Add more examples for offline ASR models with Rust API. (#3209)

This PR adds additional Rust API examples for offline ASR (Automatic Speech Recognition) models, expanding coverage to include Nemo Parakeet, Zipformer variants for different languages (Vietnamese, Chinese-English, English), and updates configuration options.

.github/scripts/test-rust.sh

@@ -6,6 +6,11 @@ cd rust-api-examples
 
 ./run-version.sh
 
+./run-nemo-parakeet-en.sh
+./run-zipformer-vi.sh
+./run-zipformer-zh-en.sh
+./run-zipformer-en.sh
+
 ./run-sense-voice.sh
 
 ./run-streaming-zipformer-en.sh

rust-api-examples/examples/nemo_parakeet.rs

@@ -0,0 +1,114 @@
+// Copyright (c) 2026 Xiaomi Corporation
+//
+// This file demonstrates how to use Nemo Parakeet with sherpa-onnx's Rust API
+// for offline speech recognition.
+//
+// See ../README.md for how to run it.
+
+use clap::Parser;
+use sherpa_onnx::{OfflineRecognizer, OfflineRecognizerConfig, OfflineTransducerModelConfig, Wave};
+use std::time::Instant;
+
+/// Nemo Parakeet offline example
+#[derive(Parser, Debug)]
+#[command(author, version, about, long_about = None)]
+struct Args {
+    /// Path to WAV file
+    #[arg(long)]
+    wav: String,
+
+    /// Path to encoder ONNX model
+    #[arg(long)]
+    encoder: String,
+
+    /// Path to decoder ONNX model
+    #[arg(long)]
+    decoder: String,
+
+    /// Path to joiner ONNX model
+    #[arg(long)]
+    joiner: String,
+
+    /// Path to tokens file
+    #[arg(long)]
+    tokens: String,
+
+    /// Provider (default: cpu)
+    #[arg(long, default_value = "cpu")]
+    provider: String,
+
+    /// Enable debug logs
+    #[arg(long, default_value_t = false)]
+    debug: bool,
+
+    /// Number of threads
+    #[arg(long, default_value_t = 2)]
+    num_threads: i32,
+}
+
+fn main() {
+    let args = Args::parse();
+
+    let wave = Wave::read(&args.wav).expect("Failed to read WAV file");
+    let audio_duration = wave.samples().len() as f64 / wave.sample_rate() as f64;
+
+    // Create default recognizer config
+    let mut recognizer_config = OfflineRecognizerConfig::default();
+
+    // Set the transducer model
+    recognizer_config.model_config.transducer = OfflineTransducerModelConfig {
+        encoder: Some(args.encoder.clone()),
+        decoder: Some(args.decoder.clone()),
+        joiner: Some(args.joiner.clone()),
+    };
+
+    recognizer_config.model_config.tokens = Some(args.tokens.clone());
+    recognizer_config.model_config.provider = Some(args.provider.clone());
+    recognizer_config.model_config.debug = args.debug;
+    recognizer_config.model_config.num_threads = args.num_threads;
+
+    // Measure recognizer creation time
+    println!("Creating recognizer ...");
+    let start_creation = Instant::now();
+    let recognizer =
+        OfflineRecognizer::create(&recognizer_config).expect("Failed to create OfflineRecognizer");
+    let creation_elapsed = start_creation.elapsed().as_secs_f64();
+    println!("Recognizer created in {:.3} seconds.", creation_elapsed);
+
+    let stream = recognizer.create_stream();
+
+    // Measure recognition time
+    let start_recognition = Instant::now();
+    stream.accept_waveform(wave.sample_rate(), wave.samples());
+    recognizer.decode(&stream);
+    let recognition_elapsed = start_recognition.elapsed().as_secs_f64();
+
+    // Get recognition result
+    if let Some(result) = stream.get_result() {
+        println!("Decoded text: {}", result.text);
+
+        let total_elapsed = creation_elapsed + recognition_elapsed;
+        let rtf = recognition_elapsed / audio_duration;
+
+        println!("\n=== Performance Summary ===");
+        println!("Audio duration          : {:.3} seconds", audio_duration);
+        println!("Recognizer creation time: {:.3} seconds", creation_elapsed);
+        println!(
+            "Recognition time        : {:.3} seconds",
+            recognition_elapsed
+        );
+        println!("Total elapsed time      : {:.3} seconds", total_elapsed);
+
+        println!(
+            "Real-Time Factor (RTF)  : {:.3} (recognition_elapsed / audio_duration = {:.3} / {:.3})",
+            rtf, recognition_elapsed, audio_duration
+        );
+
+        println!(
+            "Number of threads       : {}",
+            recognizer_config.model_config.num_threads
+        );
+    } else {
+        eprintln!("Failed to get recognition result");
+    }
+}

rust-api-examples/examples/sense_voice.rs

@@ -40,6 +40,10 @@ struct Args {
     /// Enable inverse text normalization
     #[arg(long, default_value_t = true)]
     use_itn: bool,
+
+    /// Number of threads
+    #[arg(long, default_value_t = 2)]
+    num_threads: i32,
 }
 
 fn main() {
@@ -59,7 +63,7 @@ fn main() {
     recognizer_config.model_config.tokens = Some(args.tokens.clone());
     recognizer_config.model_config.provider = Some(args.provider.clone());
     recognizer_config.model_config.debug = args.debug;
-    recognizer_config.model_config.num_threads = 2;
+    recognizer_config.model_config.num_threads = args.num_threads;
 
