Card: data line

7a6fd02 verified 5 days ago

10.4 kB

---
language:
- tr
license: cc-by-sa-4.0
library_name: sherpa-onnx
pipeline_tag: automatic-speech-recognition
base_model: atasoglu/seda-v0.2
tags:
- speech
- asr
- streaming
- zipformer
- transducer
- icefall
- sherpa-onnx
- onnx
- k2
- knowledge-distillation
datasets:
- issai/Turkish_Speech_Corpus
- cloud0day3/alania-synthetic-speech-tr
- ysdede/commonvoice_17_tr_fixed
- espnet/yodas2
- atasoglu/yodas-tr-whisper-labels
model-index:
- name: seda-v0.2-tiny
  results:
  - task:
      type: automatic-speech-recognition
      name: Speech Recognition
    dataset:
      type: google/fleurs
      name: FLEURS (tr_tr), test
      config: tr_tr
      split: test
    metrics:
    - type: wer
      value: 10.51
      name: WER (mode "accurate")
    - type: wer
      value: 11.24
      name: WER (mode "balanced", streaming)
    - type: wer
      value: 17.30
      name: WER (mode "fast", streaming)
---
# seda-v0.2-tiny: streaming Turkish speech recognition in 4.5M parameters

**seda-v0.2-tiny** is a 15× smaller version of [seda-v0.2](https://huggingface.co/atasoglu/seda-v0.2) for slow CPUs, phones and the browser. It is a streaming Zipformer2 transducer with **4.5M parameters** (seda-v0.2: 66M), distilled from seda-v0.2 and exported to int8 ONNX for [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx). The acoustic model is about **7 MB**.

- **About as accurate as Whisper small on Turkish**, with 50× fewer parameters and streaming output: 11.24 WER on FLEURS-TR test against 11.87 for Whisper small, and 16.88 against 23.25 on broadcast audio (MediaSpeech).
- **Better than seda-v0.1** (66M parameters, 11.60 WER), while 15× smaller.
- **Runs in a phone's browser:** with sherpa-onnx WASM on one thread of a low-end Android phone, the 0.64 s chunk model with beam search runs at 0.77× real time.
- The same language model, voice activity detector and text formatter setup as seda-v0.2, so every [sedalib](https://github.com/atasoglu/seda) mode works.

## Usage

With [sedalib](https://github.com/atasoglu/seda) 1.1 or newer:

```bash
pip install "sedalib[format]"  # without [format]: no punctuation, casing or digits
```

```python
from sedalib import Recognizer, read_wav
samples, sample_rate = read_wav("audio.wav")
recognizer = Recognizer.from_pretrained("atasoglu/seda-v0.2-tiny", mode="fast")
print(recognizer.transcribe(samples, sample_rate))
```

For live audio, feed the stream small pieces as they arrive:

```python
stream = recognizer.stream()
for chunk in microphone_chunks():  # e.g. 100 ms of 16 kHz audio
    print(stream.accept(chunk, sample_rate=16000))  # partial text
    if stream.is_endpoint:
        print("final:", stream.end_utterance())
```

`from_pretrained` downloads only the files of the chosen options: `fast` needs 7 MB, the language model adds 20 MB. The files are plain sherpa-onnx streaming transducer files, so `sherpa_onnx.OnlineRecognizer.from_transducer` loads them directly. The LM settings are `modified_beam_search`, `lm_scale=0.6` and `lodr_scale=-0.5`.

### Modes and chunk sizes

| Mode | Decoding | FLEURS test WER |
|---|---|---|
| `fast` | 0.64 s chunks, greedy | 17.30 |
| `balanced` (default) | 1.28 s chunks, beam 4 + LM | 11.24 |
| `accurate` | 0.64 s chunks live, then a second pass with full context, beam + LM | 10.51 |

Shorter chunks react sooner but cost more CPU per second of audio and are a little less accurate. Pick them with `chunk=`:

| Chunk | Directory | FLEURS test WER, greedy | with beam 4 + LM |
|---|---|---|---|
| 0.16 s (`chunk=8`) | `c8/` | 21.08 | 13.99 |
| 0.32 s (`chunk=16`) | `c16/` | 19.20 | 12.85 |
| 0.64 s (`chunk=32`) | `c32/` | 17.30 | 11.67 |
| 1.28 s (`chunk=64`) | root | 16.50 | 11.24 |

