Update README.md
Browse files
README.md
CHANGED
|
@@ -1,5 +1,16 @@
|
|
| 1 |
---
|
| 2 |
library_name: mlx
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
# whisper-tiny-khmer-mlx-fp32
|
|
@@ -7,6 +18,9 @@ This model was converted to MLX format from [`openai-whisper-tiny`](https://gith
|
|
| 7 |
- [seanghay/khmer_mpwt_speech](https://huggingface.co/datasets/seanghay/khmer_mpwt_speech)
|
| 8 |
- [seanghay/km-speech-corpus](https://huggingface.co/datasets/seanghay/km-speech-corpus)
|
| 9 |
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
__NOTE__ MLX format is usable for M-chip series of Apple.
|
| 12 |
|
|
@@ -32,4 +46,4 @@ You can also use command line in terminal
|
|
| 32 |
|
| 33 |
```bash
|
| 34 |
mlx_whisper --model Kimang18/whisper-tiny-khmer-mlx-fp32 --task transcribe SPEECH_FILE_NAME --fp16 False
|
| 35 |
-
```
|
|
|
|
| 1 |
---
|
| 2 |
library_name: mlx
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
datasets:
|
| 5 |
+
- google/fleurs
|
| 6 |
+
- seanghay/khmer_mpwt_speech
|
| 7 |
+
- seanghay/km-speech-corpus
|
| 8 |
+
- openslr/openslr
|
| 9 |
+
metrics:
|
| 10 |
+
- wer
|
| 11 |
+
tags:
|
| 12 |
+
- mlx
|
| 13 |
+
- Khmer
|
| 14 |
---
|
| 15 |
|
| 16 |
# whisper-tiny-khmer-mlx-fp32
|
|
|
|
| 18 |
- [seanghay/khmer_mpwt_speech](https://huggingface.co/datasets/seanghay/khmer_mpwt_speech)
|
| 19 |
- [seanghay/km-speech-corpus](https://huggingface.co/datasets/seanghay/km-speech-corpus)
|
| 20 |
|
| 21 |
+
It achieves the following __word error rate__ (`wer`) on 2 popular datasets:
|
| 22 |
+
- 0.938 on [google/fleurs](https://huggingface.co/datasets/google/fleurs) `km-kh`, `test` split
|
| 23 |
+
- 0.697 on [openslr/openslr](https://huggingface.co/datasets/openslr/openslr) `SLR42`, `train` split
|
| 24 |
|
| 25 |
__NOTE__ MLX format is usable for M-chip series of Apple.
|
| 26 |
|
|
|
|
| 46 |
|
| 47 |
```bash
|
| 48 |
mlx_whisper --model Kimang18/whisper-tiny-khmer-mlx-fp32 --task transcribe SPEECH_FILE_NAME --fp16 False
|
| 49 |
+
```
|