yacht commited on
Commit
5dd71d7
·
verified ·
1 Parent(s): 1eeba74

Card: at-a-glance table without an empty header row (HTML table, as on the OCR cards)

Browse files
Files changed (1) hide show
  1. README.md +14 -14
README.md CHANGED
@@ -25,20 +25,20 @@ Streaming Thai speech-to-text with 0.64B parameters. It writes Thai speech down
25
 
26
  **Built on NVIDIA Nemotron.** The base model is NVIDIA's [`nvidia/nemotron-3.5-asr-streaming-0.6b`](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b), released under the OpenMDW License Agreement, version 1.1. That licence is included as [`LICENSE`](LICENSE), and NVIDIA's original model card and sub-cards are kept unchanged in [`nvidia/`](nvidia/).
27
 
28
- | | |
29
- |---|---|
30
- | Architecture | Cache-aware streaming FastConformer encoder + RNN-Transducer decoder, prompt-conditioned (NeMo `EncDecRNNTBPEModelWithPrompt`) |
31
- | Parameters | 640.4 million (0.64B): encoder 609.1 million (24 FastConformer layers, 1,024 wide), language-prompt layer 4.5 million, RNN-T decoder 16.2 million (2-layer LSTM of 640) and joint network 10.7 million |
32
- | Accuracy, 560 ms | % of characters wrong: FLEURS 8.47, GigaSpeech2 15.35, TVSpeech 12.41, LOTUSDIS 20.77, Isan 34.90 (average 18.38; see [Results](#results)) |
33
- | Latency settings | 80, 320, 560 and 1120 ms (`att_context_size` `[56,0]`, `[56,3]`, `[56,6]`, `[56,13]`), chosen at decode time |
34
- | Language prompts | `th-TH`: everything in Thai script. `th-mix`: English words in Latin letters |
35
- | Tokenizer | 14,988 pieces: NVIDIA's 13,087 multilingual pieces and 1,901 new Thai pieces |
36
- | Training data | 12,373 hours of Thai speech and 961 hours of English (LibriSpeech) |
37
- | Input | 16 kHz mono audio (128 log-mel bands); any length when cut at pauses, as the included server does |
38
- | Files | Two `.nemo` files of 2.56 GB each, 32-bit weights (see [Use](#use)). An ONNX export, full precision and int8, is at [iapp/OpenThai2.0-ASR-Streaming-ONNX](https://huggingface.co/iapp/OpenThai2.0-ASR-Streaming-ONNX) |
39
- | Speed | One NVIDIA H100: files about 257× faster than real time, about 20 live streams at once at 560 ms. A CPU with 8 threads: files about 4× real time, a live stream keeps up at 560 and 320 ms |
40
- | Memory | About 2.6 GB for the weights; about 6.5 GB of RAM in all on a CPU |
41
- | Licence | OpenMDW-1.1 |
42
 
43
  ## Results
44
 
 
25
 
26
  **Built on NVIDIA Nemotron.** The base model is NVIDIA's [`nvidia/nemotron-3.5-asr-streaming-0.6b`](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b), released under the OpenMDW License Agreement, version 1.1. That licence is included as [`LICENSE`](LICENSE), and NVIDIA's original model card and sub-cards are kept unchanged in [`nvidia/`](nvidia/).
27
 
28
+ <table>
29
+ <tr><td><b>Architecture</b></td><td>Cache-aware streaming FastConformer encoder + RNN-Transducer decoder, prompt-conditioned (NeMo <code>EncDecRNNTBPEModelWithPrompt</code>)</td></tr>
30
+ <tr><td><b>Parameters</b></td><td>640.4 million (0.64B): encoder 609.1 million (24 FastConformer layers, 1,024 wide), language-prompt layer 4.5 million, RNN-T decoder 16.2 million (2-layer LSTM of 640) and joint network 10.7 million</td></tr>
31
+ <tr><td><b>Accuracy, 560 ms</b></td><td>% of characters wrong: FLEURS 8.47, GigaSpeech2 15.35, TVSpeech 12.41, LOTUSDIS 20.77, Isan 34.90 (average 18.38; see <a href="#results">Results</a>)</td></tr>
32
+ <tr><td><b>Latency settings</b></td><td>80, 320, 560 and 1120 ms (<code>att_context_size</code> <code>[56,0]</code>, <code>[56,3]</code>, <code>[56,6]</code>, <code>[56,13]</code>), chosen at decode time</td></tr>
33
+ <tr><td><b>Language prompts</b></td><td><code>th-TH</code>: everything in Thai script. <code>th-mix</code>: English words in Latin letters</td></tr>
34
+ <tr><td><b>Tokenizer</b></td><td>14,988 pieces: NVIDIA's 13,087 multilingual pieces and 1,901 new Thai pieces</td></tr>
35
+ <tr><td><b>Training data</b></td><td>12,373 hours of Thai speech and 961 hours of English (LibriSpeech)</td></tr>
36
+ <tr><td><b>Input</b></td><td>16 kHz mono audio (128 log-mel bands); any length when cut at pauses, as the included server does</td></tr>
37
+ <tr><td><b>Files</b></td><td>Two <code>.nemo</code> files of 2.56 GB each, 32-bit weights (see <a href="#use">Use</a>). An ONNX export, full precision and int8, is at <a href="https://huggingface.co/iapp/OpenThai2.0-ASR-Streaming-ONNX">iapp/OpenThai2.0-ASR-Streaming-ONNX</a></td></tr>
38
+ <tr><td><b>Speed</b></td><td>One NVIDIA H100: files about 257× faster than real time, about 20 live streams at once at 560 ms. A CPU with 8 threads: files about 4× real time, a live stream keeps up at 560 and 320 ms</td></tr>
39
+ <tr><td><b>Memory</b></td><td>About 2.6 GB for the weights; about 6.5 GB of RAM in all on a CPU</td></tr>
40
+ <tr><td><b>Licence</b></td><td>OpenMDW-1.1</td></tr>
41
+ </table>
42
 
43
  ## Results
44