Align card with canonical release metadata
Browse files
README.md
CHANGED
|
@@ -7,6 +7,11 @@ license_link: https://huggingface.co/RyeAI/ekko-pnc/blob/main/MODEL_LICENSE.md
|
|
| 7 |
library_name: transformers
|
| 8 |
pipeline_tag: token-classification
|
| 9 |
base_model: jonfd/electra-small-nordic
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
tags:
|
| 11 |
- token-classification
|
| 12 |
- danish
|
|
@@ -27,7 +32,7 @@ A 21,880,202-parameter Danish punctuation and capitalization model with one
|
|
| 27 |
```bash
|
| 28 |
git clone https://github.com/Rye-A1/ekko-danish-stt.git
|
| 29 |
cd ekko-danish-stt
|
| 30 |
-
uv sync --extra pnc
|
| 31 |
uv run ekko tiny audio.wav --pnc --format json
|
| 32 |
```
|
| 33 |
|
|
@@ -35,7 +40,7 @@ The runtime uses the self-contained 22 MB `pnc.int8.onnx`, processes long text
|
|
| 35 |
in overlapping windows, and keeps one output word for every input word so Tiny's
|
| 36 |
native timestamps remain attached.
|
| 37 |
|
| 38 |
-
Pinned revision: `
|
| 39 |
|
| 40 |
Artifact checksums:
|
| 41 |
|
|
|
|
| 7 |
library_name: transformers
|
| 8 |
pipeline_tag: token-classification
|
| 9 |
base_model: jonfd/electra-small-nordic
|
| 10 |
+
datasets:
|
| 11 |
+
- alexandrainst/nst-da
|
| 12 |
+
- alexandrainst/ftspeech
|
| 13 |
+
- CoRal-project/coral-v3
|
| 14 |
+
- wikimedia/wikipedia
|
| 15 |
tags:
|
| 16 |
- token-classification
|
| 17 |
- danish
|
|
|
|
| 32 |
```bash
|
| 33 |
git clone https://github.com/Rye-A1/ekko-danish-stt.git
|
| 34 |
cd ekko-danish-stt
|
| 35 |
+
uv sync --locked --extra pnc
|
| 36 |
uv run ekko tiny audio.wav --pnc --format json
|
| 37 |
```
|
| 38 |
|
|
|
|
| 40 |
in overlapping windows, and keeps one output word for every input word so Tiny's
|
| 41 |
native timestamps remain attached.
|
| 42 |
|
| 43 |
+
Pinned revision: `b1ff0133c7a2bf63ae6288640c267462229a719f`.
|
| 44 |
|
| 45 |
Artifact checksums:
|
| 46 |
|