Add NOTICE
Browse files
NOTICE
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Afrikaans-tuned Whisper model by DigiPhyte (Pty) Ltd
|
| 2 |
+
====================================================
|
| 3 |
+
|
| 4 |
+
This model is a fine-tune / derivative work, distributed under the MIT License
|
| 5 |
+
(see LICENSE). The upstream works below are gratefully acknowledged and their
|
| 6 |
+
licences are preserved as required.
|
| 7 |
+
|
| 8 |
+
1. Base model: OpenAI Whisper
|
| 9 |
+
-----------------------------
|
| 10 |
+
openai/whisper-large-v3 (and large-v3-turbo / medium / small variants)
|
| 11 |
+
Copyright OpenAI. Licensed under the Apache License, Version 2.0.
|
| 12 |
+
https://huggingface.co/openai/whisper-large-v3
|
| 13 |
+
A copy of the Apache 2.0 licence is included as LICENSE-whisper-apache-2.0.txt.
|
| 14 |
+
Modifications by DigiPhyte: fine-tuned with a LoRA adapter for South African
|
| 15 |
+
Afrikaans, then merged into the base weights and converted to the CTranslate2
|
| 16 |
+
format for use with faster-whisper.
|
| 17 |
+
|
| 18 |
+
2. Training data: afrikaans-30s
|
| 19 |
+
-------------------------------
|
| 20 |
+
andreoosthuizen/afrikaans-30s (approx. 56 hours of Afrikaans speech).
|
| 21 |
+
Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).
|
| 22 |
+
https://huggingface.co/datasets/andreoosthuizen/afrikaans-30s
|
| 23 |
+
|
| 24 |
+
3. LoRA adapter (large-v3 model only)
|
| 25 |
+
-------------------------------------
|
| 26 |
+
andreoosthuizen/whisper-large-v3-afrikaans
|
| 27 |
+
Copyright (c) Andre Oosthuizen. Licensed under the MIT License.
|
| 28 |
+
https://huggingface.co/andreoosthuizen/whisper-large-v3-afrikaans
|
| 29 |
+
This attribution applies to the large-v3 Afrikaans model, which reuses this
|
| 30 |
+
adapter. The medium, large-v3-turbo, and small Afrikaans models were fine-tuned
|
| 31 |
+
by DigiPhyte directly on the afrikaans-30s dataset (item 2) and do not include
|
| 32 |
+
this adapter.
|