Whisper small speech-to-text (encoder + decoder, transformers.js)
Browser-ready import artifacts for automatic-speech-recognition, produced by SkillSafe's
reproducible converter (models/ in skillsafe.ai)
from a pinned upstream source. Every byte here is derivable from that source plus
the recipe below; nothing was edited by hand.
Provenance
| Upstream | https://huggingface.co/onnx-community/whisper-small/tree/36050c46d777d46dc4b5f43f6d90574fc38f8732 |
| Upstream SHA-256 / commit | 36050c46d777d46dc4b5f43f6d90574fc38f8732 |
| Recipe | recipes/whisper-small.yaml — sha256 d41958ee8e7f7506d7116e49212de3d9e5cb2428402258314f3276c6e81e7dfd |
| Toolchain | Python 3.12.13, torch 2.10.0, onnx 1.23.0, onnxruntime 1.30.0 on Darwin 25.6.0 arm64 |
| Converted | 2026-09-22T20:49:19+00:00 |
Files
| file | class | size | SHA-256 |
|---|---|---|---|
added_tokens.json |
bundle | 0.03 MB | 9715fd2243b6f06a5858b5e32950d2853f73dd5bc201aafcf76f5082a2d8acd1 |
config.json |
bundle | 0.00 MB | 457854d452f17661e197d74aee12b8e74fb75ba30ebfaa7426d0d61ea1e08a18 |
generation_config.json |
bundle | 0.00 MB | f538b28220c6a6d6f1af1458d4141cacb4ef4963df3de98a19490440c412ddf0 |
merges.txt |
bundle | 0.47 MB | 2df2990a395e35e8dfbc7511e08c12d56018d8d04691e0133e5d63b21e154dc6 |
normalizer.json |
bundle | 0.05 MB | bf1c507dc8724ca9cf9903640dacfb69dae2f00edee4f21ceba106a7392f26dd |
onnx/decoder_model_merged.onnx |
registry (fp32) | 586.82 MB | 6ed5e35feaba79ad2e89b368ddc7b4ddaa3c00b4c37a664375d3428a76fecc6a |
onnx/decoder_model_merged_fp16.onnx |
registry (fp16) | 294.29 MB | 22aba6c7f5193701cbe1519051b6ef097eb530ad6887b7093065ec59b830f61d |
onnx/decoder_model_merged_quantized.onnx |
registry (q8) | 149.49 MB | ec07c3cbb64172c39791e26ee870a65ac22b458c36722bfe2776b3dbf741e0c9 |
onnx/encoder_model.onnx |
registry (fp32, q8) | 336.48 MB | b37cd6625dc36f9178ec7539a1876b9680ea26a910097e092be39dc766320c7b |
onnx/encoder_model_fp16.onnx |
registry (fp16) | 168.43 MB | 5549cd8666ff4b694ceb128bfa48b95bdcceec29075cf2c2212f90002cc058de |
preprocessor_config.json |
bundle | 0.00 MB | a6a76d28c93edb273669eb9e0b0636a2bddbb1272c3261e47b7ca6dfdbac1b8d |
special_tokens_map.json |
bundle | 0.00 MB | e67ae3a0aaa99abcd9f187138e12db1f65c16a14761c50ef10eef2c174a7a691 |
tokenizer.json |
bundle | 2.37 MB | 27fc476bfe7f17299480be2273fc0608e4d5a99aba2ab5dec5374b4482d1a566 |
tokenizer_config.json |
bundle | 0.27 MB | 2a4c4281cf9f51ac6ccc406fdc711a087afe6530f671fa7b80953edc498275ce |
vocab.json |
bundle | 0.99 MB | 50d6a919f0a0601d56a04eb583c780d18553aa388254ba3158eb6a00f13e2c1a |
registry files are parameter files served from models.skillsafe.ai once vetted;
bundle files ship inside an app; registry-shared is a runtime library reused by
every model of the same architecture.
