Merge tag 'v1.7.1' into fix_decoding

# By Georgi Gerganov (18) and others
# Via Georgi Gerganov
* tag 'v1.7.1': (43 commits)
  release : v1.7.1
  vulkan : retry allocation with fallback flags (ggml-org#2451)
  release : v1.7.0
  scripts : bench v3-turbo
  whisper : remove mel leftover constants (396089f)
  whisper : zero-out the KV cache upon clear (ggml-org#2445)
  objc : fix build
  metal : zero-init buffer contexts (#0)
  whisper : revert mel-related changes (#0)
  whisper : adapt to latest ggml (skip) (#0)
  ggml : fix typo in example usage ggml_gallocr_new (ggml/984)
  ggml : fixes after sync (ggml/983)
  ggml-backend : add device and backend reg interfaces (llama/9707)
  Fixed dequant precision issues in Q4_1 and Q5_1 (llama/9711)
  ggml-backend : add device and backend reg interfaces (llama/9707)
  Initial cmake support of SYCL for AMD GPUs (llama/9658)
  vulkan : do not use tensor->extra (llama/9407)
  ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
  ggml: refactor cross entropy loss CPU impl. (ggml/976)
  scripts : sync ggml-backend.cpp
  ...

# Conflicts:
#	bindings/javascript/package.json

Author: bygreencn

.gitignore

@@ -3,11 +3,13 @@
 .cache/
 .coreml/
 .test/
+.venv/
 .vs/
 .vscode/
 .DS_Store
 .vimspector.json
 /CMakeSettings.json
+/talk-llama.dSYM/
 
 build/
 build-*/

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 1.6.2)
+project("whisper.cpp" VERSION 1.7.1)
 include(CheckIncludeFileCXX)
 
 set(SOVERSION 1)

Makefile

@@ -3,12 +3,11 @@ BUILD_TARGETS = \
 	main \
 	bench \
 	quantize \
-	server \
-	tests/test-c.o
+	server
 
 # Binaries only useful for tests
 TEST_TARGETS = \
-	tests/test-backend-ops
+	tests/test-c.o
 
 # Deprecation aliases
 ifdef WHISPER_CUBLAS
@@ -513,9 +512,6 @@ ifdef GGML_CUDA
 	OBJ_GGML += ggml/src/ggml-cuda.o
 	OBJ_GGML += $(patsubst %.cu,%.o,$(wildcard ggml/src/ggml-cuda/*.cu))
 	OBJ_GGML += $(OBJ_CUDA_TMPL)
-
-	#OBJ_WHISPER += src/whisper-mel-cuda.o
-
 ifdef WHISPER_FATAL_WARNINGS
 	MK_NVCCFLAGS += -Werror all-warnings
 endif # WHISPER_FATAL_WARNINGS
@@ -624,10 +620,6 @@ ggml/src/ggml-cuda.o: \
 	ggml/src/ggml-common.h \
 	$(wildcard ggml/src/ggml-cuda/*.cuh)
 	$(NVCC_COMPILE)
-
-#src/whisper-mel-cuda.o: src/whisper-mel-cuda.cu src/whisper-mel-cuda.hpp
-#	$(NVCC) $(NVCCFLAGS) $(CPPFLAGS) -Xcompiler "$(CUDA_CXXFLAGS)" -c $< -o $@
-
 endif # GGML_CUDA
 
 ifdef GGML_VULKAN
@@ -905,10 +897,10 @@ ggml/src/ggml-alloc.o: \
 	$(CC)  $(CFLAGS)   -c $< -o $@
 
 ggml/src/ggml-backend.o: \
-	ggml/src/ggml-backend.c \
+	ggml/src/ggml-backend.cpp \
 	ggml/include/ggml.h \
 	ggml/include/ggml-backend.h
-	$(CC)  $(CFLAGS)   -c $< -o $@
+	$(CXX) $(CXXFLAGS) -c $< -o $@
 
 ggml/src/ggml-quants.o: \
 	ggml/src/ggml-quants.c \
@@ -956,7 +948,6 @@ $(LIB_GGML_S): \
 
 src/whisper.o: \
 	src/whisper.cpp \
-	src/whisper-mel.hpp \
 	include/whisper.h \
 	ggml/include/ggml.h \
 	ggml/include/ggml-alloc.h \
@@ -1101,11 +1092,6 @@ tests: $(TEST_TARGETS)
 tests/test-c.o: tests/test-c.c include/whisper.h
 	$(CC) $(CFLAGS) -c $(filter-out %.h,$^) -o $@
 
-tests/test-backend-ops: tests/test-backend-ops.cpp \
-	$(OBJ_GGML)
-	$(CXX) $(CXXFLAGS) -c $< -o $(call GET_OBJ_FILE, $<)
-	$(CXX) $(CXXFLAGS) $(filter-out %.h $<,$^) $(call GET_OBJ_FILE, $<) -o $@ $(LDFLAGS)
-
 #
 # Audio samples
 #
@@ -1151,8 +1137,9 @@ samples:
 .PHONY: large-v1
 .PHONY: large-v2
 .PHONY: large-v3
+.PHONY: large-v3-turbo
 
