//go:build darwin
|
|
package engine
|
|
import (
|
"os"
|
"path/filepath"
|
"testing"
|
|
"voicesnap/internal/model"
|
)
|
|
func TestOfflineConfigForParakeetUsesNemoTransducer(t *testing.T) {
|
root := t.TempDir()
|
files := map[string]string{
|
"encoder": filepath.Join(root, "encoder.int8.onnx"),
|
"decoder": filepath.Join(root, "decoder.int8.onnx"),
|
"joiner": filepath.Join(root, "joiner.int8.onnx"),
|
"tokens": filepath.Join(root, "tokens.txt"),
|
}
|
resolved := model.ResolvedModel{
|
ModelID: model.ParakeetModelID,
|
BackendKind: model.BackendNemoTransducer,
|
Profile: model.ModelProfile{
|
DisplayName: "Parakeet English",
|
BackendKind: model.BackendNemoTransducer,
|
ProviderOrder: []string{"cpu"},
|
NumThreads: 4,
|
},
|
Files: files,
|
}
|
|
config, err := offlineConfigForResolvedModel(resolved, "cpu")
|
if err != nil {
|
t.Fatal(err)
|
}
|
if config.ModelConfig.Transducer.Encoder != files["encoder"] {
|
t.Fatalf("encoder = %q, want %q", config.ModelConfig.Transducer.Encoder, files["encoder"])
|
}
|
if config.ModelConfig.Transducer.Decoder != files["decoder"] {
|
t.Fatalf("decoder = %q, want %q", config.ModelConfig.Transducer.Decoder, files["decoder"])
|
}
|
if config.ModelConfig.Transducer.Joiner != files["joiner"] {
|
t.Fatalf("joiner = %q, want %q", config.ModelConfig.Transducer.Joiner, files["joiner"])
|
}
|
if config.ModelConfig.Tokens != files["tokens"] {
|
t.Fatalf("tokens = %q, want %q", config.ModelConfig.Tokens, files["tokens"])
|
}
|
if config.ModelConfig.ModelType != model.BackendNemoTransducer {
|
t.Fatalf("model type = %q, want %q", config.ModelConfig.ModelType, model.BackendNemoTransducer)
|
}
|
}
|
|
func TestOfflineConfigForQwen3ASRUsesQwen3Config(t *testing.T) {
|
root := t.TempDir()
|
files := map[string]string{
|
"conv_frontend": filepath.Join(root, "conv_frontend.onnx"),
|
"encoder": filepath.Join(root, "encoder.int8.onnx"),
|
"decoder": filepath.Join(root, "decoder.int8.onnx"),
|
"tokenizer": filepath.Join(root, "tokenizer"),
|
}
|
resolved := model.ResolvedModel{
|
ModelID: model.Qwen3ASRModelID,
|
BackendKind: model.BackendQwen3ASR,
|
Profile: model.ModelProfile{
|
DisplayName: "Qwen3-ASR",
|
BackendKind: model.BackendQwen3ASR,
|
ProviderOrder: []string{"cpu"},
|
NumThreads: 4,
|
},
|
Files: files,
|
}
|
|
config, err := offlineConfigForResolvedModel(resolved, "cpu")
|
if err != nil {
|
t.Fatal(err)
|
}
|
if config.ModelConfig.Qwen3ASR.ConvFrontend != files["conv_frontend"] {
|
t.Fatalf("conv frontend = %q, want %q", config.ModelConfig.Qwen3ASR.ConvFrontend, files["conv_frontend"])
|
}
|
if config.ModelConfig.Qwen3ASR.Encoder != files["encoder"] {
|
t.Fatalf("encoder = %q, want %q", config.ModelConfig.Qwen3ASR.Encoder, files["encoder"])
|
}
|
if config.ModelConfig.Qwen3ASR.Decoder != files["decoder"] {
|
t.Fatalf("decoder = %q, want %q", config.ModelConfig.Qwen3ASR.Decoder, files["decoder"])
|
}
|
if config.ModelConfig.Qwen3ASR.Tokenizer != files["tokenizer"] {
|
t.Fatalf("tokenizer = %q, want %q", config.ModelConfig.Qwen3ASR.Tokenizer, files["tokenizer"])
|
}
|
if config.ModelConfig.Qwen3ASR.MaxTotalLen != 1024 {
|
t.Fatalf("max total len = %d, want 1024", config.ModelConfig.Qwen3ASR.MaxTotalLen)
|
}
|
if config.ModelConfig.Qwen3ASR.MaxNewTokens != 256 {