- **Measurement:** sherpa-onnx int8 on CPU, audio followed by 1 s of silence. The `accurate` row decodes each utterance with full context (a 20.48 s chunk) and beam 4 + LM. The 0.16 and 0.32 s rows were measured with 0.64 s of left context; the released `c8/` and `c16/` use 1.28 s, which was 0.5 to 0.8 WER better on FLEURS dev.
- **Slow devices:** on a low-end Android phone (sherpa-onnx WASM, one thread), 0.16 s chunks with beam search needed 1.07 s of compute per second of audio and fell behind; 0.32 s greedy needed 0.89 s and 0.64 s with beam search 0.77 s. Use 0.32 or 0.64 s chunks there; 0.16 s suits fast CPUs where the earliest reaction matters.

## Evaluation

WER (%). Lower is better. References and hypotheses were normalized the same way: lowercase, no punctuation, numbers spelled out. Whisper ran with [faster-whisper](https://github.com/SYSTRAN/faster-whisper) on a GPU and decodes whole recordings offline; seda-v0.1 is in its released setup (0.64 s chunks, beam 4 + LM).

| Test set | Whisper small (244M) | seda-v0.1 (66M) | **seda-v0.2-tiny (4.5M)** `balanced` | **`accurate`** | seda-v0.2 (66M) `accurate` |
|---|---|---|---|---|---|
| FLEURS-TR test (743 utt.) | 11.87 | 11.60 | 11.24 | **10.51** | 6.38 |
| MediaSpeech TR (broadcast, 10 h) | 23.25 | 17.50 | **16.88** ¹ | | 11.27 |

¹ 0.64 s chunks, beam 4 + LM.

More test sets, 0.64 s chunks:

| Test set | Greedy | Beam 4 + LM | seda-v0.2 teacher, greedy |
|---|---|---|---|
| Khan Academy TR test ² | 14.61 | 11.63 | 8.72 |
| ISSAI TSC test | 23.33 | 20.68 | 14.86 |

² Khan Academy is not in this model's training data (it is in seda-v0.2's).

**The gap to seda-v0.2 is the price of 15× fewer parameters.** The tiny model makes about 1.6× as many errors as its teacher. Use seda-v0.2 where it fits; use this model where it does not.

## Model details

| | |
|---|---|
| Architecture | Zipformer2 encoder + stateless decoder + joiner (pruned RNN-T), causal, trained with dynamic chunks |
| Parameters | 4.49M |
| Encoder | 6 stacks, layers `1,1,2,2,1,1`, dim 128, feedforward 256, 4 heads |
| Input | 80-dim log-mel fbank, 16 kHz mono (other rates are resampled) |
| Output | 500 BPE tokens (the same as seda-v0.2). Lowercase spoken-form Turkish, without punctuation; numbers are words |
| Quantization | int8 (dynamic) encoder and joiner, fp32 decoder |
| Language model | seda-v0.2's LSTM 2×1024 RNN LM (int8, 17 MB) for shallow fusion, token bigram for LODR |
| Formatter | 2-layer Turkish BERT ([ytu-ce-cosmos/turkish-tiny-bert-uncased](https://huggingface.co/ytu-ce-cosmos/turkish-tiny-bert-uncased)) with heads for punctuation, casing and apostrophes; int8, 4.7 MB |

## Files

| File | Description |
|---|---|
| `encoder.int8.onnx`, `decoder.onnx`, `joiner.int8.onnx`, `tokens.txt` | Streaming model, 1.28 s chunks (`balanced`) |
| `c32/`, `c16/`, `c8/` | The same weights, exported with 0.64, 0.32 and 0.16 s chunks |
| `final/` | The same weights, exported with 5.12 s chunks and 20.48 s left context (second pass of `accurate`) |
| `lm/rnnlm.int8.onnx`, `lm/2gram.fst` | LSTM language model and token bigram for LODR, from seda-v0.2 |
| `vad/silero_vad.onnx` | Silero VAD v6, 16 kHz (MIT, see `vad/LICENSE`) |
| `formatter/` | Text formatter (int8 ONNX token classifier) and its tokenizer |

