Add paper link to model card (#1)
Browse files- Add paper link to model card (eea31661e18f9d44e1f00c750d684d0d0dece54d)
Co-authored-by: Niels Rogge <nielsr@users.noreply.huggingface.co>
README.md
CHANGED
|
@@ -1,9 +1,14 @@
|
|
| 1 |
---
|
| 2 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
language:
|
| 4 |
- en
|
|
|
|
| 5 |
pipeline_tag: automatic-speech-recognition
|
| 6 |
-
base_model: nvidia/stt_en_conformer_ctc_small
|
| 7 |
tags:
|
| 8 |
- esp32
|
| 9 |
- esp32-s3
|
|
@@ -15,11 +20,6 @@ tags:
|
|
| 15 |
- conformer
|
| 16 |
- ctc
|
| 17 |
- streaming
|
| 18 |
-
datasets:
|
| 19 |
-
- openslr/librispeech_asr
|
| 20 |
-
- mozilla-foundation/common_voice_17_0
|
| 21 |
-
- facebook/multilingual_librispeech
|
| 22 |
-
- MLCommons/peoples_speech
|
| 23 |
model-index:
|
| 24 |
- name: oido-ctc-small-stream-int8
|
| 25 |
results:
|
|
@@ -60,6 +60,8 @@ accelerator). The encoder processes 1.28 s chunks (32 frames) with 5 s of left c
|
|
| 60 |
stop only the last partial chunk is left to compute, and partial text appears while you talk. It was trained with noise,
|
| 61 |
music, babble and reverberation augmentation, so it keeps the original model's robustness.
|
| 62 |
|
|
|
|
|
|
|
| 63 |
**Code, firmware and tools:** [github.com/lokutor-ai/oido](https://github.com/lokutor-ai/oido) (GPLv3, commercial
|
| 64 |
licenses available). Audio samples: [lokutor-ai.github.io/oido](https://lokutor-ai.github.io/oido/).
|
| 65 |
|
|
@@ -132,4 +134,4 @@ English, AMI, VCTK, People's Speech (transcripts regenerated by parakeet-tdt-0.6
|
|
| 132 |
Derived from NVIDIA's `stt_en_conformer_ctc_small`, licensed CC-BY-4.0. Changes: re-implemented in PyTorch, fine-tuned
|
| 133 |
for streaming and robustness, quantized to int8 and repacked. Because the fine-tuning data includes share-alike
|
| 134 |
corpora (People's Speech, OpenSLR 70 and 83), this model is released under **CC-BY-SA-4.0**. The engine and firmware are
|
| 135 |
-
GPLv3, with commercial licenses from [Lokutor](https://lokutor.com).
|
|
|
|
| 1 |
---
|
| 2 |
+
base_model: nvidia/stt_en_conformer_ctc_small
|
| 3 |
+
datasets:
|
| 4 |
+
- openslr/librispeech_asr
|
| 5 |
+
- mozilla-foundation/common_voice_17_0
|
| 6 |
+
- facebook/multilingual_librispeech
|
| 7 |
+
- MLCommons/peoples_speech
|
| 8 |
language:
|
| 9 |
- en
|
| 10 |
+
license: cc-by-sa-4.0
|
| 11 |
pipeline_tag: automatic-speech-recognition
|
|
|
|
| 12 |
tags:
|
| 13 |
- esp32
|
| 14 |
- esp32-s3
|
|
|
|
| 20 |
- conformer
|
| 21 |
- ctc
|
| 22 |
- streaming
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
model-index:
|
| 24 |
- name: oido-ctc-small-stream-int8
|
| 25 |
results:
|
|
|
|
| 60 |
stop only the last partial chunk is left to compute, and partial text appears while you talk. It was trained with noise,
|
| 61 |
music, babble and reverberation augmentation, so it keeps the original model's robustness.
|
| 62 |
|
| 63 |
+
**Paper:** [Oído: Open-Vocabulary Speech Recognition on a $5 Microcontroller without a Neural Accelerator](https://huggingface.co/papers/2610.13447)
|
| 64 |
+
|
| 65 |
**Code, firmware and tools:** [github.com/lokutor-ai/oido](https://github.com/lokutor-ai/oido) (GPLv3, commercial
|
| 66 |
licenses available). Audio samples: [lokutor-ai.github.io/oido](https://lokutor-ai.github.io/oido/).
|
| 67 |
|
|
|
|
| 134 |
Derived from NVIDIA's `stt_en_conformer_ctc_small`, licensed CC-BY-4.0. Changes: re-implemented in PyTorch, fine-tuned
|
| 135 |
for streaming and robustness, quantized to int8 and repacked. Because the fine-tuning data includes share-alike
|
| 136 |
corpora (People's Speech, OpenSLR 70 and 83), this model is released under **CC-BY-SA-4.0**. The engine and firmware are
|
| 137 |
+
GPLv3, with commercial licenses from [Lokutor](https://lokutor.com).
|