Support for OpenVINO

BUILD.bazel

@@ -55,6 +55,7 @@ COPTS = [
     "-std=c++11",
     "-fPIC",
     "-Wall",
+    "-DWHISPER_USE_OPENVINO",
 ] + selects.with_or({
     "//conditions:default": [],
     "@bazel_tools//src/conditions:linux_x86_64": [

extern/openvino.BUILD

@@ -0,0 +1,54 @@
+load("@bazel_skylib//lib:selects.bzl", "selects")
+
+package(
+	default_visibility = ["//visibility:public"],
+)
+
+cc_library(
+	name = "openvino_old_headers",
+	hdrs = glob([
+		"include/ie/**/*.*"
+	]),
+	strip_include_prefix = "include/ie",
+	visibility = ["//visibility:public"],
+)
+
+cc_library(
+	name = "openvino_new_headers",
+	hdrs = glob([
+		#"include/openvino/**/*.*"
+		"include/**/*.*"
+	]),
+	strip_include_prefix = "include",
+	visibility = ["//visibility:public"],
+	deps = [
+		"@linux_openvino//:openvino_old_headers",
+	]
+)
+
+cc_library(
+	name = "ngraph",
+	hdrs = glob([
+		"include/ngraph/**/*.*"
+	]),
+	strip_include_prefix = "include",
+	visibility = ["//visibility:public"],
+)
+
+cc_library(
+	name = "openvino",
+    srcs = selects.with_or({
+        "@bazel_tools//src/conditions:linux_x86_64": [
+            "lib/intel64/libopenvino.so",
+        ],
+        "@bazel_tools//src/conditions:linux_aarch64": [
+            "lib/aarch64/libopenvino.so",
+        ],
+    }),
+	strip_include_prefix = "include/ie",
+	visibility = ["//visibility:public"],
+	deps = [
+		"@linux_openvino//:ngraph",
+		"@linux_openvino//:openvino_new_headers",
+    ],
+)

extern/whispercpp.BUILD

@@ -17,6 +17,9 @@ HEADERS = [
     "whisper.h",
 ]
 
+OPENVINO_HEADERS = glob(["openvino/*.h"])
+OPENVINO_SOURCES = glob(["openvino/*.cpp"])
+
 EXAMPLE_HEADERS = [
     "examples/common-sdl.h",
     "examples/common.h",
@@ -30,6 +33,10 @@ CFLAGS = [
     "-std=c11",
     "-fPIC",
     "-pthread",
+    "-DWHISPER_USE_OPENVINO",
+    "-D_POSIX_C_SOURCE=200809L",
+    "-D_USE_MATH_DEFINES",
+    "-D_GNU_SOURCE",
 ]
 
 CXXFLAGS = [
@@ -39,6 +46,10 @@ CXXFLAGS = [
     "-std=c++11",
     "-fPIC",
     "-pthread",
+    "-DWHISPER_USE_OPENVINO",
+    "-D_POSIX_C_SOURCE=200809L",
+    "-D_USE_MATH_DEFINES",
+    "-D_GNU_SOURCE",
 ]
 
 cc_library(
@@ -90,7 +101,9 @@ cc_library(
 
 cc_library(
     name = "ggml",
-    srcs = ["ggml.c"],
+    srcs = [
+        "ggml.c",
+    ],
     hdrs = HEADERS,
     copts = CFLAGS + selects.with_or({
         "//conditions:default": [],
@@ -113,8 +126,8 @@ cc_library(
 
 cc_library(
     name = "whisper",
-    srcs = ["whisper.cpp"],
-    hdrs = HEADERS + EXAMPLE_HEADERS,
+    srcs = ["whisper.cpp"] + OPENVINO_SOURCES,
+    hdrs = HEADERS + EXAMPLE_HEADERS + OPENVINO_HEADERS,
     copts = CXXFLAGS + selects.with_or({
         "//conditions:default": [],
         "@bazel_tools//src/conditions:linux_x86_64": [
@@ -132,5 +145,9 @@ cc_library(
             "Accelerate",
         ],
     }),
-    deps = [":ggml"],
+    deps = [
+        ":ggml",
+	"@linux_openvino//:openvino_new_headers",
+	"@linux_openvino//:openvino",
+    ],
 )