     // Measure recognizer creation time
     println!("Creating recognizer ...");

rust-api-examples/examples/zipformer.rs

@@ -0,0 +1 @@
+nemo_parakeet.rs

rust-api-examples/run-nemo-parakeet-en.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -ex
+
+# See also
+# https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-transducer/nemo-transducer-models.html#sherpa-onnx-nemo-parakeet-tdt-0-6b-v2-int8-english
+if [ ! -f "./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/encoder.int8.onnx" ]; then
+    curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2
+    tar xvf sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2
+    rm sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2
+    ls -lh sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8
+fi
+
+# Run Rust Nemo Parakeet example
+cargo run --example nemo_parakeet -- \
+    --wav "./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/test_wavs/0.wav" \
+    --encoder "./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/encoder.int8.onnx" \
+    --decoder "./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/decoder.int8.onnx" \
+    --joiner "./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/joiner.int8.onnx" \
+    --tokens "./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/tokens.txt" \
+    --provider cpu \
+    --num-threads 2 \
+    --debug

rust-api-examples/run-sense-voice.sh

@@ -14,4 +14,6 @@ fi
 cargo run --example sense_voice -- \
     --wav ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-2024-07-17/test_wavs/en.wav \
     --model ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-2024-07-17/model.int8.onnx \
-    --tokens ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-2024-07-17/tokens.txt
+    --tokens ./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-int8-2024-07-17/tokens.txt \
+    --num-threads 2 \
+    --debug

rust-api-examples/run-zipformer-en.sh

@@ -0,0 +1,23 @@
+#!/usr/bin/env bash
+set -ex
+
+# see also
+# https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-transducer/zipformer-transducer-models.html#icefall-asr-multidataset-pruned-transducer-stateless7-2023-05-04-english
+if [ ! -f "./icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04/data/lang_bpe_500/tokens.txt" ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04.tar.bz2
+
+  tar xvf icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04.tar.bz2
+  rm icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04.tar.bz2
+  ls -lh icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04
+fi
+
+# Run Zipformer transducer
+cargo run --example zipformer -- \
+    --wav "./icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04/test_wavs/1089-134686-0001.wav" \
+    --tokens=./icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04/data/lang_bpe_500/tokens.txt \
+    --encoder=./icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04/exp/encoder-epoch-30-avg-4.int8.onnx \
+    --decoder=./icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04/exp/decoder-epoch-30-avg-4.onnx \
+    --joiner=./icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04/exp/joiner-epoch-30-avg-4.int8.onnx \
+    --provider cpu \
+    --num-threads 2 \
+    --debug

rust-api-examples/run-zipformer-vi.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -ex
+
+# see also
+# https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-transducer/zipformer-transducer-models.html#sherpa-onnx-zipformer-vi-30m-int8-2026-02-09-vietnamese
+if [ ! -f "./sherpa-onnx-zipformer-vi-30M-int8-2026-02-09/encoder.int8.onnx" ]; then
+    curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-zipformer-vi-30M-int8-2026-02-09.tar.bz2
+    tar xvf sherpa-onnx-zipformer-vi-30M-int8-2026-02-09.tar.bz2
+    rm sherpa-onnx-zipformer-vi-30M-int8-2026-02-09.tar.bz2
+    ls -lh sherpa-onnx-zipformer-vi-30M-int8-2026-02-09
+fi
+
+# Run Zipformer transducer
+cargo run --example zipformer -- \
+    --wav "./sherpa-onnx-zipformer-vi-30M-int8-2026-02-09/test_wavs/0.wav" \
+    --encoder "./sherpa-onnx-zipformer-vi-30M-int8-2026-02-09/encoder.int8.onnx" \
+    --decoder "./sherpa-onnx-zipformer-vi-30M-int8-2026-02-09/decoder.onnx" \
+    --joiner "./sherpa-onnx-zipformer-vi-30M-int8-2026-02-09/joiner.int8.onnx" \
+    --tokens "./sherpa-onnx-zipformer-vi-30M-int8-2026-02-09/tokens.txt" \
+    --provider cpu \
+    --num-threads 2 \
+    --debug

rust-api-examples/run-zipformer-zh-en.sh

@@ -0,0 +1,22 @@
+#!/usr/bin/env bash
+set -ex
+
+# see also
+# https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-transducer/zipformer-transducer-models.html#sherpa-onnx-zipformer-zh-en-2023-11-22-chinese-english
+if [ ! -f "./sherpa-onnx-zipformer-zh-en-2023-11-22/encoder-epoch-34-avg-19.int8.onnx" ]; then
+    curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-zipformer-zh-en-2023-11-22.tar.bz2
+    tar xvf sherpa-onnx-zipformer-zh-en-2023-11-22.tar.bz2
+    rm sherpa-onnx-zipformer-zh-en-2023-11-22.tar.bz2
+    ls -lh sherpa-onnx-zipformer-zh-en-2023-11-22
+fi
+
+# Run Zipformer transducer
+cargo run --example zipformer -- \
+    --wav "./sherpa-onnx-zipformer-zh-en-2023-11-22/test_wavs/0.wav" \
+    --encoder "./sherpa-onnx-zipformer-zh-en-2023-11-22/encoder-epoch-34-avg-19.int8.onnx" \
+    --decoder "./sherpa-onnx-zipformer-zh-en-2023-11-22/decoder-epoch-34-avg-19.onnx" \
+    --joiner "./sherpa-onnx-zipformer-zh-en-2023-11-22/joiner-epoch-34-avg-19.int8.onnx" \
+    --tokens "./sherpa-onnx-zipformer-zh-en-2023-11-22/tokens.txt" \
+    --provider cpu \
+    --num-threads 2 \
+    --debug