0x0a, 0x04, 0x42, 0x55, 0x53, 0x59, 0x10,
-	0x01, 0x12, 0x09, 0x0a, 0x05, 0x52, 0x45, 0x41, 0x44, 0x59, 0x10, 0x02, 0x12, 0x12, 0x0a, 0x05,
-	0x45, 0x52, 0x52, 0x4f, 0x52, 0x10, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x01,
-	0x32, 0xf4, 0x04, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x12, 0x32, 0x0a, 0x06,
-	0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
-	0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x1a, 0x0e,
-	0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00,
-	0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x12, 0x17, 0x2e, 0x62, 0x61,
-	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74,
-	0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52,
-	0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a, 0x09, 0x4c, 0x6f, 0x61, 0x64, 0x4d, 0x6f,
-	0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x6f,
-	0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63,
-	0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x3c, 0x0a,
-	0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53, 0x74, 0x72, 0x65, 0x61, 0x6d, 0x12, 0x17,
+	0x03, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x64, 0x73, 0x74, 0x12, 0x14, 0x0a, 0x05, 0x76, 0x6f,
+	0x69, 0x63, 0x65, 0x18, 0x04, 0x20, 0x01, 0x28, 0x09, 0x52, 0x05, 0x76, 0x6f, 0x69, 0x63, 0x65,
+	0x22, 0x46, 0x0a, 0x14, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e,
+	0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x12, 0x16, 0x0a, 0x06, 0x6c, 0x65, 0x6e, 0x67,
+	0x74, 0x68, 0x18, 0x01, 0x20, 0x01, 0x28, 0x05, 0x52, 0x06, 0x6c, 0x65, 0x6e, 0x67, 0x74, 0x68,
+	0x12, 0x16, 0x0a, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x18, 0x02, 0x20, 0x03, 0x28, 0x05,
+	0x52, 0x06, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x22, 0xac, 0x01, 0x0a, 0x0f, 0x4d, 0x65, 0x6d,
+	0x6f, 0x72, 0x79, 0x55, 0x73, 0x61, 0x67, 0x65, 0x44, 0x61, 0x74, 0x61, 0x12, 0x14, 0x0a, 0x05,
+	0x74, 0x6f, 0x74, 0x61, 0x6c, 0x18, 0x01, 0x20, 0x01, 0x28, 0x04, 0x52, 0x05, 0x74, 0x6f, 0x74,
+	0x61, 0x6c, 0x12, 0x45, 0x0a, 0x09, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x18,
+	0x02, 0x20, 0x03, 0x28, 0x0b, 0x32, 0x27, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e,
+	0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x55, 0x73, 0x61, 0x67, 0x65, 0x44, 0x61, 0x74, 0x61, 0x2e,
+	0x42, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x45, 0x6e, 0x74, 0x72, 0x79, 0x52, 0x09,
+	0x62, 0x72, 0x65, 0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x1a, 0x3c, 0x0a, 0x0e, 0x42, 0x72, 0x65,
+	0x61, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x45, 0x6e, 0x74, 0x72, 0x79, 0x12, 0x10, 0x0a, 0x03, 0x6b,
+	0x65, 0x79, 0x18, 0x01, 0x20, 0x01, 0x28, 0x09, 0x52, 0x03, 0x6b, 0x65, 0x79, 0x12, 0x14, 0x0a,
+	0x05, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x18, 0x02, 0x20, 0x01, 0x28, 0x04, 0x52, 0x05, 0x76, 0x61,
+	0x6c, 0x75, 0x65, 0x3a, 0x02, 0x38, 0x01, 0x22, 0xbc, 0x01, 0x0a, 0x0e, 0x53, 0x74, 0x61, 0x74,
+	0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x12, 0x33, 0x0a, 0x05, 0x73, 0x74,
+	0x61, 0x74, 0x65, 0x18, 0x01, 0x20, 0x01, 0x28, 0x0e, 0x32, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b,
+	0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x61, 0x74, 0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e,
+	0x73, 0x65, 0x2e, 0x53, 0x74, 0x61, 0x74, 0x65, 0x52, 0x05, 0x73, 0x74, 0x61, 0x74, 0x65, 0x12,
+	0x30, 0x0a, 0x06, 0x6d, 0x65, 0x6d, 0x6f, 0x72, 0x79, 0x18, 0x02, 0x20, 0x01, 0x28, 0x0b, 0x32,
+	0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x65, 0x6d, 0x6f, 0x72, 0x79,
+	0x55, 0x73, 0x61, 0x67, 0x65, 0x44, 0x61, 0x74, 0x61, 0x52, 0x06, 0x6d, 0x65, 0x6d, 0x6f, 0x72,
+	0x79, 0x22, 0x43, 0x0a, 0x05, 0x53, 0x74, 0x61, 0x74, 0x65, 0x12, 0x11, 0x0a, 0x0d, 0x55, 0x4e,
+	0x49, 0x4e, 0x49, 0x54, 0x49, 0x41, 0x4c, 0x49, 0x5a, 0x45, 0x44, 0x10, 0x00, 0x12, 0x08, 0x0a,
+	0x04, 0x42, 0x55, 0x53, 0x59, 0x10, 0x01, 0x12, 0x09, 0x0a, 0x05, 0x52, 0x45, 0x41, 0x44, 0x59,
+	0x10, 0x02, 0x12, 0x12, 0x0a, 0x05, 0x45, 0x52, 0x52, 0x4f, 0x52, 0x10, 0xff, 0xff, 0xff, 0xff,
+	0xff, 0xff, 0xff, 0xff, 0xff, 0x01, 0x32, 0xf4, 0x04, 0x0a, 0x07, 0x42, 0x61, 0x63, 0x6b, 0x65,
+	0x6e, 0x64, 0x12, 0x32, 0x0a, 0x06, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x12, 0x16, 0x2e, 0x62,
+	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73,
+	0x73, 0x61, 0x67, 0x65, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52,
+	0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x34, 0x0a, 0x07, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63,
+	0x74, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64,
+	0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63,
+	0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x12, 0x35, 0x0a, 0x09,
+	0x4c, 0x6f, 0x61, 0x64, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x12, 0x15, 0x2e, 0x62, 0x61, 0x63, 0x6b,
+	0x65, 0x6e, 0x64, 0x2e, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73,
+	0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c,
+	0x74, 0x22, 0x00, 0x12, 0x3c, 0x0a, 0x0d, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x53, 0x74,
+	0x72, 0x65, 0x61, 0x6d, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50,
+	0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e,
+	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x30,
+	0x01, 0x12, 0x40, 0x0a, 0x09, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12, 0x17,
 	0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74,