Verification
Imported as published upstream (no conversion). Each file is pinned by SHA-256 to its source; every ONNX file passed onnx.checker and a CPU smoke run under onnxruntime with zero-filled inputs at the declared shapes:
| file | inputs | outputs | ms |
|---|---|---|---|
onnx/decoder_model_merged.onnx |
input_ids[1, 4], encoder_hidden_states[1, 1500, 768], past_key_values.0.decoder.key[1, 12, 1, 64], past_key_values.0.decoder.value[1, 12, 1, 64], past_key_values.0.encoder.key[1, 12, 1500, 64], past_key_values.0.encoder.value[1, 12, 1500, 64], past_key_values.1.decoder.key[1, 12, 1, 64], past_key_values.1.decoder.value[1, 12, 1, 64], past_key_values.1.encoder.key[1, 12, 1500, 64], past_key_values.1.encoder.value[1, 12, 1500, 64], past_key_values.2.decoder.key[1, 12, 1, 64], past_key_values.2.decoder.value[1, 12, 1, 64], past_key_values.2.encoder.key[1, 12, 1500, 64], past_key_values.2.encoder.value[1, 12, 1500, 64], past_key_values.3.decoder.key[1, 12, 1, 64], past_key_values.3.decoder.value[1, 12, 1, 64], past_key_values.3.encoder.key[1, 12, 1500, 64], past_key_values.3.encoder.value[1, 12, 1500, 64], past_key_values.4.decoder.key[1, 12, 1, 64], past_key_values.4.decoder.value[1, 12, 1, 64], past_key_values.4.encoder.key[1, 12, 1500, 64], past_key_values.4.encoder.value[1, 12, 1500, 64], past_key_values.5.decoder.key[1, 12, 1, 64], past_key_values.5.decoder.value[1, 12, 1, 64], past_key_values.5.encoder.key[1, 12, 1500, 64], past_key_values.5.encoder.value[1, 12, 1500, 64], past_key_values.6.decoder.key[1, 12, 1, 64], past_key_values.6.decoder.value[1, 12, 1, 64], past_key_values.6.encoder.key[1, 12, 1500, 64], past_key_values.6.encoder.value[1, 12, 1500, 64], past_key_values.7.decoder.key[1, 12, 1, 64], past_key_values.7.decoder.value[1, 12, 1, 64], past_key_values.7.encoder.key[1, 12, 1500, 64], past_key_values.7.encoder.value[1, 12, 1500, 64], past_key_values.8.decoder.key[1, 12, 1, 64], past_key_values.8.decoder.value[1, 12, 1, 64], past_key_values.8.encoder.key[1, 12, 1500, 64], past_key_values.8.encoder.value[1, 12, 1500, 64], past_key_values.9.decoder.key[1, 12, 1, 64], past_key_values.9.decoder.value[1, 12, 1, 64], past_key_values.9.encoder.key[1, 12, 1500, 64], past_key_values.9.encoder.value[1, 12, 1500, 64], past_key_values.10.decoder.key[1, 12, 1, 64], past_key_values.10.decoder.value[1, 12, 1, 64], past_key_values.10.encoder.key[1, 12, 1500, 64], past_key_values.10.encoder.value[1, 12, 1500, 64], past_key_values.11.decoder.key[1, 12, 1, 64], past_key_values.11.decoder.value[1, 12, 1, 64], past_key_values.11.encoder.key[1, 12, 1500, 64], past_key_values.11.encoder.value[1, 12, 1500, 64], use_cache_branch[1] | logits[1, 4, 51865], present.0.decoder.key[1, 12, 4, 64], present.0.decoder.value[1, 12, 4, 64], present.0.encoder.key[1, 12, 1500, 64], present.0.encoder.value[1, 12, 1500, 64], present.1.decoder.key[1, 12, 4, 64], present.1.decoder.value[1, 12, 4, 64], present.1.encoder.key[1, 12, 1500, 64], present.1.encoder.value[1, 12, 1500, 64], present.2.decoder.key[1, 12, 4, 64], present.2.decoder.value[1, 12, 4, 64], present.2.encoder.key[1, 12, 1500, 64], present.2.encoder.value[1, 12, 1500, 64], present.3.decoder.key[1, 12, 4, 64], present.3.decoder.value[1, 12, 4, 64], present.3.encoder.key[1, 12, 1500, 64], present.3.encoder.value[1, 12, 1500, 64], present.4.decoder.key[1, 12, 4, 64], present.4.decoder.value[1, 12, 4, 64], present.4.encoder.key[1, 12, 1500, 64], present.4.encoder.value[1, 12, 1500, 64], present.5.decoder.key[1, 12, 4, 64], present.5.decoder.value[1, 12, 4, 64], present.5.encoder.key[1, 12, 1500, 64], present.5.encoder.value[1, 12, 1500, 64], present.6.decoder.key[1, 12, 4, 64], present.6.decoder.value[1, 12, 4, 64], present.6.encoder.key[1, 12, 1500, 64], present.6.encoder.value[1, 12, 1500, 64], present.7.decoder.key[1, 12, 4, 64], present.7.decoder.value[1, 12, 4, 64], present.7.encoder.key[1, 12, 1500, 64], present.7.encoder.value[1, 12, 1500, 64], present.8.decoder.key[1, 12, 4, 64], present.8.decoder.value[1, 12, 4, 64], present.8.encoder.key[1, 12, 1500, 64], present.8.encoder.value[1, 12, 1500, 64], present.9.decoder.key[1, 12, 4, 64], present.9.decoder.value[1, 12, 4, 64], present.9.encoder.key[1, 12, 1500, 64], present.9.encoder.value[1, 12, 1500, 64], present.10.decoder.key[1, 12, 4, 64], present.10.decoder.value[1, 12, 4, 64], present.10.encoder.key[1, 12, 1500, 64], present.10.encoder.value[1, 12, 1500, 64], present.11.decoder.key[1, 12, 4, 64], present.11.decoder.value[1, 12, 4, 64], present.11.encoder.key[1, 12, 1500, 64], present.11.encoder.value[1, 12, 1500, 64] | 51.2 |