## Training

- **Knowledge distillation from seda-v0.2.** The student learns the transcripts (pruned RNN-T loss) and, with weight 0.5, the teacher's encoder outputs (hidden-state distillation through a linear projection). The teacher's outputs were computed once for all training audio.
- **Acoustic data: 1,913 h:**
  - [ISSAI Turkish Speech Corpus](https://huggingface.co/datasets/issai/Turkish_Speech_Corpus) (209 h)
  - [Alania synthetic speech](https://huggingface.co/datasets/cloud0day3/alania-synthetic-speech-tr), CC BY 4.0 subset (166 h)
  - [Common Voice 17](https://huggingface.co/datasets/ysdede/commonvoice_17_tr_fixed) (11 h)
  - [YODAS2](https://huggingface.co/datasets/espnet/yodas2) Turkish, 1,527 h: YouTube videos under CC BY 3.0, labeled by Whisper large-v3-turbo and filtered by agreement with the video's subtitles. The labels are published as [atasoglu/yodas-tr-whisper-labels](https://huggingface.co/datasets/atasoglu/yodas-tr-whisper-labels).
  - Evaluation sentences were held out the same way as for seda-v0.2.
- **Recipe:**
  - Trained from scratch for 11 epochs: base LR 0.035, Eden schedule, fp16, max duration 1500 s, utterances up to 30 s.
  - Chunk sizes 0.16/0.32/0.64/1.28 s and full context, sampled per batch, so one set of weights serves every export.
  - MUSAN noise augmentation on the student's input only.
  - The release weights average the model over epochs 10 and 11.
- **Language model and formatter:** the LM is seda-v0.2's. The formatter is a 2-layer BERT fine-tuned on Wikipedia, FineWeb-2 and the YODAS Whisper labels, like seda-v0.2's 4-layer one.

## Limitations

- **Less accurate than seda-v0.2:** about 1.6× the errors. Noisy, reverberant and telephone audio were not measured for this model.
- **Foreign names and abbreviations** are the main error source. The model often spells them by sound.
- **The formatter places commas poorly** (F1 0.38 on FLEURS dev); sentence ends, casing and apostrophes are close to seda-v0.2's.
- **Turkish only,** 16 kHz.

## License

The model files are released under **CC BY-SA 4.0**:
- The language model and the formatter were trained on Wikipedia (CC BY-SA).
- The acoustic model's own training data has no share-alike source (Khan Academy, CC BY-SA 3.0, was removed), but it was distilled from seda-v0.2, whose training data includes Khan Academy.
- `vad/silero_vad.onnx` is Silero VAD, MIT license (`vad/LICENSE`).
- Other sources and their licenses: YODAS2 CC BY 3.0 (derived from YouTube videos by their creators); ISSAI Turkish Speech Corpus CC BY 4.0; Alania speech CC BY 4.0; Common Voice CC0; FineWeb-2 ODC-By 1.0; MUSAN CC BY 4.0 (augmentation only); Whisper MIT (labeler only); [ytu-ce-cosmos/turkish-tiny-bert-uncased](https://huggingface.co/ytu-ce-cosmos/turkish-tiny-bert-uncased) (YTU COSMOS) MIT, base of the formatter.

## Acknowledgements and citation

Built with [k2](https://github.com/k2-fsa/k2), [icefall](https://github.com/k2-fsa/icefall), [lhotse](https://github.com/lhotse-speech/lhotse), [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx), [Silero VAD](https://github.com/snakers4/silero-vad) and [Whisper](https://github.com/openai/whisper). The formatter builds on the Turkish BERT models of [YTU COSMOS](https://huggingface.co/ytu-ce-cosmos).

See the [seda-v0.2 card](https://huggingface.co/atasoglu/seda-v0.2) for the dataset citations (YODAS, ISSAI TSC, Common Voice, MUSAN, Zipformer).