-	0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x0e, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
-	0x64, 0x2e, 0x52, 0x65, 0x70, 0x6c, 0x79, 0x22, 0x00, 0x30, 0x01, 0x12, 0x40, 0x0a, 0x09, 0x45,
-	0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
-	0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e,
-	0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x45, 0x6d, 0x62, 0x65,
-	0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x41, 0x0a,
-	0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x12, 0x1d,
-	0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74,
-	0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e,
-	0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00,
-	0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72,
-	0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
-	0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x71, 0x75, 0x65,
-	0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61,
-	0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12,
-	0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
-	0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61,
-	0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x4a,
-	0x0a, 0x0e, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x65, 0x53, 0x74, 0x72, 0x69, 0x6e, 0x67,
-	0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x50, 0x72, 0x65, 0x64, 0x69,
-	0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b,
-	0x65, 0x6e, 0x64, 0x2e, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e,
-	0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00, 0x12, 0x3b, 0x0a, 0x06, 0x53, 0x74,
-	0x61, 0x74, 0x75, 0x73, 0x12, 0x16, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48,
-	0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x1a, 0x17, 0x2e, 0x62,
-	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x61, 0x74, 0x75, 0x73, 0x52, 0x65, 0x73,
-	0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00, 0x42, 0x5a, 0x0a, 0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b,
-	0x79, 0x6e, 0x65, 0x74, 0x2e, 0x6c, 0x6f, 0x63, 0x61, 0x6c, 0x61, 0x69, 0x2e, 0x62, 0x61, 0x63,
-	0x6b, 0x65, 0x6e, 0x64, 0x42, 0x0e, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x42, 0x61, 0x63,
-	0x6b, 0x65, 0x6e, 0x64, 0x50, 0x01, 0x5a, 0x2b, 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63,
-	0x6f, 0x6d, 0x2f, 0x67, 0x6f, 0x2d, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63,
-	0x61, 0x6c, 0x41, 0x49, 0x2f, 0x70, 0x6b, 0x67, 0x2f, 0x67, 0x72, 0x70, 0x63, 0x2f, 0x70, 0x72,
-	0x6f, 0x74, 0x6f, 0x62, 0x06, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x33,
+	0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a, 0x18, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e,
+	0x64, 0x2e, 0x45, 0x6d, 0x62, 0x65, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x52, 0x65, 0x73, 0x75, 0x6c,
+	0x74, 0x22, 0x00, 0x12, 0x41, 0x0a, 0x0d, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49,
+	0x6d, 0x61, 0x67, 0x65, 0x12, 0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x47,
+	0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x65, 0x49, 0x6d, 0x61, 0x67, 0x65, 0x52, 0x65, 0x71, 0x75,
+	0x65, 0x73, 0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65,
+	0x73, 0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x4d, 0x0a, 0x12, 0x41, 0x75, 0x64, 0x69, 0x6f, 0x54,
+	0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x12, 0x1a, 0x2e, 0x62,
+	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70,
+	0x74, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74, 0x1a, 0x19, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65,
+	0x6e, 0x64, 0x2e, 0x54, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x52, 0x65, 0x73,
+	0x75, 0x6c, 0x74, 0x22, 0x00, 0x12, 0x2d, 0x0a, 0x03, 0x54, 0x54, 0x53, 0x12, 0x13, 0x2e, 0x62,
+	0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x54, 0x53, 0x52, 0x65, 0x71, 0x75, 0x65, 0x73,
+	0x74, 0x1a, 0x0f, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x52, 0x65, 0x73, 0x75,
+	0x6c, 0x74, 0x22, 0x00, 0x12, 0x4a, 0x0a, 0x0e, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69, 0x7a, 0x65,
+	0x53, 0x74, 0x72, 0x69, 0x6e, 0x67, 0x12, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64,
+	0x2e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x4f, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x1a,
+	0x1d, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x54, 0x6f, 0x6b, 0x65, 0x6e, 0x69,
+	0x7a, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00,
+	0x12, 0x3b, 0x0a, 0x06, 0x53, 0x74, 0x61, 0x74, 0x75, 0x73, 0x12, 0x16, 0x2e, 0x62, 0x61, 0x63,
+	0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x48, 0x65, 0x61, 0x6c, 0x74, 0x68, 0x4d, 0x65, 0x73, 0x73, 0x61,
+	0x67, 0x65, 0x1a, 0x17, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x2e, 0x53, 0x74, 0x61,
+	0x74, 0x75, 0x73, 0x52, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x22, 0x00, 0x42, 0x5a, 0x0a,
+	0x19, 0x69, 0x6f, 0x2e, 0x73, 0x6b, 0x79, 0x6e, 0x65, 0x74, 0x2e, 0x6c, 0x6f, 0x63, 0x61, 0x6c,
+	0x61, 0x69, 0x2e, 0x62, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x42, 0x0e, 0x4c, 0x6f, 0x63, 0x61,
+	0x6c, 0x41, 0x49, 0x42, 0x61, 0x63, 0x6b, 0x65, 0x6e, 0x64, 0x50, 0x01, 0x5a, 0x2b, 0x67, 0x69,
+	0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d, 0x2f, 0x67, 0x6f, 0x2d, 0x73, 0x6b, 0x79, 0x6e,
+	0x65, 0x74, 0x2f, 0x4c, 0x6f, 0x63, 0x61, 0x6c, 0x41, 0x49, 0x2f, 0x70, 0x6b, 0x67, 0x2f, 0x67,
+	0x72, 0x70, 0x63, 0x2f, 0x70, 0x72, 0x6f, 0x74, 0x6f, 0x62, 0x06, 0x70, 0x72, 0x6f, 0x74, 0x6f,
+	0x33,
 }
 
