Commit
·
7c51b83
1
Parent(s):
51409ba
update model card README.md
Browse files
README.md
CHANGED
|
@@ -7,9 +7,24 @@ tags:
|
|
| 7 |
- generated_from_trainer
|
| 8 |
datasets:
|
| 9 |
- addy88/sanskrit-asr-84
|
|
|
|
|
|
|
| 10 |
model-index:
|
| 11 |
- name: Whisper Small Sanskasr - bvkb
|
| 12 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
---
|
| 14 |
|
| 15 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -18,6 +33,9 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 18 |
# Whisper Small Sanskasr - bvkb
|
| 19 |
|
| 20 |
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the addy88/sanskrit-asr-84 dataset.
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
## Model description
|
| 23 |
|
|
@@ -37,23 +55,33 @@ More information needed
|
|
| 37 |
|
| 38 |
The following hyperparameters were used during training:
|
| 39 |
- learning_rate: 1e-05
|
| 40 |
-
- train_batch_size:
|
| 41 |
- eval_batch_size: 10
|
| 42 |
- seed: 42
|
| 43 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 44 |
- lr_scheduler_type: linear
|
| 45 |
-
-
|
|
|
|
| 46 |
|
| 47 |
### Training results
|
| 48 |
|
| 49 |
-
| Training Loss | Epoch | Step | Validation Loss | Wer
|
| 50 |
-
|
| 51 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
|
| 53 |
|
| 54 |
### Framework versions
|
| 55 |
|
| 56 |
- Transformers 4.30.2
|
| 57 |
-
- Pytorch 2.0.
|
| 58 |
- Datasets 2.13.0
|
| 59 |
- Tokenizers 0.13.3
|
|
|
|
| 7 |
- generated_from_trainer
|
| 8 |
datasets:
|
| 9 |
- addy88/sanskrit-asr-84
|
| 10 |
+
metrics:
|
| 11 |
+
- wer
|
| 12 |
model-index:
|
| 13 |
- name: Whisper Small Sanskasr - bvkb
|
| 14 |
+
results:
|
| 15 |
+
- task:
|
| 16 |
+
name: Automatic Speech Recognition
|
| 17 |
+
type: automatic-speech-recognition
|
| 18 |
+
dataset:
|
| 19 |
+
name: addy88/sanskrit-asr-84
|
| 20 |
+
type: addy88/sanskrit-asr-84
|
| 21 |
+
config: null
|
| 22 |
+
split: None
|
| 23 |
+
args: 'split: test'
|
| 24 |
+
metrics:
|
| 25 |
+
- name: Wer
|
| 26 |
+
type: wer
|
| 27 |
+
value: 87.73735987188287
|
| 28 |
---
|
| 29 |
|
| 30 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 33 |
# Whisper Small Sanskasr - bvkb
|
| 34 |
|
| 35 |
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the addy88/sanskrit-asr-84 dataset.
|
| 36 |
+
It achieves the following results on the evaluation set:
|
| 37 |
+
- Loss: 0.7583
|
| 38 |
+
- Wer: 87.7374
|
| 39 |
|
| 40 |
## Model description
|
| 41 |
|
|
|
|
| 55 |
|
| 56 |
The following hyperparameters were used during training:
|
| 57 |
- learning_rate: 1e-05
|
| 58 |
+
- train_batch_size: 10
|
| 59 |
- eval_batch_size: 10
|
| 60 |
- seed: 42
|
| 61 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 62 |
- lr_scheduler_type: linear
|
| 63 |
+
- lr_scheduler_warmup_steps: 5
|
| 64 |
+
- training_steps: 100
|
| 65 |
|
| 66 |
### Training results
|
| 67 |
|
| 68 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 69 |
+
|:-------------:|:-----:|:----:|:---------------:|:-------:|
|
| 70 |
+
| 1.516 | 0.02 | 10 | 1.1614 | 96.7284 |
|
| 71 |
+
| 1.0285 | 0.04 | 20 | 0.9667 | 99.4967 |
|
| 72 |
+
| 0.9284 | 0.06 | 30 | 0.8885 | 93.6399 |
|
| 73 |
+
| 0.8789 | 0.08 | 40 | 0.8437 | 90.8259 |
|
| 74 |
+
| 0.8579 | 0.1 | 50 | 0.8159 | 89.4532 |
|
| 75 |
+
| 0.8001 | 0.12 | 60 | 0.7957 | 89.4990 |
|
| 76 |
+
| 0.8113 | 0.14 | 70 | 0.7807 | 89.1558 |
|
| 77 |
+
| 0.7542 | 0.16 | 80 | 0.7698 | 87.9890 |
|
| 78 |
+
| 0.7672 | 0.18 | 90 | 0.7610 | 87.7374 |
|
| 79 |
+
| 0.7576 | 0.2 | 100 | 0.7583 | 87.7374 |
|
| 80 |
|
| 81 |
|
| 82 |
### Framework versions
|
| 83 |
|
| 84 |
- Transformers 4.30.2
|
| 85 |
+
- Pytorch 2.0.0
|
| 86 |
- Datasets 2.13.0
|
| 87 |
- Tokenizers 0.13.3
|