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

@@ -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,