docs: drop stale pre-publication verification note from model card
Browse files
README.md
CHANGED
|
@@ -89,12 +89,6 @@ Python at inference, all decoding local. It ships in **fp16** (maximum fidelity,
|
|
| 89 |
without punctuation marks, so the decoder never predicts a punctuation token -- there is no
|
| 90 |
setting to enable it. Transcripts are plain, unpunctuated text by design.
|
| 91 |
|
| 92 |
-
**Verification status:** this pack is staged in a private repo, not yet publicly listed. Local
|
| 93 |
-
verification so far covers Mandarin (`zh`) sanity-checked against the upstream architecture and
|
| 94 |
-
bit-stable at fp16/q8_0, with a small (~9% CER) drift at q4_k versus fp16 on the sanity clip;
|
| 95 |
-
Japanese (`ja`), one of the 40 advertised languages, has not yet had a native-speaker listening
|
| 96 |
-
review and must get one before this model is made public.
|
| 97 |
-
|
| 98 |
## ⚙️ How these packs were made
|
| 99 |
|
| 100 |
Converted from [DataoceanAI/dolphin-base](https://huggingface.co/DataoceanAI/dolphin-base) with the OpenASR importer:
|
|
|
|
| 89 |
without punctuation marks, so the decoder never predicts a punctuation token -- there is no
|
| 90 |
setting to enable it. Transcripts are plain, unpunctuated text by design.
|
| 91 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
## ⚙️ How these packs were made
|
| 93 |
|
| 94 |
Converted from [DataoceanAI/dolphin-base](https://huggingface.co/DataoceanAI/dolphin-base) with the OpenASR importer:
|