Support paraformer on iOS (#265)

* Fix C API to support streaming paraformer

* Fix Swift API

* Support paraformer in iOS
This commit is contained in:
Fangjun Kuang
2023-08-14 14:38:41 +08:00
committed by GitHub
parent 35526e26e1
commit a8bdb4b38a
12 changed files with 204 additions and 86 deletions

View File

@@ -7,6 +7,8 @@
objects = {
/* Begin PBXBuildFile section */
C93989AE2A89FE13009AB859 /* sherpa-onnx.xcframework in Frameworks */ = {isa = PBXBuildFile; fileRef = C984A81B29AA11C500D74C52 /* sherpa-onnx.xcframework */; };
C93989B02A89FE33009AB859 /* onnxruntime.xcframework in Frameworks */ = {isa = PBXBuildFile; fileRef = C93989AF2A89FE33009AB859 /* onnxruntime.xcframework */; };
C984A7E829A9EEB700D74C52 /* AppDelegate.swift in Sources */ = {isa = PBXBuildFile; fileRef = C984A7E729A9EEB700D74C52 /* AppDelegate.swift */; };
C984A7EA29A9EEB700D74C52 /* SceneDelegate.swift in Sources */ = {isa = PBXBuildFile; fileRef = C984A7E929A9EEB700D74C52 /* SceneDelegate.swift */; };
C984A7F129A9EEB900D74C52 /* Assets.xcassets in Resources */ = {isa = PBXBuildFile; fileRef = C984A7F029A9EEB900D74C52 /* Assets.xcassets */; };
@@ -18,8 +20,6 @@
C984A82829AA196100D74C52 /* Main.storyboard in Resources */ = {isa = PBXBuildFile; fileRef = C984A82629AA196100D74C52 /* Main.storyboard */; };
C984A82A29AA19AC00D74C52 /* Model.swift in Sources */ = {isa = PBXBuildFile; fileRef = C984A82929AA19AC00D74C52 /* Model.swift */; };
C984A83C29AA430B00D74C52 /* ViewController.swift in Sources */ = {isa = PBXBuildFile; fileRef = C984A83B29AA430B00D74C52 /* ViewController.swift */; };
C984A83D29AA43D900D74C52 /* sherpa-onnx.xcframework in Frameworks */ = {isa = PBXBuildFile; fileRef = C984A81B29AA11C500D74C52 /* sherpa-onnx.xcframework */; };
C984A83F29AA43EE00D74C52 /* onnxruntime.xcframework in Frameworks */ = {isa = PBXBuildFile; fileRef = C984A83E29AA43EE00D74C52 /* onnxruntime.xcframework */; };
/* End PBXBuildFile section */
/* Begin PBXContainerItemProxy section */
@@ -40,6 +40,10 @@
/* End PBXContainerItemProxy section */
/* Begin PBXFileReference section */
C93989AF2A89FE33009AB859 /* onnxruntime.xcframework */ = {isa = PBXFileReference; lastKnownFileType = wrapper.xcframework; name = onnxruntime.xcframework; path = "../../build-ios/ios-onnxruntime/1.15.1/onnxruntime.xcframework"; sourceTree = "<group>"; };
C93989B12A89FF78009AB859 /* decoder.int8.onnx */ = {isa = PBXFileReference; lastKnownFileType = file; name = decoder.int8.onnx; path = "../../../icefall-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en/decoder.int8.onnx"; sourceTree = "<group>"; };
C93989B22A89FF78009AB859 /* encoder.int8.onnx */ = {isa = PBXFileReference; lastKnownFileType = file; name = encoder.int8.onnx; path = "../../../icefall-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en/encoder.int8.onnx"; sourceTree = "<group>"; };
C93989B32A89FF78009AB859 /* tokens.txt */ = {isa = PBXFileReference; lastKnownFileType = text; name = tokens.txt; path = "../../../icefall-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en/tokens.txt"; sourceTree = "<group>"; };
C984A7E429A9EEB700D74C52 /* SherpaOnnx.app */ = {isa = PBXFileReference; explicitFileType = wrapper.application; includeInIndex = 0; path = SherpaOnnx.app; sourceTree = BUILT_PRODUCTS_DIR; };
C984A7E729A9EEB700D74C52 /* AppDelegate.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AppDelegate.swift; sourceTree = "<group>"; };
C984A7E929A9EEB700D74C52 /* SceneDelegate.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SceneDelegate.swift; sourceTree = "<group>"; };
@@ -66,8 +70,8 @@
isa = PBXFrameworksBuildPhase;
buildActionMask = 2147483647;
files = (
C984A83F29AA43EE00D74C52 /* onnxruntime.xcframework in Frameworks */,
C984A83D29AA43D900D74C52 /* sherpa-onnx.xcframework in Frameworks */,
C93989B02A89FE33009AB859 /* onnxruntime.xcframework in Frameworks */,
C93989AE2A89FE13009AB859 /* sherpa-onnx.xcframework in Frameworks */,
);
runOnlyForDeploymentPostprocessing = 0;
};
@@ -146,8 +150,12 @@
C984A81A29AA11C500D74C52 /* Frameworks */ = {
isa = PBXGroup;
children = (
C93989B12A89FF78009AB859 /* decoder.int8.onnx */,
C93989B22A89FF78009AB859 /* encoder.int8.onnx */,
C93989B32A89FF78009AB859 /* tokens.txt */,
C984A82029AA139600D74C52 /* onnxruntime.xcframework */,
C984A83E29AA43EE00D74C52 /* onnxruntime.xcframework */,
C93989AF2A89FE33009AB859 /* onnxruntime.xcframework */,
C984A81B29AA11C500D74C52 /* sherpa-onnx.xcframework */,
);
name = Frameworks;

