dbg: fix the forward pass (#139)

.github/workflows/build.yml

@@ -4,7 +4,6 @@ on:
  push:
  branches:
  - main
- - encodec-submodule-fix-ci
  paths:
  [
  ".github/workflows/**",
@@ -56,7 +55,6 @@ jobs:
  - name: Build
  id: cmake_build
  run: |
- cd bark
  mkdir build
  cd build
  cmake ..
@@ -82,7 +80,6 @@ jobs:
  id: cmake_build
  run: |
  sysctl -a
- cd bark
  mkdir build
  cd build
  cmake ..

.gitignore

@@ -7,7 +7,6 @@
 *.wav
 *.bin
 .build/
-build/
 .cache/
 *.dSYM/
 .direnv/
@@ -18,21 +17,5 @@ build/
 .vs/
 .vscode/
 
-ggml_weights/*
-ggml_weights_*/*
-models/
-
-build-info.h
-
-!ggml_weights/ggml_vocab.bin
-!ggml_weights/vocab.txt
-
-tests/test-gpt-eval
-tests/test-fine-gpt-eval
-
-tests/test-forward-coarse
-tests/test-forward-semantic
-tests/test-forward-fine
-tests/test-forward-encodec
-
-tests/test-tokenizer
+bark_weights/
+build/

bark/CMakeLists.txt → CMakeLists.txt

@@ -1,6 +1,8 @@
 cmake_minimum_required(VERSION 3.12)
 project("bark" C CXX)
 
+set(BARK_BUILD_EXAMPLES ON)
+
 if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
  set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
  set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
@@ -9,33 +11,24 @@ endif()
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 set(CMAKE_CXX_FLAGS_RELEASE "-O3")
 
-if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
- set(BARK_STANDALONE ON)
-else()
- set(BARK_STANDALONE OFF)
-endif()
-
-option(BARK_BUILD_TESTS "bark: build tests" ${BARK_STANDALONE})
-option(BARK_BUILD_EXAMPLES "bark: build examples" ${BARK_STANDALONE})
-
-# Build libraries
-
 set(BARK_LIB bark)
 
-# add_subdirectory(../ggml ${CMAKE_BINARY_DIR}/ggml)
-add_subdirectory(../encodec.cpp ${CMAKE_BINARY_DIR}/encodec.cpp)
+add_subdirectory(encodec.cpp)
 
-add_library(${BARK_LIB} bark.cpp bark.h)
+add_library(${BARK_LIB} STATIC bark.cpp bark.h)
 
 if (BARK_BUILD_EXAMPLES)
  add_subdirectory(examples)
 endif()
 
-if (BARK_BUILD_TESTS)
- include(CTest)
- add_subdirectory(tests)
-endif ()
-
 target_link_libraries(${BARK_LIB} PUBLIC ggml encodec)
 target_include_directories(${BARK_LIB} PUBLIC .)
 target_compile_features(${BARK_LIB} PUBLIC cxx_std_11)
+
+if (GGML_CUBLAS)
+ add_compile_definitions(GGML_USE_CUBLAS)
+endif()
+
+if (GGML_METAL)
+ add_compile_definitions(GGML_USE_METAL)
+endif()

README.md

@@ -11,7 +11,7 @@ Inference of [SunoAI's bark model](https://github.com/suno-ai/bark) in pure C/C+
 
 ## Description
 
-With `bark.cpp`, my goal is to bring **real-time realistic multilingual** text-to-speech generation to the community. Currently, I am focused on porting the [Bark](https://github.com/suno-ai/bark) model in C++.
+With `bark.cpp`, our goal is to bring **real-time realistic multilingual** text-to-speech generation to the community. Currently, I am focused on porting the [Bark](https://github.com/suno-ai/bark) model in C++.
 
 - [x] Plain C/C++ implementation without dependencies
 - [x] AVX, AVX2 and AVX512 for x86 architectures
@@ -93,8 +93,8 @@ git submodule update --init --recursive
 In order to build bark.cpp you must use `CMake`:
 
 ```bash
-mkdir bark/build
-cd bark/build
+mkdir build
+cd build
 cmake ..
 cmake --build . --config Release
 ```
@@ -106,16 +106,13 @@ cmake --build . --config Release
 python3 -m pip install -r bark/requirements.txt
 
 # obtain the original bark and encodec weights and place them in ./models
-python3 bark/download_weights.py --download-dir ./models
+python3 download_weights.py --download-dir ./models
 
 # convert the model to ggml format
-python3 bark/convert.py \
- --dir-model ./models \
- --vocab-path ./ggml_weights/ \
- --out-dir ./ggml_weights/
+python3 convert.py --dir-model ./models --out-dir ./ggml_weights/
 
 # run the inference
-./bark/build/examples/main/main -m ./ggml_weights/ -p "this is an audio"
+./build/examples/main/main -m ./ggml_weights/ -p "this is an audio"
 ```
 
 ### (Optional) Quantize weights
@@ -125,11 +122,7 @@ Weights can be quantized using the following strategy: `q4_0`, `q4_1`, `q5_0`, `
 Note that to preserve audio quality, we do not quantize the codec model. The bulk of the computation is in the forward pass of the GPT models.
 
 ```bash
-mkdir ggml_weights_q4
-cp ggml_weights/*vocab* ggml_weights_q4
-./bark/build/examples/quantize/quantize ./ggml_weights/ggml_weights_text.bin ./ggml_weights_q4/ggml_weights_text.bin q4_0
-./bark/build/examples/quantize/quantize ./ggml_weights/ggml_weights_coarse.bin ./ggml_weights_q4/ggml_weights_coarse.bin q4_0
-./bark/build/examples/quantize/quantize ./ggml_weights/ggml_weights_fine.bin ./ggml_weights_q4/ggml_weights_fine.bin q4_0
+./build/examples/quantize/quantize ./ggml_weights.bin ./ggml_weights_q4.bin q4_0
 ```
 
 ### Seminal papers