PereLluis13
commited on
Commit
•
fa16ca3
1
Parent(s):
e2a01cb
update model card README.md
Browse files
README.md
CHANGED
@@ -8,70 +8,9 @@ tags:
|
|
8 |
- collectivat/tv3_parla
|
9 |
- projecte-aina/parlament_parla
|
10 |
- generated_from_trainer
|
11 |
-
- robust-speech-event
|
12 |
-
datasets:
|
13 |
-
- mozilla-foundation/common_voice_8_0
|
14 |
-
- collectivat/tv3_parla
|
15 |
-
- projecte-aina/parlament_parla
|
16 |
model-index:
|
17 |
- name: wav2vec2-xls-r-300m-ca
|
18 |
-
results:
|
19 |
-
- task:
|
20 |
-
name: Speech Recognition
|
21 |
-
type: automatic-speech-recognition
|
22 |
-
dataset:
|
23 |
-
name: mozilla-foundation/common_voice_8_0 ca
|
24 |
-
type: mozilla-foundation/common_voice_8_0
|
25 |
-
args: ca
|
26 |
-
metrics:
|
27 |
-
- name: Test WER
|
28 |
-
type: wer
|
29 |
-
value: 0.1522665117742443
|
30 |
-
- name: Test CER
|
31 |
-
type: cer
|
32 |
-
value: 0.04078709154868726
|
33 |
-
- task:
|
34 |
-
name: Speech Recognition
|
35 |
-
type: automatic-speech-recognition
|
36 |
-
dataset:
|
37 |
-
name: projecte-aina/parlament_parla ca
|
38 |
-
type: projecte-aina/parlament_parla
|
39 |
-
args: clean
|
40 |
-
metrics:
|
41 |
-
- name: Test WER
|
42 |
-
type: wer
|
43 |
-
value: 0.06541946111307212
|
44 |
-
- name: Test CER
|
45 |
-
type: cer
|
46 |
-
value: 0.02205785796827398
|
47 |
-
- task:
|
48 |
-
name: Speech Recognition
|
49 |
-
type: automatic-speech-recognition
|
50 |
-
dataset:
|
51 |
-
name: collectivat/tv3_parla ca
|
52 |
-
type: collectivat/tv3_parla
|
53 |
-
args: ca
|
54 |
-
metrics:
|
55 |
-
- name: Test WER
|
56 |
-
type: wer
|
57 |
-
value: 0.24485121453593564
|
58 |
-
- name: Test CER
|
59 |
-
type: cer
|
60 |
-
value: 0.10753510718204506
|
61 |
-
- task:
|
62 |
-
name: Speech Recognition
|
63 |
-
type: automatic-speech-recognition
|
64 |
-
dataset:
|
65 |
-
name: Robust Speech Event - Catalan Dev Data
|
66 |
-
type: speech-recognition-community-v2/dev_data
|
67 |
-
args: ca
|
68 |
-
metrics:
|
69 |
-
- name: Test WER
|
70 |
-
type: wer
|
71 |
-
value: 0.3325532856871798
|
72 |
-
- name: Test CER
|
73 |
-
type: cer
|
74 |
-
value: 0.15916561314791403
|
75 |
---
|
76 |
|
77 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
@@ -81,8 +20,8 @@ should probably proofread and complete it, then remove this comment. -->
|
|
81 |
|
82 |
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - CA dataset.
|
83 |
It achieves the following results on the evaluation set:
|
84 |
-
- Loss: 0.
|
85 |
-
- Wer: 0.