	---
	language:
	- tr
	license: cc-by-sa-4.0
	library_name: sherpa-onnx
	pipeline_tag: automatic-speech-recognition
	base_model: atasoglu/seda-v0.2
	tags:
	- speech
	- asr
	- streaming
	- zipformer
	- transducer
	- icefall
	- sherpa-onnx
	- onnx
	- k2
	- knowledge-distillation
	datasets:
	- issai/Turkish_Speech_Corpus
	- cloud0day3/alania-synthetic-speech-tr
	- ysdede/commonvoice_17_tr_fixed
	- espnet/yodas2
	- atasoglu/yodas-tr-whisper-labels
	model-index:
	- name: seda-v0.2-tiny
	results:
	- task:
	type: automatic-speech-recognition
	name: Speech Recognition
	dataset:
	type: google/fleurs
	name: FLEURS (tr_tr), test
	config: tr_tr
	split: test
	metrics:
	- type: wer
	value: 10.51
	name: WER (mode "accurate")
	- type: wer
	value: 11.24
	name: WER (mode "balanced", streaming)
	- type: wer
	value: 17.30
	name: WER (mode "fast", streaming)
	---

	# seda-v0.2-tiny: streaming Turkish speech recognition in 4.5M parameters

	seda-v0.2-tiny is a 15× smaller version of [seda-v0.2](https://huggingface.co/atasoglu/seda-v0.2) for slow CPUs, phones and the browser. It is a streaming Zipformer2 transducer with 4.5M parameters (seda-v0.2: 66M), distilled from seda-v0.2 and exported to int8 ONNX for [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx). The acoustic model is about 7 MB.

	- About as accurate as Whisper small on Turkish, with 50× fewer parameters and streaming output: 11.24 WER on FLEURS-TR test against 11.87 for Whisper small, and 16.88 against 23.25 on broadcast audio (MediaSpeech).
	- Better than seda-v0.1 (66M parameters, 11.60 WER), while 15× smaller.
	- Runs in a phone's browser: with sherpa-onnx WASM on one thread of a low-end Android phone, the 0.64 s chunk model with beam search runs at 0.77× real time.
	- The same language model, voice activity detector and text formatter setup as seda-v0.2, so every [sedalib](https://github.com/atasoglu/seda) mode works.

	## Usage

	With [sedalib](https://github.com/atasoglu/seda) 1.1 or newer:

	```bash
	pip install "sedalib[format]" # without [format]: no punctuation, casing or digits
	```

	```python
	from sedalib import Recognizer, read_wav

	samples, sample_rate = read_wav("audio.wav")

	recognizer = Recognizer.from_pretrained("atasoglu/seda-v0.2-tiny", mode="fast")
	print(recognizer.transcribe(samples, sample_rate))
	```

	For live audio, feed the stream small pieces as they arrive:

	```python
	stream = recognizer.stream()

	for chunk in microphone_chunks(): # e.g. 100 ms of 16 kHz audio
	print(stream.accept(chunk, sample_rate=16000)) # partial text
	if stream.is_endpoint:
	print("final:", stream.end_utterance())
	```

	`from_pretrained` downloads only the files of the chosen options: `fast` needs 7 MB, the language model adds 20 MB. The files are plain sherpa-onnx streaming transducer files, so `sherpa_onnx.OnlineRecognizer.from_transducer` loads them directly. The LM settings are `modified_beam_search`, `lm_scale=0.6` and `lodr_scale=-0.5`.

	### Modes and chunk sizes

	\| Mode \| Decoding \| FLEURS test WER \|
	\|---\|---\|---\|
	\| `fast` \| 0.64 s chunks, greedy \| 17.30 \|
	\| `balanced` (default) \| 1.28 s chunks, beam 4 + LM \| 11.24 \|
	\| `accurate` \| 0.64 s chunks live, then a second pass with full context, beam + LM \| 10.51 \|

	Shorter chunks react sooner but cost more CPU per second of audio and are a little less accurate. Pick them with `chunk=`:

	\| Chunk \| Directory \| FLEURS test WER, greedy \| with beam 4 + LM \|
	\|---\|---\|---\|---\|
	\| 0.16 s (`chunk=8`) \| `c8/` \| 21.08 \| 13.99 \|
	\| 0.32 s (`chunk=16`) \| `c16/` \| 19.20 \| 12.85 \|
	\| 0.64 s (`chunk=32`) \| `c32/` \| 17.30 \| 11.67 \|
	\| 1.28 s (`chunk=64`) \| root \| 16.50 \| 11.24 \|

