whisper.cpp 0.3.2 → 0.3.4

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 3d1b91cc9ac6e7682429f261d3e9b283c4c34ef246a2e88d07f1c294d145d1f4
-  data.tar.gz: be2a0ff1c4eda542685c209e6c64b7f3f3f902c4353534747922ea89c42ef24d
+  metadata.gz: a9cd333686c64e91cb061ac8cf093e32bd76e977854ca84ee25c3929929c2f57
+  data.tar.gz: aff6dd2d092a9ad86f6f3c1453aaed13c42fc6f8d686886c86fcccc30667a15a
 SHA512:
-  metadata.gz: c992abd9e3e9b48d6b31cfa6489b5b09489781d251c2d3ca24d4932fca579b7d19b7483a1d72aefe77ac0655d5d5af19708bf6342cf6e5d4922d3663b1244acc
-  data.tar.gz: 8890fa1e7653f8690c3686a7751e597958c06adfaa3a0b3a3e74c947920f1da2014ada7eebe4fd78e85cc2a74fe235b244c43817937072ffe3b7c75792567ec1
+  metadata.gz: 01cf0c815ec1e7630dc42787f424b521e2b172d67aa78662791c99d126bd4eed941d550ffe0601f88fec5bb6046332bbb0fcdb63eeaea1b3e538f6253768366d
+  data.tar.gz: 5e894cad714070459949b15cc7c527124ada408639dd5220cd14ba6bc92eafd2bde138f27683addf5d07bc2d193bc21a6d3813971a70881113e5e4696cf95023

data/.ruby-version

@@ -1 +1 @@
-3.3.0
+3.3.6

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    whisper.cpp (0.1.0)
+    whisper.cpp (0.3.4)
       ffi (~> 1.15)
 
 GEM
@@ -28,7 +28,7 @@ DEPENDENCIES
   whisper.cpp!
 
 RUBY VERSION
-   ruby 3.3.0p0
+   ruby 3.3.6p108
 
 BUNDLED WITH
    2.5.11

data/ext/extconf.rb

@@ -43,26 +43,45 @@ unless Dir.exist?(whisper_dir)
   abort "Failed to find or create the whisper.cpp directory at #{whisper_dir}"
 end
 
-# Now, proceed to build libwhispercpp.so using the whisper.cpp Makefile
+# Now, proceed to modify the Makefile and build libwhispercpp.so
 Dir.chdir(whisper_dir) do
   # Set environment variables for build settings
-  ENV['GGML_CUDA'] = '1'  # Enable CUDA support
+  ENV['GGML_CUDA'] = '1'  # Enable CUDA support if desired
 
-  puts "Building libwhispercpp.so with GGML_CUDA=#{ENV['GGML_CUDA']}..."
+  # Modify the Makefile to add libwhispercpp.so target
+  makefile_path = File.join(Dir.pwd, 'Makefile')
+
+  makefile_content = File.read(makefile_path)
+
+  # Check if 'libwhispercpp.so' target is already defined
+  unless makefile_content.include?('libwhispercpp.so:')
+    # Append the new target at the end of the Makefile
+    new_content = <<~MAKEFILE
+
+      # Custom target for building libwhispercpp.so
+      libwhispercpp.so: $(OBJ_GGML) $(OBJ_WHISPER) $(OBJ_COMMON) $(OBJ_SDL) $(OBJ_WHISPER_EXTRA)
+      \t$(CXX) $(CXXFLAGS) -shared -fPIC -o $@ $^ $(LDFLAGS)
+
+      BUILD_TARGETS += libwhispercpp.so
+    MAKEFILE
+
+    makefile_content << new_content
+
+    # Write back the modified Makefile
+    File.open(makefile_path, 'w') do |f|
+      f.write(makefile_content)
+    end
 
-  # Build libwhisper.a and libggml.a
-  unless system 'make clean && make -j libwhisper.a libggml.a'
-    abort "Failed to build libwhisper.a and libggml.a"
+    puts "Modified Makefile to add libwhispercpp.so target."
+  else
+    puts "Makefile already contains libwhispercpp.so target."
   end
 