rules/deps.bzl

@@ -74,10 +74,16 @@ def internal_deps():
         init_submodules = True,
         recursive_init_submodules = True,
         remote = "https://github.com/ggerganov/whisper.cpp.git",
-        commit = "2f889132c66051b14c6f8770e9b3d4e3f159821d",
+        commit = "c5f9acf4b79780171e25b91d227c615e828568f6",
         shallow_since = "1678217790 +0200",
     )
 
+    native.new_local_repository(
+        name = "linux_openvino",
+        path = "/home/pi/code/v3/openvino/l_openvino_toolkit_debian9_2023.2.0.13089.cfd42bd2cb0_arm64/runtime/",
+        build_file = "//extern:openvino.BUILD",
+    )
+
     http_archive(
         name = "com_github_libsdl_sdl2",
         build_file = Label("//extern:sdl2.BUILD"),

src/whispercpp/api.pyi

@@ -9,7 +9,6 @@ from numpy.typing import NDArray
 
 SAMPLE_RATE: int = ...
 N_FFT: int = ...
-N_MEL: int = ...
 HOP_LENGTH: int = ...
 CHUNK_SIZE: int = ...
 

src/whispercpp/api_cpp2py_export.cc

@@ -20,7 +20,6 @@ PYBIND11_MODULE(api_cpp2py_export, m) {
     // NOTE: default attributes
     m.attr("SAMPLE_RATE") = py::int_(WHISPER_SAMPLE_RATE);
     m.attr("N_FFT") = py::int_(WHISPER_N_FFT);
-    m.attr("N_MEL") = py::int_(WHISPER_N_MEL);
     m.attr("HOP_LENGTH") = py::int_(WHISPER_HOP_LENGTH);
     m.attr("CHUNK_SIZE") = py::int_(WHISPER_CHUNK_SIZE);
 

src/whispercpp/context.cc

@@ -67,6 +67,7 @@ Context Context::from_file(const char *filename, bool no_state) {
         c.set_init_with_state(true);
     }
     RAISE_IF_NULL(c.wctx);
+    whisper_ctx_init_openvino_encoder(c.wctx, nullptr, "CPU", nullptr);
     return c;
 }
 
@@ -81,6 +82,7 @@ Context Context::from_buffer(void *buffer, size_t buffer_size, bool no_state) {
         c.set_init_with_state(true);
     }
     RAISE_IF_NULL(c.wctx);
+    whisper_ctx_init_openvino_encoder(c.wctx, nullptr, "CPU", nullptr);
     return c;
 }
 
@@ -303,6 +305,8 @@ size_t Context::n_len() {
     }
 }
 
+// Get number of mels per segment
+size_t Context::model_n_mels() { return whisper_model_n_mels(wctx); }
 // Get number of vocab
 size_t Context::n_vocab() { return whisper_n_vocab(wctx); }
 // Get number of text context

src/whispercpp/context.h

@@ -526,6 +526,7 @@ struct Context {
     // language functions. Returns a vector of probabilities for each language.
     std::vector<float> lang_detect(size_t offset_ms, size_t threads);
 
+    size_t model_n_mels();
     size_t n_len();
     size_t n_vocab();
     size_t n_text_ctx();
@@ -551,8 +552,8 @@ struct Context {
     whisper_token solm_token() { return whisper_token_solm(wctx); }
     whisper_token not_token() { return whisper_token_not(wctx); }
     whisper_token beg_token() { return whisper_token_beg(wctx); }
-    whisper_token token_translate() { return whisper_token_translate(); }
-    whisper_token token_transcribe() { return whisper_token_transcribe(); }
+    whisper_token token_translate() { return whisper_token_translate(wctx); }
+    whisper_token token_transcribe() { return whisper_token_transcribe(wctx); }
     whisper_token lang_token(int lang_id) {
         return whisper_token_lang(wctx, lang_id);
     }