	- Measurement: sherpa-onnx int8 on CPU, audio followed by 1 s of silence. The `accurate` row decodes each utterance with full context (a 20.48 s chunk) and beam 4 + LM. The 0.16 and 0.32 s rows were measured with 0.64 s of left context; the released `c8/` and `c16/` use 1.28 s, which was 0.5 to 0.8 WER better on FLEURS dev.
	- Slow devices: on a low-end Android phone (sherpa-onnx WASM, one thread), 0.16 s chunks with beam search needed 1.07 s of compute per second of audio and fell behind; 0.32 s greedy needed 0.89 s and 0.64 s with beam search 0.77 s. Use 0.32 or 0.64 s chunks there; 0.16 s suits fast CPUs where the earliest reaction matters.

	## Evaluation

	WER (%). Lower is better. References and hypotheses were normalized the same way: lowercase, no punctuation, numbers spelled out. Whisper ran with [faster-whisper](https://github.com/SYSTRAN/faster-whisper) on a GPU and decodes whole recordings offline; seda-v0.1 is in its released setup (0.64 s chunks, beam 4 + LM).

	\| Test set \| Whisper small (244M) \| seda-v0.1 (66M) \| seda-v0.2-tiny (4.5M) `balanced` \| `accurate` \| seda-v0.2 (66M) `accurate` \|
	\|---\|---\|---\|---\|---\|---\|
	\| FLEURS-TR test (743 utt.) \| 11.87 \| 11.60 \| 11.24 \| 10.51 \| 6.38 \|
	\| MediaSpeech TR (broadcast, 10 h) \| 23.25 \| 17.50 \| 16.88 ¹ \| \| 11.27 \|

	¹ 0.64 s chunks, beam 4 + LM.

	More test sets, 0.64 s chunks:

	\| Test set \| Greedy \| Beam 4 + LM \| seda-v0.2 teacher, greedy \|
	\|---\|---\|---\|---\|
	\| Khan Academy TR test ² \| 14.61 \| 11.63 \| 8.72 \|
	\| ISSAI TSC test \| 23.33 \| 20.68 \| 14.86 \|

	² Khan Academy is not in this model's training data (it is in seda-v0.2's).

	The gap to seda-v0.2 is the price of 15× fewer parameters. The tiny model makes about 1.6× as many errors as its teacher. Use seda-v0.2 where it fits; use this model where it does not.

	## Model details

	\| \| \|
	\|---\|---\|
	\| Architecture \| Zipformer2 encoder + stateless decoder + joiner (pruned RNN-T), causal, trained with dynamic chunks \|
	\| Parameters \| 4.49M \|
	\| Encoder \| 6 stacks, layers `1,1,2,2,1,1`, dim 128, feedforward 256, 4 heads \|
	\| Input \| 80-dim log-mel fbank, 16 kHz mono (other rates are resampled) \|
	\| Output \| 500 BPE tokens (the same as seda-v0.2). Lowercase spoken-form Turkish, without punctuation; numbers are words \|
	\| Quantization \| int8 (dynamic) encoder and joiner, fp32 decoder \|
	\| Language model \| seda-v0.2's LSTM 2×1024 RNN LM (int8, 17 MB) for shallow fusion, token bigram for LODR \|
	\| Formatter \| 2-layer Turkish BERT ([ytu-ce-cosmos/turkish-tiny-bert-uncased](https://huggingface.co/ytu-ce-cosmos/turkish-tiny-bert-uncased)) with heads for punctuation, casing and apostrophes; int8, 4.7 MB \|

	## Files

	\| File \| Description \|
	\|---\|---\|
	\| `encoder.int8.onnx`, `decoder.onnx`, `joiner.int8.onnx`, `tokens.txt` \| Streaming model, 1.28 s chunks (`balanced`) \|
	\| `c32/`, `c16/`, `c8/` \| The same weights, exported with 0.64, 0.32 and 0.16 s chunks \|
	\| `final/` \| The same weights, exported with 5.12 s chunks and 20.48 s left context (second pass of `accurate`) \|
	\| `lm/rnnlm.int8.onnx`, `lm/2gram.fst` \| LSTM language model and token bigram for LODR, from seda-v0.2 \|
	\| `vad/silero_vad.onnx` \| Silero VAD v6, 16 kHz (MIT, see `vad/LICENSE`) \|
	\| `formatter/` \| Text formatter (int8 ONNX token classifier) and its tokenizer \|