-  # Link the static libraries into a single shared library using g++
-  gcc_command = 'g++ -shared -o libwhispercpp.so ' \
-                '-Wl,--whole-archive libwhisper.a -Wl,--no-whole-archive libggml.a ' \
-                '-L$( [ -d /opt/cuda ] && echo /opt/cuda/lib || echo /usr/local/cuda/lib ) ' \
-                '-lcuda -lcudart -lcublas -lc -lm -lstdc++'
+  puts "Building libwhispercpp.so with GGML_CUDA=#{ENV['GGML_CUDA']}..."
 
-  unless system gcc_command
-    abort "Failed to link libwhispercpp.so"
+  # Build libwhispercpp.so
+  unless system 'make clean && make -j libwhispercpp.so'
+    abort "Failed to build libwhispercpp.so"
   end
 
   # Verify that libwhispercpp.so was created
@@ -71,10 +90,11 @@ Dir.chdir(whisper_dir) do
     abort "libwhispercpp.so not found after compilation"
   end
 
-  # Copy the compiled library to the gem's lib directory
-  FileUtils.cp(source_lib, root_dir)
+  # Copy the compiled library to the gem's root directory
+  destination_lib = File.join(root_dir, 'libwhispercpp.so')
+  FileUtils.cp(source_lib, destination_lib)
 
-  puts "Copied libwhispercpp.so to #{root_dir}"
+  puts "Copied libwhispercpp.so to #{destination_lib}"
 end
 
 puts "Compilation completed."
@@ -82,9 +102,9 @@ puts "Compilation completed."
 # Create a no-op Makefile to prevent RubyGems from attempting further compilation
 makefile_content = <<~MAKEFILE
   all:
-  	@echo 'libwhispercpp.so already built.'
+  \t@echo 'libwhispercpp.so already built.'
   install:
-  	@echo 'libwhispercpp.so already installed.'
+  \t@echo 'libwhispercpp.so already installed.'
 MAKEFILE
 
 File.open('Makefile', 'w') do |f|
@@ -93,15 +113,10 @@ end
 
 puts "Created a no-op Makefile to prevent further compilation."
 
-# After copying libwhispercpp.so
-
-# Path to the cloned whisper.cpp directory
+# Remove the cloned whisper.cpp directory
 cloned_dir = whisper_dir
 
-# Remove the cloned whisper.cpp directory
 puts "Removing cloned whisper.cpp directory at #{cloned_dir}..."
 FileUtils.rm_rf(cloned_dir)
-
 puts "Removed cloned whisper.cpp directory."
 
-

data/lib/whisper/model.rb

@@ -8,42 +8,44 @@ module Whisper
     def initialize(model_path)
       @model_path = model_path
       @ctx = nil
+      @state = nil
       init_whisper_context
+      init_whisper_state
     end
 
     def transcribe_from_file(audio_file_path, format: 'plaintext', **params)
       # Load audio file and convert to float array
-      audio_data = Whisper::AudioProcessor.convert_to_float_array(audio_file_path)
-      transcribe_from_audio_data(audio_data, format: format, **params)
+      audio_data = Whisper::AudioProcessor.convert_to_float_array audio_file_path
+      transcribe_from_audio_data audio_data, format: format, **params
     end
 
     def transcribe_from_audio_data(audio_data, format: 'plaintext', **params)
       # Prepare full params
-      full_params = default_full_params(params)
+      full_params = default_full_params params
 
       # Prepare audio data pointer
       n_samples = audio_data.size
-      samples_ptr = FFI::MemoryPointer.new(:float, n_samples)
-      samples_ptr.write_array_of_float(audio_data)
+      samples_ptr = FFI::MemoryPointer.new :float, n_samples
+      samples_ptr.write_array_of_float audio_data
 
-      # Call the whisper_full_parallel function
-      n_processors = params.fetch :n_processors, ENV['WHISPER_N_PROCS']&.to_i || 1
-      result = Whisper.whisper_full_parallel(@ctx, full_params, samples_ptr, n_samples, n_processors)
+      # Call the whisper_full_with_state function
+      result = Whisper.whisper_full_with_state @ctx, @state, full_params, samples_ptr, n_samples
       raise 'Transcription failed' if result != 0
 
