Merge branch 'master' into whisper-cpp-1-7-x

# Conflicts:
#	sys/build.rs
This commit is contained in:
arizhih 2024-10-22 12:23:11 +02:00
commit cc92f1bcb1
3 changed files with 44 additions and 31 deletions

4
.gitignore vendored
View file

@ -1,4 +1,6 @@
**/target **/target
**/Cargo.lock **/Cargo.lock
/.idea /.idea
/.vscode /.vscode
*.bin
*.wav

View file

@ -1,54 +1,59 @@
#![allow(clippy::uninlined_format_args)] /*
wget https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-tiny.bin
wget https://github.com/ggerganov/whisper.cpp/raw/master/samples/jfk.wav
cargo run --example basic_use ggml-tiny.bin jfk.wav
*/
use whisper_rs::{FullParams, SamplingStrategy, WhisperContext, WhisperContextParameters}; use whisper_rs::{FullParams, SamplingStrategy, WhisperContext, WhisperContextParameters};
// note that running this example will not do anything, as it is just a fn main() {
// demonstration of how to use the library, and actual usage requires let model_path = std::env::args()
// more dependencies than the base library. .nth(1)
pub fn usage() -> Result<(), &'static str> { .expect("Please specify path to model");
let wav_path = std::env::args()
.nth(2)
.expect("Please specify path to wav file");
let language = "en";
let samples: Vec<i16> = hound::WavReader::open(wav_path)
.unwrap()
.into_samples::<i16>()
.map(|x| x.unwrap())
.collect();
// load a context and model // load a context and model
let ctx = WhisperContext::new_with_params("path/to/model", WhisperContextParameters::default()) let ctx = WhisperContext::new_with_params(&model_path, WhisperContextParameters::default())
.expect("failed to load model"); .expect("failed to load model");
// make a state
let mut state = ctx.create_state().expect("failed to create state"); let mut state = ctx.create_state().expect("failed to create state");
// create a params object
// note that currently the only implemented strategy is Greedy, BeamSearch is a WIP
// n_past defaults to 0
let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 1 }); let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 1 });
// edit things as needed
// here we set the number of threads to use to 1
params.set_n_threads(1);
// we also enable translation
params.set_translate(true);
// and set the language to translate to to english // and set the language to translate to to english
params.set_language(Some("en")); params.set_language(Some(&language));
// we also explicitly disable anything that prints to stdout // we also explicitly disable anything that prints to stdout
params.set_print_special(false); params.set_print_special(false);
params.set_print_progress(false); params.set_print_progress(false);
params.set_print_realtime(false); params.set_print_realtime(false);
params.set_print_timestamps(false); params.set_print_timestamps(false);
// assume we have a buffer of audio data
// here we'll make a fake one, integer samples, 16 bit, 16KHz, stereo
let audio_data = vec![0_i16; 16000 * 2];
// we must convert to 16KHz mono f32 samples for the model // we must convert to 16KHz mono f32 samples for the model
// some utilities exist for this // some utilities exist for this
// note that you don't need to use these, you can do it yourself or any other way you want // note that you don't need to use these, you can do it yourself or any other way you want
// these are just provided for convenience // these are just provided for convenience
// SIMD variants of these functions are also available, but only on nightly Rust: see the docs // SIMD variants of these functions are also available, but only on nightly Rust: see the docs
let mut inter_audio_data = Vec::with_capacity(audio_data.len()); let mut inter_samples = vec![Default::default(); samples.len()];
whisper_rs::convert_integer_to_float_audio(&audio_data, &mut inter_audio_data)
whisper_rs::convert_integer_to_float_audio(&samples, &mut inter_samples)
.expect("failed to convert audio data"); .expect("failed to convert audio data");
let audio_data = whisper_rs::convert_stereo_to_mono_audio(&inter_audio_data) let samples = whisper_rs::convert_stereo_to_mono_audio(&inter_samples)
.expect("failed to convert audio data"); .expect("failed to convert audio data");
// now we can run the model // now we can run the model
// note the key we use here is the one we created above // note the key we use here is the one we created above
state state
.full(params, &audio_data[..]) .full(params, &samples[..])
.expect("failed to run model"); .expect("failed to run model");
// fetch the results // fetch the results
@ -67,10 +72,4 @@ pub fn usage() -> Result<(), &'static str> {
.expect("failed to get segment end timestamp"); .expect("failed to get segment end timestamp");
println!("[{} - {}]: {}", start_timestamp, end_timestamp, segment); println!("[{} - {}]: {}", start_timestamp, end_timestamp, segment);
} }
Ok(())
}
fn main() {
println!("running this example does nothing! see the source code for usage");
} }

View file

@ -32,6 +32,7 @@ fn main() {
#[cfg(feature = "coreml")] #[cfg(feature = "coreml")]
println!("cargo:rustc-link-lib=static=whisper.coreml"); println!("cargo:rustc-link-lib=static=whisper.coreml");
#[cfg(feature = "openblas")] #[cfg(feature = "openblas")]
{ {
if let Ok(openblas_path) = env::var("OPENBLAS_PATH") { if let Ok(openblas_path) = env::var("OPENBLAS_PATH") {
@ -194,6 +195,17 @@ fn main() {
}; };
let vulkan_lib_path = vulkan_path.join("Lib"); let vulkan_lib_path = vulkan_path.join("Lib");
println!("cargo:rustc-link-search={}", vulkan_lib_path.display()); println!("cargo:rustc-link-search={}", vulkan_lib_path.display());
} else if cfg!(target_os = "macos") {
println!("cargo:rerun-if-env-changed=VULKAN_SDK");
println!("cargo:rustc-link-lib=vulkan");
let vulkan_path = match env::var("VULKAN_SDK") {
Ok(path) => PathBuf::from(path),
Err(_) => panic!(
"Please install Vulkan SDK and ensure that VULKAN_SDK env variable is set"
),
};
let vulkan_lib_path = vulkan_path.join("lib");
println!("cargo:rustc-link-search={}", vulkan_lib_path.display());
} else { } else {
println!("cargo:rustc-link-lib=vulkan"); println!("cargo:rustc-link-lib=vulkan");
} }