	## Training

	- Knowledge distillation from seda-v0.2. The student learns the transcripts (pruned RNN-T loss) and, with weight 0.5, the teacher's encoder outputs (hidden-state distillation through a linear projection). The teacher's outputs were computed once for all training audio.
	- Acoustic data: 1,913 h:
	- [ISSAI Turkish Speech Corpus](https://huggingface.co/datasets/issai/Turkish_Speech_Corpus) (209 h)
	- [Alania synthetic speech](https://huggingface.co/datasets/cloud0day3/alania-synthetic-speech-tr), CC BY 4.0 subset (166 h)
	- [Common Voice 17](https://huggingface.co/datasets/ysdede/commonvoice_17_tr_fixed) (11 h)
	- [YODAS2](https://huggingface.co/datasets/espnet/yodas2) Turkish, 1,527 h: YouTube videos under CC BY 3.0, labeled by Whisper large-v3-turbo and filtered by agreement with the video's subtitles. The labels are published as [atasoglu/yodas-tr-whisper-labels](https://huggingface.co/datasets/atasoglu/yodas-tr-whisper-labels).
	- Evaluation sentences were held out the same way as for seda-v0.2.
	- Recipe:
	- Trained from scratch for 11 epochs: base LR 0.035, Eden schedule, fp16, max duration 1500 s, utterances up to 30 s.
	- Chunk sizes 0.16/0.32/0.64/1.28 s and full context, sampled per batch, so one set of weights serves every export.
	- MUSAN noise augmentation on the student's input only.
	- The release weights average the model over epochs 10 and 11.
	- Language model and formatter: the LM is seda-v0.2's. The formatter is a 2-layer BERT fine-tuned on Wikipedia, FineWeb-2 and the YODAS Whisper labels, like seda-v0.2's 4-layer one.

	## Limitations

	- Less accurate than seda-v0.2: about 1.6× the errors. Noisy, reverberant and telephone audio were not measured for this model.
	- Foreign names and abbreviations are the main error source. The model often spells them by sound.
	- The formatter places commas poorly (F1 0.38 on FLEURS dev); sentence ends, casing and apostrophes are close to seda-v0.2's.
	- Turkish only, 16 kHz.

	## License

	The model files are released under CC BY-SA 4.0:
	- The language model and the formatter were trained on Wikipedia (CC BY-SA).
	- The acoustic model's own training data has no share-alike source (Khan Academy, CC BY-SA 3.0, was removed), but it was distilled from seda-v0.2, whose training data includes Khan Academy.
	- `vad/silero_vad.onnx` is Silero VAD, MIT license (`vad/LICENSE`).
	- Other sources and their licenses: YODAS2 CC BY 3.0 (derived from YouTube videos by their creators); ISSAI Turkish Speech Corpus CC BY 4.0; Alania speech CC BY 4.0; Common Voice CC0; FineWeb-2 ODC-By 1.0; MUSAN CC BY 4.0 (augmentation only); Whisper MIT (labeler only); [ytu-ce-cosmos/turkish-tiny-bert-uncased](https://huggingface.co/ytu-ce-cosmos/turkish-tiny-bert-uncased) (YTU COSMOS) MIT, base of the formatter.

	## Acknowledgements and citation

	Built with [k2](https://github.com/k2-fsa/k2), [icefall](https://github.com/k2-fsa/icefall), [lhotse](https://github.com/lhotse-speech/lhotse), [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx), [Silero VAD](https://github.com/snakers4/silero-vad) and [Whisper](https://github.com/openai/whisper). The formatter builds on the Turkish BERT models of [YTU COSMOS](https://huggingface.co/ytu-ce-cosmos).

	See the [seda-v0.2 card](https://huggingface.co/atasoglu/seda-v0.2) for the dataset citations (YODAS, ISSAI TSC, Common Voice, MUSAN, Zipformer).