software-mansion/react-native-executorch-whisper-small

whisper-small

This repository hosts the whisper-small models exported for the React Native ExecuTorch library as ExecuTorch .pte programs, ready to run on device.

Upstream model: whisper-small

Variants

Path	Backend	Precision
coreml/whisper_small_coreml_fp16.pte	coreml	fp16
mlx/whisper_small_mlx_int8.pte	mlx	int8
vulkan/whisper_small_vulkan_fp16.pte	vulkan	fp16
vulkan/whisper_small_vulkan_int8.pte	vulkan	int8
xnnpack/whisper_small_xnnpack_fp32.pte	xnnpack	fp32
xnnpack/whisper_small_xnnpack_int8.pte	xnnpack	int8

XNNPACK int8

xnnpack/whisper_small_xnnpack_int8.pte is 8da4w (int8 dynamic activations, int4 weights, group 64). Greedy WER (CER for zh) against the fp32 file, LibriSpeech test-clean (150 clips) and FLEURS (80 per language):

	en	de	es	pl	fr	zh
fp32	2.84	12.01	8.05	22.57	20.71	10.81
int8	3.30	12.71	8.15	25.26	21.92	12.89

Repository structure

config.json                             35 B
coreml/config.json                      1.5 kB
coreml/whisper_small_coreml_fp16.pte    462 MB
mlx/config.json                         1.5 kB
mlx/whisper_small_mlx_int8.pte          276 MB
tokenizer.json                          2.4 MB
vulkan/config.json                      2.6 kB
vulkan/whisper_small_vulkan_fp16.pte    881 MB
vulkan/whisper_small_vulkan_int8.pte    484 MB
xnnpack/config.json                     2.6 kB
xnnpack/whisper_small_xnnpack_fp32.pte  1.1 GB
xnnpack/whisper_small_xnnpack_int8.pte  421 MB

Compatibility

These files are published for the ExecuTorch v1.4.1 runtime. ExecuTorch gives no forward compatibility guarantee, so an older runtime may fail to load them.

To use them in React Native ExecuTorch, pass the model constant shipped in the library's model registry to the corresponding task pipeline. See the documentation.

To load these files in your own ExecuTorch runtime, read the compatibility note first.