onnx/decoder_model_merged_fp16.onnx |
input_ids[1, 4], encoder_hidden_states[1, 1500, 768], past_key_values.0.decoder.key[1, 12, 1, 64], past_key_values.0.decoder.value[1, 12, 1, 64], past_key_values.0.encoder.key[1, 12, 1500, 64], past_key_values.0.encoder.value[1, 12, 1500, 64], past_key_values.1.decoder.key[1, 12, 1, 64], past_key_values.1.decoder.value[1, 12, 1, 64], past_key_values.1.encoder.key[1, 12, 1500, 64], past_key_values.1.encoder.value[1, 12, 1500, 64], past_key_values.2.decoder.key[1, 12, 1, 64], past_key_values.2.decoder.value[1, 12, 1, 64], past_key_values.2.encoder.key[1, 12, 1500, 64], past_key_values.2.encoder.value[1, 12, 1500, 64], past_key_values.3.decoder.key[1, 12, 1, 64], past_key_values.3.decoder.value[1, 12, 1, 64], past_key_values.3.encoder.key[1, 12, 1500, 64], past_key_values.3.encoder.value[1, 12, 1500, 64], past_key_values.4.decoder.key[1, 12, 1, 64], past_key_values.4.decoder.value[1, 12, 1, 64], past_key_values.4.encoder.key[1, 12, 1500, 64], past_key_values.4.encoder.value[1, 12, 1500, 64], past_key_values.5.decoder.key[1, 12, 1, 64], past_key_values.5.decoder.value[1, 12, 1, 64], past_key_values.5.encoder.key[1, 12, 1500, 64], past_key_values.5.encoder.value[1, 12, 1500, 64], past_key_values.6.decoder.key[1, 12, 1, 64], past_key_values.6.decoder.value[1, 12, 1, 64], past_key_values.6.encoder.key[1, 12, 1500, 64], past_key_values.6.encoder.value[1, 12, 1500, 64], past_key_values.7.decoder.key[1, 12, 1, 64], past_key_values.7.decoder.value[1, 12, 1, 64], past_key_values.7.encoder.key[1, 12, 1500, 64], past_key_values.7.encoder.value[1, 12, 1500, 64], past_key_values.8.decoder.key[1, 12, 1, 64], past_key_values.8.decoder.value[1, 12, 1, 64], past_key_values.8.encoder.key[1, 12, 1500, 64], past_key_values.8.encoder.value[1, 12, 1500, 64], past_key_values.9.decoder.key[1, 12, 1, 64], past_key_values.9.decoder.value[1, 12, 1, 64], past_key_values.9.encoder.key[1, 12, 1500, 64], past_key_values.9.encoder.value[1, 12, 1500, 64], past_key_values.10.decoder.key[1, 12, 1, 64], past_key_values.10.decoder.value[1, 12, 1, 64], past_key_values.10.encoder.key[1, 12, 1500, 64], past_key_values.10.encoder.value[1, 12, 1500, 64], past_key_values.11.decoder.key[1, 12, 1, 64], past_key_values.11.decoder.value[1, 12, 1, 64], past_key_values.11.encoder.key[1, 12, 1500, 64], past_key_values.11.encoder.value[1, 12, 1500, 64], use_cache_branch[1] | logits[1, 4, 51865], present.0.decoder.key[1, 12, 4, 64], present.0.decoder.value[1, 12, 4, 64], present.0.encoder.key[1, 12, 1500, 64], present.0.encoder.value[1, 12, 1500, 64], present.1.decoder.key[1, 12, 4, 64], present.1.decoder.value[1, 12, 4, 64], present.1.encoder.key[1, 12, 1500, 64], present.1.encoder.value[1, 12, 1500, 64], present.2.decoder.key[1, 12, 4, 64], present.2.decoder.value[1, 12, 4, 64], present.2.encoder.key[1, 12, 1500, 64], present.2.encoder.value[1, 12, 1500, 64], present.3.decoder.key[1, 12, 4, 64], present.3.decoder.value[1, 12, 4, 64], present.3.encoder.key[1, 12, 1500, 64], present.3.encoder.value[1, 12, 1500, 64], present.4.decoder.key[1, 12, 4, 64], present.4.decoder.value[1, 12, 4, 64], present.4.encoder.key[1, 12, 1500, 64], present.4.encoder.value[1, 12, 1500, 64], present.5.decoder.key[1, 12, 4, 64], present.5.decoder.value[1, 12, 4, 64], present.5.encoder.key[1, 12, 1500, 64], present.5.encoder.value[1, 12, 1500, 64], present.6.decoder.key[1, 12, 4, 64], present.6.decoder.value[1, 12, 4, 64], present.6.encoder.key[1, 12, 