-tiny.en tiny base.en base small.en small medium.en medium large-v1 large-v2 large-v3: main
+tiny.en tiny base.en base small.en small medium.en medium large-v1 large-v2 large-v3 large-v3-turbo: main
 	bash ./models/download-ggml-model.sh $@
 	@echo ""
 	@echo "==============================================="

Package.swift

@@ -34,7 +34,7 @@ let package = Package(
                 "src/whisper.cpp",
                 "ggml/src/ggml-aarch64.c",
                 "ggml/src/ggml-alloc.c",
-                "ggml/src/ggml-backend.c",
+                "ggml/src/ggml-backend.cpp",
                 "ggml/src/ggml-quants.c",
                 "ggml/src/ggml-metal.m"
             ],

README.md

@@ -7,7 +7,7 @@
 [![Conan Center](https://shields.io/conan/v/whisper-cpp)](https://conan.io/center/whisper-cpp)
 [![npm](https://img.shields.io/npm/v/whisper.cpp.svg)](https://www.npmjs.com/package/whisper.cpp/)
 
-Stable: [v1.6.2](https://github.com/ggerganov/whisper.cpp/releases/tag/v1.6.0) / [Roadmap | F.A.Q.](https://github.com/ggerganov/whisper.cpp/discussions/126)
+Stable: [v1.7.1](https://github.com/ggerganov/whisper.cpp/releases/tag/v1.7.1) / [Roadmap | F.A.Q.](https://github.com/ggerganov/whisper.cpp/discussions/126)
 
 High-performance inference of [OpenAI's Whisper](https://github.com/openai/whisper) automatic speech recognition (ASR) model:
 
@@ -236,6 +236,7 @@ make medium
 make large-v1
 make large-v2
 make large-v3
+make large-v3-turbo
 ```
 
 ## Memory usage

bindings/go/examples/go-model-download/main.go

@@ -24,7 +24,7 @@ const (
 
 var (
 	// The models which will be downloaded, if no model is specified as an argument
-	modelNames = []string{"ggml-tiny.en", "ggml-tiny", "ggml-base.en", "ggml-base", "ggml-small.en", "ggml-small", "ggml-medium.en", "ggml-medium", "ggml-large-v1", "ggml-large-v2", "ggml-large-v3"}
+	modelNames = []string{"ggml-tiny.en", "ggml-tiny", "ggml-base.en", "ggml-base", "ggml-small.en", "ggml-small", "ggml-medium.en", "ggml-medium", "ggml-large-v1", "ggml-large-v2", "ggml-large-v3", "large-v3-turbo"}
 )
 
 var (

bindings/javascript/package.json

@@ -1,26 +1,26 @@
-{
-  "name": "whisper.cpp",
-  "version": "1.6.2",
-  "description": "Whisper speech recognition",
-  "main": "whisper.js",
-  "scripts": {
-    "test": "echo \"todo: add tests\" && exit 0"
-  },
-  "repository": {
-    "type": "git",
-    "url": "git+https://github.com/ggerganov/whisper.cpp"
-  },
-  "keywords": [
-    "openai",
-    "whisper",
-    "speech-to-text",
-    "speech-recognition",
-    "transformer"
-  ],
-  "author": "Georgi Gerganov",
-  "license": "MIT",
-  "bugs": {
-    "url": "https://github.com/ggerganov/whisper.cpp/issues"
-  },
-  "homepage": "https://github.com/ggerganov/whisper.cpp#readme"
-}
+{
+  "name": "whisper.cpp",
+  "version": "1.7.1",
+  "description": "Whisper speech recognition",
+  "main": "whisper.js",
+  "scripts": {
+    "test": "echo \"todo: add tests\" && exit 0"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/ggerganov/whisper.cpp"
+  },
+  "keywords": [
+    "openai",
+    "whisper",
+    "speech-to-text",
+    "speech-recognition",
+    "transformer"
+  ],
+  "author": "Georgi Gerganov",
+  "license": "MIT",
+  "bugs": {
+    "url": "https://github.com/ggerganov/whisper.cpp/issues"
+  },
+  "homepage": "https://github.com/ggerganov/whisper.cpp#readme"
+}

bindings/ruby/ext/extconf.rb

@@ -1,7 +1,6 @@
 require 'mkmf'
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','whisper.cpp')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','whisper.h')} .")
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','whisper-mel.hpp')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml.h')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml.c')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-impl.h')} .")
@@ -11,7 +10,7 @@
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-alloc.c')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-backend-impl.h')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-backend.h')} .")
-system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-backend.c')} .")
+system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-backend.cpp')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-common.h')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-quants.h')} .")
 system("cp #{File.join(File.dirname(__FILE__),'..','..','..','ggml-quants.c')} .")