Commit ·
a04c2fe
1
Parent(s): 7c41b61
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,11 +1,56 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
| 2 |
tags:
|
| 3 |
-
-
|
|
|
|
| 4 |
datasets:
|
| 5 |
- common_voice
|
|
|
|
|
|
|
|
|
|
| 6 |
model-index:
|
| 7 |
- name: wav2vec2-large-xlsr-53-urdu
|
| 8 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
---
|
| 10 |
|
| 11 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -21,15 +66,11 @@ It achieves the following results on the evaluation set:
|
|
| 21 |
|
| 22 |
## Model description
|
| 23 |
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
## Intended uses & limitations
|
| 27 |
-
|
| 28 |
-
More information needed
|
| 29 |
|
| 30 |
## Training and evaluation data
|
| 31 |
|
| 32 |
-
|
| 33 |
|
| 34 |
## Training procedure
|
| 35 |
|
|
|
|
| 1 |
---
|
| 2 |
+
language:
|
| 3 |
+
- ur
|
| 4 |
+
license: apache-2.0
|
| 5 |
tags:
|
| 6 |
+
- automatic-speech-recognition
|
| 7 |
+
- robust-speech-event
|
| 8 |
datasets:
|
| 9 |
- common_voice
|
| 10 |
+
metrics:
|
| 11 |
+
- wer
|
| 12 |
+
- cer
|
| 13 |
model-index:
|
| 14 |
- name: wav2vec2-large-xlsr-53-urdu
|
| 15 |
+
results:
|
| 16 |
+
- task:
|
| 17 |
+
type: automatic-speech-recognition # Required. Example: automatic-speech-recognition
|
| 18 |
+
name: Urdu Speech Recognition # Optional. Example: Speech Recognition
|
| 19 |
+
dataset:
|
| 20 |
+
type: common_voice # Required. Example: common_voice. Use dataset id from https://hf.co/datasets
|
| 21 |
+
name: Urdu # Required. Example: Common Voice zh-CN
|
| 22 |
+
args: ur # Optional. Example: zh-CN
|
| 23 |
+
metrics:
|
| 24 |
+
- type: wer # Required. Example: wer
|
| 25 |
+
value: 57.7 # Required. Example: 20.90
|
| 26 |
+
name: Test WER # Optional. Example: Test WER
|
| 27 |
+
args:
|
| 28 |
+
- learning_rate: 0.0003
|
| 29 |
+
- train_batch_size: 16
|
| 30 |
+
- eval_batch_size: 8
|
| 31 |
+
- seed: 42
|
| 32 |
+
- gradient_accumulation_steps: 2
|
| 33 |
+
- total_train_batch_size: 32
|
| 34 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 35 |
+
- lr_scheduler_type: linear
|
| 36 |
+
- lr_scheduler_warmup_steps: 200
|
| 37 |
+
- num_epochs: 50
|
| 38 |
+
- mixed_precision_training: Native AMP # Optional. Example for BLEU: max_order
|
| 39 |
+
- type: cer # Required. Example: wer
|
| 40 |
+
value: 33.8 # Required. Example: 20.90
|
| 41 |
+
name: Test CER # Optional. Example: Test WER
|
| 42 |
+
args:
|
| 43 |
+
- learning_rate: 0.0003
|
| 44 |
+
- train_batch_size: 16
|
| 45 |
+
- eval_batch_size: 8
|
| 46 |
+
- seed: 42
|
| 47 |
+
- gradient_accumulation_steps: 2
|
| 48 |
+
- total_train_batch_size: 32
|
| 49 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 50 |
+
- lr_scheduler_type: linear
|
| 51 |
+
- lr_scheduler_warmup_steps: 200
|
| 52 |
+
- num_epochs: 50
|
| 53 |
+
- mixed_precision_training: Native AMP # Optional. Example for BLEU: max_order
|
| 54 |
---
|
| 55 |
|
| 56 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 66 |
|
| 67 |
## Model description
|
| 68 |
|
| 69 |
+
The training and valid dataset is 0.58 hours. It was hard to train any model on lower number of so I decided to take Urdu checkpoint and finetune the XLSR model.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
|
| 71 |
## Training and evaluation data
|
| 72 |
|
| 73 |
+
Trained on Harveenchadha/vakyansh-wav2vec2-urdu-urm-60 due to lesser number of samples. Persian and Urdu are quite similar.
|
| 74 |
|
| 75 |
## Training procedure
|
| 76 |
|