1500, 64], present.6.encoder.value[1, 12, 1500, 64], present.7.decoder.key[1, 12, 4, 64], present.7.decoder.value[1, 12, 4, 64], present.7.encoder.key[1, 12, 1500, 64], present.7.encoder.value[1, 12, 1500, 64], present.8.decoder.key[1, 12, 4, 64], present.8.decoder.value[1, 12, 4, 64], present.8.encoder.key[1, 12, 1500, 64], present.8.encoder.value[1, 12, 1500, 64], present.9.decoder.key[1, 12, 4, 64], present.9.decoder.value[1, 12, 4, 64], present.9.encoder.key[1, 12, 1500, 64], present.9.encoder.value[1, 12, 1500, 64], present.10.decoder.key[1, 12, 4, 64], present.10.decoder.value[1, 12, 4, 64], present.10.encoder.key[1, 12, 1500, 64], present.10.encoder.value[1, 12, 1500, 64], present.11.decoder.key[1, 12, 4, 64], present.11.decoder.value[1, 12, 4, 64], present.11.encoder.key[1, 12, 1500, 64], present.11.encoder.value[1, 12, 1500, 64] | 104.0 |
onnx/decoder_model_merged_quantized.onnx |
input_ids[1, 4], encoder_hidden_states[1, 1500, 768], past_key_values.0.decoder.key[1, 12, 1, 64], past_key_values.0.decoder.value[1, 12, 1, 64], past_key_values.0.encoder.key[1, 12, 1500, 64], past_key_values.0.encoder.value[1, 12, 1500, 64], past_key_values.1.decoder.key[1, 12, 1, 64], past_key_values.1.decoder.value[1, 12, 1, 64], past_key_values.1.encoder.key[1, 12, 1500, 64], past_key_values.1.encoder.value[1, 12, 1500, 64], past_key_values.2.decoder.key[1, 12, 1, 64], past_key_values.2.decoder.value[1, 12, 1, 64], past_key_values.2.encoder.key[1, 12, 1500, 64], past_key_values.2.encoder.value[1, 12, 1500, 64], past_key_values.3.decoder.key[1, 12, 1, 64], past_key_values.3.decoder.value[1, 12, 1, 64], past_key_values.3.encoder.key[1, 12, 1500, 64], past_key_values.3.encoder.value[1, 12, 1500, 64], past_key_values.4.decoder.key[1, 12, 1, 64], past_key_values.4.decoder.value[1, 12, 1, 64], past_key_values.4.encoder.key[1, 12, 1500, 64], past_key_values.4.encoder.value[1, 12, 1500, 64], past_key_values.5.decoder.key[1, 12, 1, 64], past_key_values.5.decoder.value[1, 12, 1, 64], past_key_values.5.encoder.key[1, 12, 1500, 64], past_key_values.5.encoder.value[1, 12, 1500, 64], past_key_values.6.decoder.key[1, 12, 1, 64], past_key_values.6.decoder.value[1, 12, 1, 64], past_key_values.6.encoder.key[1, 12, 1500, 64], past_key_values.6.encoder.value[1, 12, 1500, 64], past_key_values.7.decoder.key[1, 12, 1, 64], past_key_values.7.decoder.value[1, 12, 1, 64], past_key_values.7.encoder.key[1, 12, 1500, 64], past_key_values.7.encoder.value[1, 12, 1500, 64], past_key_values.8.decoder.key[1, 12, 1, 64], past_key_values.8.decoder.value[1, 12, 1, 64], past_key_values.8.encoder.key[1, 12, 1500, 64], past_key_values.8.encoder.value[1, 12, 1500, 64], past_key_values.9.decoder.key[1, 12, 1, 64], past_key_values.9.decoder.value[1, 12, 1, 64], past_key_values.9.encoder.key[1, 12, 1500, 64], past_key_values.9.encoder.value[1, 12, 1500, 64], past_key_values.10.decoder.key[1, 12, 1, 64], past_key_values.10.decoder.value[1, 12, 1, 64], past_key_values.10.encoder.key[1, 12, 1500, 64], past_key_values.10.encoder.value[1, 12, 1500, 64], past_key_values.11.decoder.key[1, 12, 1, 64], past_key_values.11.decoder.value[1, 12, 1, 64], past_key_values.11.encoder.key[1, 12, 1500, 64], past_key_values.11.encoder.value[1, 12, 1500, 64], use_cache_branch[1] | logits[1, 4, 51865], present.0.decoder.key[1, 12, 4, 64], present.0.decoder.value[1, 12, 4, 64], present.0.encoder.key[1, 12, 1500, 64], present.0.encoder.value[1, 12, 1500, 64], present.1.decoder.key[1, 12, 4, 64], present.1.decoder.value[1, 12, 4, 64], present.1.encoder.key[1, 12, 1500, 64], present.1.encoder.value[1, 12, 1500, 64], present.2.decoder.key[1, 12, 4, 64], present.2.decoder.value[1, 12, 4, 64], present.2.encoder.key[1, 12, 1500, 64], present.2.encoder.value[1, 12, 1500, 64], present.3.decoder.key[1, 12, 4, 64], present.3.decoder.value[1, 12, 4, 64], present.3.encoder.key[1, 12, 1500, 64], present.3.encoder.value[1, 12, 1500, 64], present.4.decoder.key[1, 12, 4, 64], present.4.decoder.value[1, 12, 4, 64], present.4.encoder.key[1, 12, 1500, 