View File

@@ -15,70 +15,91 @@ func getResource(_ forResource: String, _ ofType: String) -> String {
/// sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20 (Bilingual, Chinese + English)
/// https://k2-fsa.github.io/sherpa/onnx/pretrained_models/zipformer-transducer-models.html
func getBilingualStreamZhEnZipformer20230220() -> SherpaOnnxOnlineTransducerModelConfig {
func getBilingualStreamZhEnZipformer20230220() -> SherpaOnnxOnlineModelConfig {
let encoder = getResource("encoder-epoch-99-avg-1", "onnx")
let decoder = getResource("decoder-epoch-99-avg-1", "onnx")
let joiner = getResource("joiner-epoch-99-avg-1", "onnx")
let tokens = getResource("tokens", "txt")
return sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner,
return sherpaOnnxOnlineModelConfig(
tokens: tokens,
numThreads: 2,
transducer: sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner
),
numThreads: 1,
modelType: "zipformer"
)
}
func getZhZipformer20230615() -> SherpaOnnxOnlineTransducerModelConfig {
func getZhZipformer20230615() -> SherpaOnnxOnlineModelConfig {
let encoder = getResource("encoder-epoch-12-avg-4-chunk-16-left-128", "onnx")
let decoder = getResource("decoder-epoch-12-avg-4-chunk-16-left-128", "onnx")
let joiner = getResource("joiner-epoch-12-avg-4-chunk-16-left-128", "onnx")
let tokens = getResource("tokens", "txt")
return sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner,
return sherpaOnnxOnlineModelConfig(
tokens: tokens,
numThreads: 2,
transducer: sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner
),
numThreads: 1,
modelType: "zipformer2"
)
}
func getZhZipformer20230615Int8() -> SherpaOnnxOnlineTransducerModelConfig {
func getZhZipformer20230615Int8() -> SherpaOnnxOnlineModelConfig {
let encoder = getResource("encoder-epoch-12-avg-4-chunk-16-left-128.int8", "onnx")
let decoder = getResource("decoder-epoch-12-avg-4-chunk-16-left-128", "onnx")
let joiner = getResource("joiner-epoch-12-avg-4-chunk-16-left-128", "onnx")
let tokens = getResource("tokens", "txt")
return sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner,
return sherpaOnnxOnlineModelConfig(
tokens: tokens,
numThreads: 2,
transducer: sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner),
numThreads: 1,
modelType: "zipformer2"
)
}
func getEnZipformer20230626() -> SherpaOnnxOnlineTransducerModelConfig {
func getEnZipformer20230626() -> SherpaOnnxOnlineModelConfig {
let encoder = getResource("encoder-epoch-99-avg-1-chunk-16-left-128", "onnx")
let decoder = getResource("decoder-epoch-99-avg-1-chunk-16-left-128", "onnx")
let joiner = getResource("joiner-epoch-99-avg-1-chunk-16-left-128", "onnx")
let tokens = getResource("tokens", "txt")
return sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner,
return sherpaOnnxOnlineModelConfig(
tokens: tokens,
numThreads: 2,
transducer: sherpaOnnxOnlineTransducerModelConfig(
encoder: encoder,
decoder: decoder,
joiner: joiner),
numThreads: 1,
modelType: "zipformer2"
)
}
func getBilingualStreamingZhEnParaformer() -> SherpaOnnxOnlineModelConfig {
let encoder = getResource("encoder.int8", "onnx")
let decoder = getResource("decoder.int8", "onnx")
let tokens = getResource("tokens", "txt")
return sherpaOnnxOnlineModelConfig(
tokens: tokens,
paraformer: sherpaOnnxOnlineParaformerModelConfig(
encoder: encoder,
decoder: decoder),
numThreads: 1,
modelType: "paraformer"
)
}
/// Please refer to
/// https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
/// to add more models if you need

View File

@@ -87,7 +87,8 @@ class ViewController: UIViewController {
// let modelConfig = getBilingualStreamZhEnZipformer20230220()
// let modelConfig = getZhZipformer20230615()
let modelConfig = getEnZipformer20230626()
// let modelConfig = getEnZipformer20230626()
let modelConfig = getBilingualStreamingZhEnParaformer()
let featConfig = sherpaOnnxFeatureConfig(
sampleRate: 16000,