huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-15 02:16:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

Improve the quantized whisper setup. (#1018)

Browse Source 
* Improve the quantized whisper setup.

* Fix the config file paths.

* Use the standard matmul where possible.
This commit is contained in:
Laurent Mazare
2023-10-02 17:17:46 +01:00
committed by GitHub
parent e04c789230
commit 089fc3b584
8 changed files with 66 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
candle-examples/examples/whisper/main.rs
View File

@ -484,17 +484,25 @@ fn main() -> Result<()> {
println!("No audio file submitted: Downloading https://huggingface.co/datasets/Narsil/candle_demo/blob/main/samples_jfk.wav");
dataset.get("samples_jfk.wav")?
};
let config = if args.quantized {
repo.get("config-tiny.json")?
let (config, tokenizer, model) = if args.quantized {
let ext = match args.model {
WhichModel::TinyEn => "tiny-en",
WhichModel::Tiny => "tiny",
_ => unimplemented!("no quantized support for {:?}", args.model),
};
(
repo.get(&format!("config-{ext}.json"))?,
repo.get(&format!("tokenizer-{ext}.json"))?,
repo.get(&format!("model-{ext}-q40.gguf"))?,
)
} else {
repo.get("config.json")?
(
repo.get("config.json")?,
repo.get("tokenizer.json")?,
repo.get("model.safetensors")?,
)
};
let model = if args.quantized {
repo.get("model-tiny-q40.gguf")?
} else {
repo.get("model.safetensors")?
};
(config, repo.get("tokenizer.json")?, model, sample)
(config, tokenizer, model, sample)
};
let tokenizer = Tokenizer::from_file(tokenizer_filename).map_err(E::msg)?;