64], present.4.encoder.value[1, 12, 1500, 64], present.5.decoder.key[1, 12, 4, 64], present.5.decoder.value[1, 12, 4, 64], present.5.encoder.key[1, 12, 1500, 64], present.5.encoder.value[1, 12, 1500, 64], present.6.decoder.key[1, 12, 4, 64], present.6.decoder.value[1, 12, 4, 64], present.6.encoder.key[1, 12, 1500, 64], present.6.encoder.value[1, 12, 1500, 64], present.7.decoder.key[1, 12, 4, 64], present.7.decoder.value[1, 12, 4, 64], present.7.encoder.key[1, 12, 1500, 64], present.7.encoder.value[1, 12, 1500, 64], present.8.decoder.key[1, 12, 4, 64], present.8.decoder.value[1, 12, 4, 64], present.8.encoder.key[1, 12, 1500, 64], present.8.encoder.value[1, 12, 1500, 64], present.9.decoder.key[1, 12, 4, 64], present.9.decoder.value[1, 12, 4, 64], present.9.encoder.key[1, 12, 1500, 64], present.9.encoder.value[1, 12, 1500, 64], present.10.decoder.key[1, 12, 4, 64], present.10.decoder.value[1, 12, 4, 64], present.10.encoder.key[1, 12, 1500, 64], present.10.encoder.value[1, 12, 1500, 64], present.11.decoder.key[1, 12, 4, 64], present.11.decoder.value[1, 12, 4, 64], present.11.encoder.key[1, 12, 1500, 64], present.11.encoder.value[1, 12, 1500, 64] | 40.1 |
onnx/encoder_model.onnx |
input_features[1, 80, 3000] | last_hidden_state[1, 1500, 768] | 327.5 |
onnx/encoder_model_fp16.onnx |
input_features[1, 80, 3000] | last_hidden_state[1, 1500, 768] | 504.1 |
Use in the browser
import * as ort from "onnxruntime-web";
const session = await ort.InferenceSession.create("https://huggingface.co/skillsafe-ai/whisper-small/resolve/main/onnx/decoder_model_merged.onnx", { executionProviders: ["webgpu", "wasm"] });
Contract (onnx/decoder_model_merged.onnx): input input_ids int64 ['batch_size', 'decoder_sequence_length'], encoder_hidden_states float32 ['batch_size', 'encoder_sequence_length / 2', 768], past_key_values.0.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.0.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.0.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.0.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.1.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.1.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.1.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.1.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.2.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.2.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.2.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.2.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.3.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.3.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.3.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.3.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.4.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.4.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.4.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.4.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.5.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.5.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.5.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.5.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.6.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.6.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.6.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.6.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.7.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.7.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.7.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.7.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.8.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.8.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.8.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.8.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.9.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.9.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.9.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.9.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.10.