|
86 |
|
87 |
## Model description
|
88 |
|
@@ -110,7 +49,7 @@ The following hyperparameters were used during training:
|
|
110 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
111 |
- lr_scheduler_type: linear
|
112 |
- lr_scheduler_warmup_steps: 2000
|
113 |
-
- num_epochs:
|
114 |
- mixed_precision_training: Native AMP
|
115 |
|
116 |
### Training results
|
@@ -162,11 +101,28 @@ The following hyperparameters were used during training:
|
|
162 |
| 1.0805 | 11.45 | 21500 | 0.2561 | 0.1524 |
|
163 |
| 1.0722 | 11.72 | 22000 | 0.2540 | 0.1566 |
|
164 |
| 1.0763 | 11.99 | 22500 | 0.2549 | 0.1572 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
165 |
|
166 |
|
167 |
### Framework versions
|
168 |
|
169 |
- Transformers 4.16.0.dev0
|
170 |
- Pytorch 1.10.1+cu102
|
171 |
-
- Datasets 1.18.
|
172 |
- Tokenizers 0.11.0
|
|
|
8 |
- collectivat/tv3_parla
|
9 |
- projecte-aina/parlament_parla
|
10 |
- generated_from_trainer
|
|
|
|
|
|
|
|
|
|
|
11 |
model-index:
|
12 |
- name: wav2vec2-xls-r-300m-ca
|
13 |
+
results: []
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
---
|
15 |
|
16 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
20 |
|
21 |
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - CA dataset.
|
22 |
It achieves the following results on the evaluation set:
|
23 |
+
- Loss: 0.2472
|
24 |
+
- Wer: 0.1499
|
25 |
|
26 |
## Model description
|
27 |
|
|
|
49 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
50 |
- lr_scheduler_type: linear
|
51 |
- lr_scheduler_warmup_steps: 2000
|
52 |
+
- num_epochs: 18.0
|
53 |
- mixed_precision_training: Native AMP
|
54 |
|
55 |
### Training results
|
|
|
101 |
| 1.0805 | 11.45 | 21500 | 0.2561 | 0.1524 |
|
102 |
| 1.0722 | 11.72 | 22000 | 0.2540 | 0.1566 |
|
103 |
| 1.0763 | 11.99 | 22500 | 0.2549 | 0.1572 |
|
104 |
+
| 1.0835 | 12.25 | 23000 | 0.2586 | 0.1521 |
|
105 |
+
| 1.0883 | 12.52 | 23500 | 0.2583 | 0.1519 |
|
106 |
+
| 1.0888 | 12.79 | 24000 | 0.2551 | 0.1582 |
|
107 |
+
| 1.0933 | 13.05 | 24500 | 0.2628 | 0.1537 |
|
108 |
+
| 1.0799 | 13.32 | 25000 | 0.2600 | 0.1508 |
|
109 |
+
| 1.0804 | 13.59 | 25500 | 0.2620 | 0.1475 |
|
110 |
+
| 1.0814 | 13.85 | 26000 | 0.2537 | 0.1517 |
|
111 |
+
| 1.0693 | 14.12 | 26500 | 0.2560 | 0.1542 |
|
112 |
+
| 1.0724 | 14.38 | 27000 | 0.2540 | 0.1574 |
|
113 |
+
| 1.0704 | 14.65 | 27500 | 0.2548 | 0.1626 |
|
114 |
+
| 1.0729 | 14.92 | 28000 | 0.2548 | 0.1601 |
|
115 |
+
| 1.0724 | 15.18 | 28500 | 0.2511 | 0.1512 |
|
116 |
+
| 1.0655 | 15.45 | 29000 | 0.2498 | 0.1490 |
|
117 |
+
| 1.0608 | 15.98 | 30000 | 0.2487 | 0.1481 |
|
118 |
+
| 1.0541 | 16.52 | 31000 | 0.2468 | 0.1504 |
|
119 |
+
| 1.0584 | 17.05 | 32000 | 0.2467 | 0.1493 |
|
120 |
+
| 1.0507 | 17.58 | 33000 | 0.2481 | 0.1517 |
|
121 |
|
122 |
|
123 |
### Framework versions
|
124 |
|
125 |
- Transformers 4.16.0.dev0
|
126 |
- Pytorch 1.10.1+cu102
|
127 |
+
- Datasets 1.18.3
|
128 |
- Tokenizers 0.11.0
|