       # Retrieve detected language
-      lang_id = Whisper.whisper_full_lang_id(@ctx)
-      language = Whisper.whisper_lang_str(lang_id)
+      lang_id = Whisper.whisper_full_lang_id_from_state @state
+      language = Whisper.whisper_lang_str lang_id
 
       # Retrieve the transcription output
-      n_segments = Whisper.whisper_full_n_segments(@ctx)
-      output = format_transcription(format, n_segments: n_segments)
+      n_segments = Whisper.whisper_full_n_segments_from_state @state
+      output = format_transcription format, n_segments: n_segments
 
-      TranscriptionResult.new(language, output)
+      TranscriptionResult.new language, output
     end
 
     def close
-      Whisper.whisper_free(@ctx) unless @ctx.nil?
+      Whisper.whisper_free_state @state unless @state.nil?
+      Whisper.whisper_free @ctx unless @ctx.nil?
     end
 
     private
@@ -59,18 +61,23 @@ module Whisper
       ctx_params[:gpu_device] = ENV['WHISPER_GPU']&.to_i || 0
 
       # Initialize context
-      @ctx = Whisper.whisper_init_from_file_with_params(@model_path, ctx_params)
+      @ctx = Whisper.whisper_init_from_file_with_params @model_path, ctx_params
       raise 'Failed to initialize Whisper model' if @ctx.null?
     end
 
+    def init_whisper_state
+      @state = Whisper.whisper_init_state @ctx
+      raise 'Failed to initialize Whisper state' if @state.null?
+    end
+
     def default_full_params params = {}
       # Get default full params
-      strategy = params.fetch(:sampling_strategy, Whisper::WHISPER_SAMPLING_GREEDY)
-      full_params = Whisper.whisper_full_default_params(strategy)
+      strategy = params.fetch :sampling_strategy, Whisper::WHISPER_SAMPLING_GREEDY
+      full_params = Whisper.whisper_full_default_params strategy
 
       # Set translate to false to prevent translation to English
       full_params[:translate] = false
-      full_params[:language]  = FFI::MemoryPointer.from_string 'auto'
+      full_params[:language] = FFI::MemoryPointer.from_string 'auto'
 
       # Set user-provided full params
       params.select{ |k, _| full_params.members.include? k }.each do |key, value|
@@ -85,17 +92,17 @@ module Whisper
       case format.downcase
       when 'plaintext'
         n_segments.times do |i|
-          segment_text = Whisper.whisper_full_get_segment_text(@ctx, i)
+          segment_text = Whisper.whisper_full_get_segment_text_from_state @state, i
           output += segment_text
         end
       when 'srt'
         n_segments.times do |i|
-          start_time = Whisper.whisper_full_get_segment_t0(@ctx, i) / 100.0
-          end_time = Whisper.whisper_full_get_segment_t1(@ctx, i) / 100.0
-          segment_text = Whisper.whisper_full_get_segment_text(@ctx, i)
+          start_time = Whisper.whisper_full_get_segment_t0_from_state(@state, i) / 100.0
+          end_time = Whisper.whisper_full_get_segment_t1_from_state(@state, i) / 100.0
+          segment_text = Whisper.whisper_full_get_segment_text_from_state @state, i
 
           output += "#{i + 1}\n"
-          output += "#{format_time_srt(start_time)} --> #{format_time_srt(end_time)}\n"
+          output += "#{format_time_srt start_time} --> #{format_time_srt end_time}\n"
           output += "#{segment_text.strip}\n\n"
         end
       else
@@ -109,7 +116,7 @@ module Whisper
       minutes = ((seconds % 3600) / 60).to_i
       secs = (seconds % 60).to_i
       millis = ((seconds - seconds.to_i) * 1000).to_i
-      format('%02d:%02d:%02d,%03d', hours, minutes, secs, millis)
+      format '%02d:%02d:%02d,%03d', hours, minutes, secs, millis
     end
   end
 end