 var (

pkg/grpc/proto/backend_grpc.pb.go

@@ -1,7 +1,7 @@
 // Code generated by protoc-gen-go-grpc. DO NOT EDIT.
 // versions:
 // - protoc-gen-go-grpc v1.2.0
-// - protoc             v4.25.3
+// - protoc             v4.23.4
 // source: backend.proto
 
 package proto

pkg/model/initializers.go

@@ -15,12 +15,11 @@ import (
 )
 
 var Aliases map[string]string = map[string]string{
-	"go-llama": GoLlamaBackend,
+	"go-llama": LLamaCPP,
 	"llama":    LLamaCPP,
 }
 
 const (
-	GoLlamaBackend      = "llama"
 	LlamaGGML           = "llama-ggml"
 	LLamaCPP            = "llama-cpp"
 	Gpt4AllLlamaBackend = "gpt4all-llama"
@@ -35,15 +34,11 @@ const (
 	TinyDreamBackend       = "tinydream"
 	PiperBackend           = "piper"
 	LCHuggingFaceBackend   = "langchain-huggingface"
-
-	// External Backends that need special handling within LocalAI:
-	TransformersMusicGen = "transformers-musicgen"
 )
 
 var AutoLoadBackends []string = []string{
 	LLamaCPP,
 	LlamaGGML,
-	GoLlamaBackend,
 	Gpt4All,
 	BertEmbeddingsBackend,
 	RwkvBackend,

tests/models_fixtures/config.yaml

@@ -1,6 +1,6 @@
 - name: list1
   parameters:
-    model: testmodel
+    model: testmodel.ggml
     top_p: 80
     top_k: 0.9
     temperature: 0.1
@@ -19,7 +19,7 @@
     top_p: 80
     top_k: 0.9
     temperature: 0.1
-    model: testmodel
+    model: testmodel.ggml
   context_size: 200
   stopwords:
   - "HUMAN:"

tests/models_fixtures/gpt4.yaml

@@ -1,6 +1,6 @@
 name: gpt4all
 parameters:
-  model: testmodel
+  model: testmodel.ggml
   top_p: 80
   top_k: 0.9
   temperature: 0.1

tests/models_fixtures/gpt4_2.yaml

@@ -1,6 +1,6 @@
 name: gpt4all-2
 parameters:
-  model: testmodel
+  model: testmodel.ggml
   top_p: 80
   top_k: 0.9
   temperature: 0.1