Commit
•
f02cdd7
1
Parent(s):
b0a49d1
update model card README.md
Browse files
README.md
CHANGED
|
@@ -1,13 +1,7 @@
|
|
| 1 |
---
|
| 2 |
-
language:
|
| 3 |
-
- ur
|
| 4 |
license: apache-2.0
|
| 5 |
tags:
|
| 6 |
-
- automatic-speech-recognition
|
| 7 |
-
- mozilla-foundation/common_voice_8_0
|
| 8 |
- generated_from_trainer
|
| 9 |
-
- ur
|
| 10 |
-
- robust-speech-event
|
| 11 |
datasets:
|
| 12 |
- common_voice
|
| 13 |
model-index:
|
|
@@ -20,10 +14,10 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 20 |
|
| 21 |
#
|
| 22 |
|
| 23 |
-
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the
|
| 24 |
It achieves the following results on the evaluation set:
|
| 25 |
-
- Loss:
|
| 26 |
-
- Wer: 0.
|
| 27 |
|
| 28 |
## Model description
|
| 29 |
|
|
@@ -46,11 +40,11 @@ The following hyperparameters were used during training:
|
|
| 46 |
- train_batch_size: 8
|
| 47 |
- eval_batch_size: 8
|
| 48 |
- seed: 42
|
| 49 |
-
- gradient_accumulation_steps:
|
| 50 |
-
- total_train_batch_size:
|
| 51 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 52 |
- lr_scheduler_type: linear
|
| 53 |
-
- lr_scheduler_warmup_steps:
|
| 54 |
- num_epochs: 50.0
|
| 55 |
- mixed_precision_training: Native AMP
|
| 56 |
|
|
@@ -58,13 +52,36 @@ The following hyperparameters were used during training:
|
|
| 58 |
|
| 59 |
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 60 |
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 61 |
-
|
|
| 62 |
-
| 3.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
|
| 65 |
### Framework versions
|
| 66 |
|
| 67 |
- Transformers 4.17.0.dev0
|
| 68 |
- Pytorch 1.10.2+cu102
|
| 69 |
-
- Datasets 1.18.
|
| 70 |
- Tokenizers 0.11.0
|
|
|
|
| 1 |
---
|
|
|
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
tags:
|
|
|
|
|
|
|
| 4 |
- generated_from_trainer
|
|
|
|
|
|
|
| 5 |
datasets:
|
| 6 |
- common_voice
|
| 7 |
model-index:
|
|
|
|
| 14 |
|
| 15 |
#
|
| 16 |
|
| 17 |
+
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
|
| 18 |
It achieves the following results on the evaluation set:
|
| 19 |
+
- Loss: 0.9604
|
| 20 |
+
- Wer: 0.6542
|
| 21 |
|
| 22 |
## Model description
|
| 23 |
|
|
|
|
| 40 |
- train_batch_size: 8
|
| 41 |
- eval_batch_size: 8
|
| 42 |
- seed: 42
|
| 43 |
+
- gradient_accumulation_steps: 2
|
| 44 |
+
- total_train_batch_size: 16
|
| 45 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 46 |
- lr_scheduler_type: linear
|
| 47 |
+
- lr_scheduler_warmup_steps: 50
|
| 48 |
- num_epochs: 50.0
|
| 49 |
- mixed_precision_training: Native AMP
|
| 50 |
|
|
|
|
| 52 |
|
| 53 |
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 54 |
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 55 |
+
| 9.5036 | 1.96 | 100 | 4.0538 | 1.0 |
|
| 56 |
+
| 3.3669 | 3.92 | 200 | 3.2041 | 1.0 |
|
| 57 |
+
| 3.1499 | 5.88 | 300 | 3.1220 | 1.0 |
|
| 58 |
+
| 3.0271 | 7.84 | 400 | 2.9935 | 0.9970 |
|
| 59 |
+
| 2.9565 | 9.8 | 500 | 2.9357 | 0.9993 |
|
| 60 |
+
| 2.9184 | 11.76 | 600 | 2.9165 | 0.9963 |
|
| 61 |
+
| 2.8832 | 13.73 | 700 | 2.8762 | 0.9911 |
|
| 62 |
+
| 2.8407 | 15.69 | 800 | 2.8102 | 0.9970 |
|
| 63 |
+
| 2.7007 | 17.65 | 900 | 2.4364 | 0.9963 |
|
| 64 |
+
| 2.4206 | 19.61 | 1000 | 1.9852 | 0.9421 |
|
| 65 |
+
| 2.0699 | 21.57 | 1100 | 1.4849 | 0.8343 |
|
| 66 |
+
| 1.8311 | 23.53 | 1200 | 1.3084 | 0.7801 |
|
| 67 |
+
| 1.7127 | 25.49 | 1300 | 1.2040 | 0.7446 |
|
| 68 |
+
| 1.6239 | 27.45 | 1400 | 1.1359 | 0.7280 |
|
| 69 |
+
| 1.5654 | 29.41 | 1500 | 1.0688 | 0.7159 |
|
| 70 |
+
| 1.4965 | 31.37 | 1600 | 1.0520 | 0.6985 |
|
| 71 |
+
| 1.445 | 33.33 | 1700 | 1.0314 | 0.6878 |
|
| 72 |
+
| 1.4095 | 35.29 | 1800 | 1.0063 | 0.6712 |
|
| 73 |
+
| 1.3853 | 37.25 | 1900 | 0.9848 | 0.6701 |
|
| 74 |
+
| 1.3558 | 39.22 | 2000 | 0.9738 | 0.6731 |
|
| 75 |
+
| 1.3415 | 41.18 | 2100 | 0.9656 | 0.6646 |
|
| 76 |
+
| 1.3102 | 43.14 | 2200 | 0.9632 | 0.6557 |
|
| 77 |
+
| 1.309 | 45.1 | 2300 | 0.9496 | 0.6557 |
|
| 78 |
+
| 1.2993 | 47.06 | 2400 | 0.9609 | 0.6550 |
|
| 79 |
+
| 1.2695 | 49.02 | 2500 | 0.9604 | 0.6542 |
|
| 80 |
|
| 81 |
|
| 82 |
### Framework versions
|
| 83 |
|
| 84 |
- Transformers 4.17.0.dev0
|
| 85 |
- Pytorch 1.10.2+cu102
|
| 86 |
+
- Datasets 1.18.3
|
| 87 |
- Tokenizers 0.11.0
|