Update README.md
Browse files
README.md
CHANGED
|
@@ -1,7 +1,155 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
base_model: openai/whisper-small
|
| 4 |
pipeline_tag: automatic-speech-recognition
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
|
| 7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
- zh
|
| 6 |
+
- de
|
| 7 |
+
- es
|
| 8 |
+
- ru
|
| 9 |
+
- ko
|
| 10 |
+
- fr
|
| 11 |
+
- ja
|
| 12 |
+
- pt
|
| 13 |
+
- tr
|
| 14 |
+
- pl
|
| 15 |
+
- ca
|
| 16 |
+
- nl
|
| 17 |
+
- ar
|
| 18 |
+
- sv
|
| 19 |
+
- it
|
| 20 |
+
- id
|
| 21 |
+
- hi
|
| 22 |
+
- fi
|
| 23 |
+
- vi
|
| 24 |
+
- he
|
| 25 |
+
- uk
|
| 26 |
+
- el
|
| 27 |
+
- ms
|
| 28 |
+
- cs
|
| 29 |
+
- ro
|
| 30 |
+
- da
|
| 31 |
+
- hu
|
| 32 |
+
- ta
|
| 33 |
+
- 'no'
|
| 34 |
+
- th
|
| 35 |
+
- ur
|
| 36 |
+
- hr
|
| 37 |
+
- bg
|
| 38 |
+
- lt
|
| 39 |
+
- la
|
| 40 |
+
- mi
|
| 41 |
+
- ml
|
| 42 |
+
- cy
|
| 43 |
+
- sk
|
| 44 |
+
- te
|
| 45 |
+
- fa
|
| 46 |
+
- lv
|
| 47 |
+
- bn
|
| 48 |
+
- sr
|
| 49 |
+
- az
|
| 50 |
+
- sl
|
| 51 |
+
- kn
|
| 52 |
+
- et
|
| 53 |
+
- mk
|
| 54 |
+
- br
|
| 55 |
+
- eu
|
| 56 |
+
- is
|
| 57 |
+
- hy
|
| 58 |
+
- ne
|
| 59 |
+
- mn
|
| 60 |
+
- bs
|
| 61 |
+
- kk
|
| 62 |
+
- sq
|
| 63 |
+
- sw
|
| 64 |
+
- gl
|
| 65 |
+
- mr
|
| 66 |
+
- pa
|
| 67 |
+
- si
|
| 68 |
+
- km
|
| 69 |
+
- sn
|
| 70 |
+
- yo
|
| 71 |
+
- so
|
| 72 |
+
- af
|
| 73 |
+
- oc
|
| 74 |
+
- ka
|
| 75 |
+
- be
|
| 76 |
+
- tg
|
| 77 |
+
- sd
|
| 78 |
+
- gu
|
| 79 |
+
- am
|
| 80 |
+
- yi
|
| 81 |
+
- lo
|
| 82 |
+
- uz
|
| 83 |
+
- fo
|
| 84 |
+
- ht
|
| 85 |
+
- ps
|
| 86 |
+
- tk
|
| 87 |
+
- nn
|
| 88 |
+
- mt
|
| 89 |
+
- sa
|
| 90 |
+
- lb
|
| 91 |
+
- my
|
| 92 |
+
- bo
|
| 93 |
+
- tl
|
| 94 |
+
- mg
|
| 95 |
+
- as
|
| 96 |
+
- tt
|
| 97 |
+
- haw
|
| 98 |
+
- ln
|
| 99 |
+
- ha
|
| 100 |
+
- ba
|
| 101 |
+
- jw
|
| 102 |
+
- su
|
| 103 |
+
tags:
|
| 104 |
+
- audio
|
| 105 |
+
- automatic-speech-recognition
|
| 106 |
+
- hf-asr-leaderboard
|
| 107 |
base_model: openai/whisper-small
|
| 108 |
pipeline_tag: automatic-speech-recognition
|
| 109 |
+
model-index:
|
| 110 |
+
- name: whisper-small
|
| 111 |
+
results:
|
| 112 |
+
- task:
|
| 113 |
+
name: Automatic Speech Recognition
|
| 114 |
+
type: automatic-speech-recognition
|
| 115 |
+
dataset:
|
| 116 |
+
name: LibriSpeech (clean)
|
| 117 |
+
type: librispeech_asr
|
| 118 |
+
config: clean
|
| 119 |
+
split: test
|
| 120 |
+
args:
|
| 121 |
+
language: en
|
| 122 |
+
metrics:
|
| 123 |
+
- name: Test WER
|
| 124 |
+
type: wer
|
| 125 |
+
value: 3.432213777886737
|
| 126 |
---
|
| 127 |
+
# Whisper-small OpenVINO IR
|
| 128 |
+
Whisper is a pre-trained model for automatic speech recognition (ASR) and speech translation. Trained on 680k hours
|
| 129 |
+
of labelled data, Whisper models demonstrate a strong ability to generalise to many datasets and domains **without** the need
|
| 130 |
+
for fine-tuning.