data/lib/whisper/version.rb

@@ -1,5 +1,5 @@
 module Whisper
 
-  VERSION = '0.3.2'
+  VERSION = '0.3.4'
 
 end

data/lib/whisper.rb

@@ -26,7 +26,19 @@ module Whisper
     :WHISPER_AHEADS_MEDIUM,
     :WHISPER_AHEADS_LARGE_V1,
     :WHISPER_AHEADS_LARGE_V2,
-    :WHISPER_AHEADS_LARGE_V3
+    :WHISPER_AHEADS_LARGE_V3,
+    :WHISPER_AHEADS_LARGE_V3_TURBO  # Added new enum value
+  ]
+
+  # Enums for grammar element type
+  enum :whisper_gretype, [
+    :WHISPER_GRETYPE_END,            0,
+    :WHISPER_GRETYPE_ALT,
+    :WHISPER_GRETYPE_RULE_REF,
+    :WHISPER_GRETYPE_CHAR,
+    :WHISPER_GRETYPE_CHAR_NOT,
+    :WHISPER_GRETYPE_CHAR_RNG_UPPER,
+    :WHISPER_GRETYPE_CHAR_ALT
   ]
 
   # Enums for sampling strategy
@@ -93,7 +105,7 @@ module Whisper
 
   # whisper_model_loader struct
   #class WhisperModelLoader < FFI::Struct
-  #  callback :read_callback, [:pointer, :pointer, :size_t], :size_t
+  #  callback :read_callback, [:pointer, :pointer], :size_t
   #  callback :eof_callback, [:pointer], :bool
   #  callback :close_callback, [:pointer], :void
 
@@ -108,7 +120,7 @@ module Whisper
   # whisper_grammar_element struct
   class WhisperGrammarElement < FFI::Struct
     layout(
-      :type, :int,
+      :type, :whisper_gretype,
       :value, :uint32
     )
   end
@@ -131,7 +143,7 @@ module Whisper
   # whisper_full_params struct
   class WhisperFullParams < FFI::Struct
     layout(
-      :strategy, :int,
+      :strategy, :whisper_sampling_strategy,
       :n_threads, :int,
       :n_max_text_ctx, :int,
       :offset_ms, :int,
@@ -189,8 +201,10 @@ module Whisper
 
   # Get default context params
   attach_function :whisper_context_default_params, [], WhisperContextParams.by_value
+  attach_function :whisper_context_default_params_by_ref, [], :pointer
   # Get default full params
-  attach_function :whisper_full_default_params, [:int], WhisperFullParams.by_value
+  attach_function :whisper_full_default_params, [:whisper_sampling_strategy], WhisperFullParams.by_value
+  attach_function :whisper_full_default_params_by_ref, [:whisper_sampling_strategy], :pointer
 
   # Function Bindings
 
@@ -208,8 +222,8 @@ module Whisper
   attach_function :whisper_init_state, [:pointer], :pointer
 
   # OpenVINO functions
-  #attach_function :whisper_ctx_init_openvino_encoder_with_state, [:pointer, :pointer, :string, :string, :string], :int
-  #attach_function :whisper_ctx_init_openvino_encoder, [:pointer, :string, :string, :string], :int
+  attach_function :whisper_ctx_init_openvino_encoder_with_state, [:pointer, :pointer, :string, :string, :string], :int
+  attach_function :whisper_ctx_init_openvino_encoder, [:pointer, :string, :string, :string], :int
 
   # Free functions
   attach_function :whisper_free, [:pointer], :void

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: whisper.cpp
 version: !ruby/object:Gem::Version
-  version: 0.3.2
+  version: 0.3.4
 platform: ruby
 authors:
 - Braulio Oliveira
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2024-10-12 00:00:00.000000000 Z
+date: 2024-11-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi
@@ -95,7 +95,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.5.3
+rubygems_version: 3.5.22
 signing_key:
 specification_version: 4
 summary: Ruby bindings for whisper.cpp