Merge remote-tracking branch 'ggerganov/master'

* ggerganov/master: (49 commits)
  cli : add --suppress_nst support (ggml-org#2664)
  cli : add no_speech_thold (ggml-org#2663)
  cmake : remove hardcoded install rpath
  server : fix help print
  ruby : bug fix on callbacks and no_speech_prob (ggml-org#2656)
  server : add no-speech threshold parameter and functionality (ggml-org#2654)
  whisper : rename suppress_non_speech_tokens to suppress_nst (ggml-org#2653)
  server : add option to suppress non-speech tokens (ggml-org#2649)
  whisper : rename binaries + fix install (ggml-org#2648)
  ruby : update gem version to v1.3.1
  release : v1.7.3
  ci : msys enable SDL2 build (ggml-org#2635)
  ruby : sync ggml (ggml-org#2643)
  android : try to fix build
  files : remove old sources
  sync : ggml
  talk-llama : sync llama.cpp
  sync : ggml
  ggml : update ggml_backend_cpu_device_supports_op (llama/10867)
  vulkan: bugfixes for small subgroup size systems + llvmpipe test (llama/10809)
  ...

Author: bygreencn

.github/workflows/build.yml

@@ -283,7 +283,7 @@ jobs:
       - name: Build using CMake
         shell: msys2 {0}
         run: |
-            cmake -B build
+            cmake -B build -DWHISPER_SDL2=ON
             cmake --build build --config ${{ matrix.build }} -j $(nproc)
 
       - name: Clean after building using CMake

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.5) # for add_link_options and implicit target directories.
 project("whisper.cpp" C CXX)
-project("whisper.cpp" VERSION 1.7.2)
+project("whisper.cpp" VERSION 1.7.3)
 include(CheckIncludeFileCXX)
 
 set(SOVERSION 1)

README.md

@@ -181,11 +181,11 @@ public void suppressBlanks(boolean enable) {
     }
 
     /** Flag to suppress non-speech tokens. */
-    public CBool suppress_non_speech_tokens;
+    public CBool suppress_nst;
 
     /** Flag to suppress non-speech tokens. */
     public void suppressNonSpeechTokens(boolean enable) {
-        suppress_non_speech_tokens = enable ? CBool.TRUE : CBool.FALSE;
+        suppress_nst = enable ? CBool.TRUE : CBool.FALSE;
     }
 
     /** Initial decoding temperature. */
@@ -315,7 +315,7 @@ protected List<String> getFieldOrder() {
                 "print_special", "print_progress", "print_realtime", "print_timestamps",  "token_timestamps",
                 "thold_pt", "thold_ptsum", "max_len", "split_on_word", "max_tokens", "audio_ctx",
                 "tdrz_enable", "suppress_regex", "initial_prompt", "prompt_tokens", "prompt_n_tokens", "language", "detect_language",
-                "suppress_blank", "suppress_non_speech_tokens", "temperature", "max_initial_ts", "length_penalty",
+                "suppress_blank", "suppress_nst", "temperature", "max_initial_ts", "length_penalty",
                 "temperature_inc", "entropy_thold", "logprob_thold", "no_speech_thold", "greedy", "beam_search",
                 "new_segment_callback", "new_segment_callback_user_data",
                 "progress_callback", "progress_callback_user_data",

bindings/java/src/main/java/io/github/ggerganov/whispercpp/params/WhisperFullParams.java

@@ -1,6 +1,6 @@
 {
   "name": "whisper.cpp",
-  "version": "1.7.2",
+  "version": "1.7.3",
   "description": "Whisper speech recognition",
   "main": "whisper.js",
   "scripts": {

bindings/javascript/package.json

@@ -22,7 +22,7 @@ Usage
 ```ruby
 require "whisper"
 
-whisper = Whisper::Context.new(Whisper::Model["base"])
+whisper = Whisper::Context.new("base")
 
 params = Whisper::Params.new
 params.language = "en"
@@ -44,20 +44,26 @@ end
 Some models are prepared up-front:
 
 ```ruby
-base_en = Whisper::Model["base.en"]
+base_en = Whisper::Model.pre_converted_models["base.en"]
 whisper = Whisper::Context.new(base_en)
 ```
 
 At first time you use a model, it is downloaded automatically. After that, downloaded cached file is used. To clear cache, call `#clear_cache`:
 
 ```ruby
-Whisper::Model["base"].clear_cache
+Whisper::Model.pre_converted_models["base"].clear_cache
 ```
 
-You can see the list of prepared model names by `Whisper::Model.preconverted_model_names`:
+You also can use shorthand for pre-converted models:
 
 ```ruby
-puts Whisper::Model.preconverted_model_names
+whisper = Whisper::Context.new("base.en")
+```
+
+You can see the list of prepared model names by `Whisper::Model.preconverted_models.keys`:
+
+```ruby
+puts Whisper::Model.preconverted_models.keys
 # tiny
 # tiny.en
 # tiny-q5_1
@@ -124,13 +130,6 @@ end
 You can also add hook to params called on new segment:
 
 ```ruby
-def format_time(time_ms)
-  sec, decimal_part = time_ms.divmod(1000)
-  min, sec = sec.divmod(60)
-  hour, min = min.divmod(60)
-  "%02d:%02d:%02d.%03d" % [hour, min, sec, decimal_part]
-end
-
 # Add hook before calling #transcribe
 params.on_new_segment do |segment|
   line = "[%{st} --> %{ed}] %{text}" % {
@@ -151,7 +150,7 @@ whisper.transcribe("path/to/audio.wav", params)
 You can see model information:
 