|
| 131 |
+
|
| 132 |
+
Whisper was proposed in the paper [Robust Speech Recognition via Large-Scale Weak Supervision](https://arxiv.org/abs/2212.04356)
|
| 133 |
+
by Alec Radford et al from OpenAI. The original code repository can be found [here](https://github.com/openai/whisper).
|
| 134 |
+
|
| 135 |
+
**Disclaimer**: Content for this model card has partly been copied and pasted from [this model card](https://huggingface.co/openai/whisper-small).
|
| 136 |
+
|
| 137 |
+
# Model details
|
| 138 |
+
Whisper is a Transformer based encoder-decoder model, also referred to as a sequence-to-sequence model.
|
| 139 |
+
This model architecture is used in [THIS REPO(Intel)](https://github.com/intel-sandbox/whisper-openvino-cpp/)
|
| 140 |
|
| 141 |
|
| 142 |
+
|
| 143 |
+
| Model Type | n_vocab | n_audio_ctx | n_audio_state | n_audio_head | n_audio_layer | n_text_ctx | n_text_state | n_text_head | n_text_layer | n_mels | Parameters |
|
| 144 |
+
|---------------------------|---------|-------------|---------------|--------------|---------------|------------|--------------|-------------|--------------|--------|------------|
|
| 145 |
+
| whisper_tiny | 51864 | 1500 | 512 | 6 | 4 | 128 | 512 | 6 | 4 | 80 | 39 M |
|
| 146 |
+
| whisper_tiny.en | 51864 | 1500 | 512 | 6 | 4 | 128 | 512 | 6 | 4 | 80 | 39 M |
|
| 147 |
+
| whisper_base | 51864 | 1500 | 512 | 8 | 6 | 128 | 512 | 8 | 6 | 80 | 74 M |
|
| 148 |
+
| whisper_base.en | 51864 | 1500 | 512 | 8 | 6 | 128 | 512 | 8 | 6 | 80 | 74 M |
|
| 149 |
+
| whisper_small | 51864 | 1500 | 512 | 12 | 12 | 128 | 512 | 12 | 12 | 80 | 244 M |
|
| 150 |
+
| whisper_small.en | 51864 | 1500 | 512 | 12 | 12 | 128 | 512 | 12 | 12 | 80 | 244 M |
|
| 151 |
+
| whisper_medium | 51864 | 1500 | 512 | 16 | 24 | 128 | 512 | 16 | 16 | 80 | 769 M |
|
| 152 |
+
| whisper_medium.en | 51864 | 1500 | 512 | 16 | 24 | 128 | 512 | 16 | 16 | 80 | 769 M |
|
| 153 |
+
| whisper_large_v1 | 51864 | 1500 | 512 | 20 | 32 | 128 | 512 | 20 | 20 | 80 | 1550 M |
|
| 154 |
+
| whisper_large_v2 | 51864 | 1500 | 512 | 20 | 32 | 128 | 512 | 20 | 20 | 80 | 1550 M |
|
| 155 |
+
| whisper_large_v3 | 51864 | 1500 | 512 | 20 | 32 | 128 | 512 | 20 | 20 | 80 | 1550 M |
|