elmamounedieye
commited on
Commit
•
0373b38
1
Parent(s):
c710c03
elmamounedieye/whisper-wol-V5
Browse files
README.md
CHANGED
@@ -1,6 +1,4 @@
|
|
1 |
---
|
2 |
-
language:
|
3 |
-
- multilingual
|
4 |
license: apache-2.0
|
5 |
base_model: serge-wilson/whisper-small-wolof
|
6 |
tags:
|
@@ -10,7 +8,7 @@ datasets:
|
|
10 |
metrics:
|
11 |
- wer
|
12 |
model-index:
|
13 |
-
- name:
|
14 |
results:
|
15 |
- task:
|
16 |
name: Automatic Speech Recognition
|
@@ -24,19 +22,19 @@ model-index:
|
|
24 |
metrics:
|
25 |
- name: Wer
|
26 |
type: wer
|
27 |
-
value:
|
28 |
---
|
29 |
|
30 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
31 |
should probably proofread and complete it, then remove this comment. -->
|
32 |
|
33 |
-
#
|
34 |
|
35 |
This model is a fine-tuned version of [serge-wilson/whisper-small-wolof](https://huggingface.co/serge-wilson/whisper-small-wolof) on the audiofolder dataset.
|
36 |
It achieves the following results on the evaluation set:
|
37 |
-
- Loss: 0.
|
38 |
-
- Wer:
|
39 |
-
- Cer:
|
40 |
|
41 |
## Model description
|
42 |
|
@@ -60,25 +58,25 @@ The following hyperparameters were used during training:
|
|
60 |
- eval_batch_size: 16
|
61 |
- seed: 42
|
62 |
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-05
|
63 |
-
- lr_scheduler_type:
|
64 |
- lr_scheduler_warmup_steps: 50
|
65 |
- training_steps: 1990
|
66 |
- mixed_precision_training: Native AMP
|
67 |
|
68 |
### Training results
|
69 |
|
70 |
-
| Training Loss | Epoch | Step | Validation Loss | Wer
|
71 |
-
|
72 |
-
| 1.
|
73 |
-
| 0.
|
74 |
-
| 0.
|
75 |
-
| 0.
|
76 |
-
| 0.
|
77 |
-
| 0.
|
78 |
-
| 0.
|
79 |
-
| 0.
|
80 |
-
| 0.
|
81 |
-
| 0.
|
82 |
|
83 |
|
84 |
### Framework versions
|
|
|
1 |
---
|
|
|
|
|
2 |
license: apache-2.0
|
3 |
base_model: serge-wilson/whisper-small-wolof
|
4 |
tags:
|
|
|
8 |
metrics:
|
9 |
- wer
|
10 |
model-index:
|
11 |
+
- name: ASR
|
12 |
results:
|
13 |
- task:
|
14 |
name: Automatic Speech Recognition
|
|
|
22 |
metrics:
|
23 |
- name: Wer
|
24 |
type: wer
|
25 |
+
value: 36.047170881052274
|
26 |
---
|
27 |
|
28 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
29 |
should probably proofread and complete it, then remove this comment. -->
|
30 |
|
31 |
+
# ASR
|
32 |
|
33 |
This model is a fine-tuned version of [serge-wilson/whisper-small-wolof](https://huggingface.co/serge-wilson/whisper-small-wolof) on the audiofolder dataset.
|
34 |
It achieves the following results on the evaluation set:
|
35 |
+
- Loss: 0.3569
|
36 |
+
- Wer: 36.0472
|
37 |
+
- Cer: 22.5967
|
38 |
|
39 |
## Model description
|
40 |
|
|
|
58 |
- eval_batch_size: 16
|
59 |
- seed: 42
|
60 |
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-05
|
61 |
+
- lr_scheduler_type: linear
|
62 |
- lr_scheduler_warmup_steps: 50
|
63 |
- training_steps: 1990
|
64 |
- mixed_precision_training: Native AMP
|
65 |
|
66 |
### Training results
|
67 |
|
68 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
69 |
+
|:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|
|
70 |
+
| 1.2672 | 1.0 | 208 | 1.2009 | 85.3838 | 63.2065 |
|
71 |
+
| 0.875 | 2.0 | 416 | 0.8801 | 95.6117 | 69.2841 |
|
72 |
+
| 0.5964 | 3.0 | 624 | 0.6979 | 88.4681 | 63.1476 |
|
73 |
+
| 0.3953 | 4.0 | 832 | 0.6112 | 69.2255 | 57.6000 |
|
74 |
+
| 0.2465 | 5.0 | 1040 | 0.5015 | 55.4825 | 44.1314 |
|
75 |
+
| 0.161 | 6.0 | 1248 | 0.4401 | 53.7476 | 36.3715 |
|
76 |
+
| 0.0903 | 7.0 | 1456 | 0.4081 | 47.1822 | 31.0320 |
|
77 |
+
| 0.0553 | 8.0 | 1664 | 0.3751 | 44.7783 | 29.2044 |
|
78 |
+
| 0.024 | 9.0 | 1872 | 0.3604 | 38.7686 | 25.2606 |
|
79 |
+
| 0.011 | 9.57 | 1990 | 0.3569 | 36.0472 | 22.5967 |
|
80 |
|
81 |
|
82 |
### Framework versions
|