|
t.Fatalf("max new tokens = %d, want 256", config.ModelConfig.Qwen3ASR.MaxNewTokens)
|
}
|
if config.ModelConfig.Tokens != "" {
|
t.Fatalf("tokens = %q, want empty for qwen3-asr", config.ModelConfig.Tokens)
|
}
|
}
|
|
func TestOnlineConfigForXASRUsesStreamingTransducer(t *testing.T) {
|
root := t.TempDir()
|
files := map[string]string{
|
"encoder": filepath.Join(root, "encoder-960ms.onnx"),
|
"decoder": filepath.Join(root, "decoder-960ms.onnx"),
|
"joiner": filepath.Join(root, "joiner-960ms.onnx"),
|
"tokens": filepath.Join(root, "tokens.txt"),
|
}
|
resolved := model.ResolvedModel{
|
ModelID: model.XASRZhEn960ModelID,
|
BackendKind: model.BackendXASRStreaming,
|
Profile: model.ModelProfile{
|
DisplayName: "X-ASR zh-en 960ms",
|
BackendKind: model.BackendXASRStreaming,
|
ProviderOrder: []string{"cpu"},
|
NumThreads: 1,
|
},
|
Files: files,
|
}
|
|
config, err := onlineConfigForResolvedModel(resolved, "cpu")
|
if err != nil {
|
t.Fatal(err)
|
}
|
if config.ModelConfig.Transducer.Encoder != files["encoder"] {
|
t.Fatalf("encoder = %q, want %q", config.ModelConfig.Transducer.Encoder, files["encoder"])
|
}
|
if config.ModelConfig.Transducer.Decoder != files["decoder"] {
|
t.Fatalf("decoder = %q, want %q", config.ModelConfig.Transducer.Decoder, files["decoder"])
|
}
|
if config.ModelConfig.Transducer.Joiner != files["joiner"] {
|
t.Fatalf("joiner = %q, want %q", config.ModelConfig.Transducer.Joiner, files["joiner"])
|
}
|
if config.ModelConfig.Tokens != files["tokens"] {
|
t.Fatalf("tokens = %q, want %q", config.ModelConfig.Tokens, files["tokens"])
|
}
|
if config.ModelConfig.ModelType != "zipformer2" {
|
t.Fatalf("model type = %q, want zipformer2", config.ModelConfig.ModelType)
|
}
|
if config.DecodingMethod != "greedy_search" {
|
t.Fatalf("decoding method = %q, want greedy_search", config.DecodingMethod)
|
}
|
}
|
|
func TestRememberNonEmptyTextKeepsLastPartial(t *testing.T) {
|
got := rememberNonEmptyText("最后一句", " ")
|
if got != "最后一句" {
|
t.Fatalf("rememberNonEmptyText() = %q, want last non-empty text", got)
|
}
|
|
got = rememberNonEmptyText("旧内容", "新内容")
|
if got != "新内容" {
|
t.Fatalf("rememberNonEmptyText() = %q, want new non-empty text", got)
|
}
|
}
|
|
func TestXASRTailPaddingMatchesFlushWindow(t *testing.T) {
|
if xasrTailPaddingSamples != 24000 {
|
t.Fatalf("tail padding samples = %d, want 24000", xasrTailPaddingSamples)
|
}
|
if len(xasrTailPadding) != xasrTailPaddingSamples {
|
t.Fatalf("tail padding len = %d, want %d", len(xasrTailPadding), xasrTailPaddingSamples)
|
}
|
}
|
|
func TestXASRRealModelSmoke(t *testing.T) {
|
if os.Getenv("PRIVATEVOICE_XASR_SMOKE") != "1" {
|
t.Skip("set PRIVATEVOICE_XASR_SMOKE=1 to run the real X-ASR model smoke test")
|
}
|
|
resolved, err := model.ResolveModel(model.XASRZhEn960ModelID)
|
if err != nil {
|
t.Fatal(err)
|
}
|
if !resolved.IsUsable() {
|
t.Fatalf("x-asr model is not usable: status=%s missing=%v problems=%v", resolved.Status, resolved.Missing, resolved.Problems)
|
}
|
|
eng, err := NewWithResolvedModel(resolved)
|
if err != nil {
|
t.Fatal(err)
|
}
|
defer eng.Close()
|
|
text, err := eng.Recognize(make([]float32, 16000))
|
if err != nil {
|
t.Fatal(err)
|
}
|
t.Logf("hardware=%s silence_text=%q", eng.HardwareInfo(), text)
|
}
|