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.10.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.10.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.10.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.11.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.11.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length', 64], past_key_values.11.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], past_key_values.11.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], use_cache_branch bool [1] → output logits float32 ['batch_size', 'decoder_sequence_length', 51865], present.0.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.0.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.0.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.0.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.1.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.1.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.1.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.1.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.2.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.2.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.2.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.2.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.3.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.3.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.3.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.3.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.4.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.4.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.4.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.4.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.5.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.5.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.5.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.5.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.6.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.6.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.6.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.6.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.7.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.7.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.7.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.7.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.8.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.8.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.8.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.8.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.9.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.9.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.9.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.9.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.10.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.10.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.10.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.10.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.11.decoder.key float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.11.decoder.value float32 ['batch_size', 12, 'past_decoder_sequence_length + 1', 64], present.11.encoder.key float32 ['batch_size', 12, 'encoder_sequence_length_out', 64], present.11.encoder.value float32 ['batch_size', 12, 'encoder_sequence_length_out', 64]. Opset 14.
Licence and attribution
Whisper small: Copyright 2022 OpenAI, Apache License 2.0 (https://github.com/openai/whisper); ONNX export by onnx-community (https://huggingface.co/onnx-community/whisper-small).
Licence: Apache-2.0 — notice: https://github.com/openai/whisper/blob/main/LICENSE. The conversion recipe and this model card are part of the SkillSafe repository and carry its licence; the weights remain under the upstream licence above.
The full manifest.json in this repo records the recipe, sources, toolchain
(including the uv.lock hash) and per-file verification numbers.
- Downloads last month
- 27