 ```ruby
-whisper = Whisper::Context.new(Whisper::Model["base"])
+whisper = Whisper::Context.new("base")
 model = whisper.model
 
 model.n_vocab # => 51864
@@ -200,7 +199,7 @@ Using this feature, you are also able to suppress log:
 Whisper.log_set ->(level, buffer, user_data) {
   # do nothing
 }, nil
-Whisper::Context.new(MODEL)
+Whisper::Context.new("base")
 ```
 
 ### Low-level API to transcribe ###
@@ -214,14 +213,14 @@ require "wavefile"
 reader = WaveFile::Reader.new("path/to/audio.wav", WaveFile::Format.new(:mono, :float, 16000))
 samples = reader.enum_for(:each_buffer).map(&:samples).flatten
 
-whisper = Whisper::Context.new(Whisper::Model["base"])
+whisper = Whisper::Context.new("base")
 whisper.full(Whisper::Params.new, samples)
 whisper.each_segment do |segment|
   puts segment.text
 end
 ```
 
-The second argument `samples` may be an array, an object with `length` method, or a MemoryView. If you can prepare audio data as C array and export it as a MemoryView, whispercpp accepts and works with it with zero copy.
+The second argument `samples` may be an array, an object with `length` and `each` method, or a MemoryView. If you can prepare audio data as C array and export it as a MemoryView, whispercpp accepts and works with it with zero copy.
 
 License
 -------

bindings/ruby/README.md

@@ -25,7 +25,6 @@ task build: ["ext/Makefile", "ext/ruby_whisper.h", "ext/ruby_whisper.cpp", "whis
 directory "pkg"
 CLOBBER.include "pkg"
 
-TEST_MODEL = "../../models/ggml-base.en.bin"
 LIB_NAME = "whisper".ext(RbConfig::CONFIG["DLEXT"])
 SO_FILE = File.join("ext", LIB_NAME)
 LIB_FILE = File.join("lib", LIB_NAME)
@@ -41,23 +40,17 @@ file SO_FILE => "ext/Makefile" do |t|
     sh "make"
   end
 end
-CLEAN.include LIB_FILE
+CLEAN.include SO_FILE
 
 directory "lib"
 file LIB_FILE => [SO_FILE, "lib"] do |t|
   copy t.source, t.name
 end
+CLEAN.include LIB_FILE
 
 Rake::TestTask.new do |t|
   t.test_files = FileList["tests/test_*.rb"]
 end
-task test: [TEST_MODEL, LIB_FILE]
-
-file TEST_MODEL do
-  Dir.chdir "../.." do
-    sh "./models/download-ggml-model.sh base.en"
-  end
-end
 
 TEST_MEMORY_VIEW = "tests/jfk_reader/jfk_reader.#{RbConfig::CONFIG['DLEXT']}"
 file TEST_MEMORY_VIEW => "tests/jfk_reader/jfk_reader.c" do |t|
@@ -67,4 +60,5 @@ file TEST_MEMORY_VIEW => "tests/jfk_reader/jfk_reader.c" do |t|
   end
 end
 CLEAN.include "tests/jfk_reader/jfk_reader.{o,#{RbConfig::CONFIG['DLEXT']}}"
-task test: TEST_MEMORY_VIEW
+
+task test: [LIB_FILE, TEST_MEMORY_VIEW]

bindings/ruby/Rakefile

@@ -111,11 +111,6 @@
     $MK_CFLAGS     << ' -march=native -mtune=native'
     $HOST_CXXFLAGS << ' -march=native -mtune=native'
   end
-
-  if $UNAME_M.match? /aarch64.*/
-    $MK_CFLAGS   << ' -mcpu=native'
-    $MK_CXXFLAGS << ' -mcpu=native'
-  end
 else
   $MK_CFLAGS   << ' -march=rv64gcv -mabi=lp64d'
   $MK_CXXFLAGS << ' -march=rv64gcv -mabi=lp64d'
@@ -162,7 +157,6 @@
 
 $OBJ_GGML <<
   'ggml/src/ggml.o' <<
-  'ggml/src/ggml-aarch64.o' <<
   'ggml/src/ggml-alloc.o' <<
   'ggml/src/ggml-backend.o' <<
   'ggml/src/ggml-backend-reg.o' <<
@@ -172,7 +166,9 @@
   'ggml/src/ggml-cpu/ggml-cpu.o' <<
   'ggml/src/ggml-cpu/ggml-cpu-cpp.o' <<
   'ggml/src/ggml-cpu/ggml-cpu-aarch64.o' <<
-  'ggml/src/ggml-cpu/ggml-cpu-quants.o'
+  'ggml/src/ggml-cpu/ggml-cpu-hbm.o' <<
+  'ggml/src/ggml-cpu/ggml-cpu-quants.o' <<
+  'ggml/src/ggml-cpu/ggml-cpu-traits.o'
 
 $OBJ_WHISPER <<
   'src/whisper.o'