End of training
Browse files
README.md
CHANGED
|
@@ -8,9 +8,22 @@ tags:
|
|
| 8 |
- generated_from_trainer
|
| 9 |
datasets:
|
| 10 |
- lord-reso/inbrowser-proctor-dataset
|
|
|
|
|
|
|
| 11 |
model-index:
|
| 12 |
- name: Whisper-Small-Inbrowser-Proctor
|
| 13 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
---
|
| 15 |
|
| 16 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -20,13 +33,8 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 20 |
|
| 21 |
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the Inbrowser Procotor Dataset dataset.
|
| 22 |
It achieves the following results on the evaluation set:
|
| 23 |
-
-
|
| 24 |
-
-
|
| 25 |
-
- eval_runtime: 53.0859
|
| 26 |
-
- eval_samples_per_second: 1.319
|
| 27 |
-
- eval_steps_per_second: 0.17
|
| 28 |
-
- epoch: 8.9286
|
| 29 |
-
- step: 250
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
@@ -46,15 +54,31 @@ More information needed
|
|
| 46 |
|
| 47 |
The following hyperparameters were used during training:
|
| 48 |
- learning_rate: 5e-06
|
| 49 |
-
- train_batch_size:
|
| 50 |
- eval_batch_size: 8
|
| 51 |
- seed: 42
|
| 52 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 53 |
- lr_scheduler_type: linear
|
| 54 |
-
- lr_scheduler_warmup_steps:
|
| 55 |
-
- training_steps:
|
| 56 |
- mixed_precision_training: Native AMP
|
| 57 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
### Framework versions
|
| 59 |
|
| 60 |
- Transformers 4.44.2
|
|
|
|
| 8 |
- generated_from_trainer
|
| 9 |
datasets:
|
| 10 |
- lord-reso/inbrowser-proctor-dataset
|
| 11 |
+
metrics:
|
| 12 |
+
- wer
|
| 13 |
model-index:
|
| 14 |
- name: Whisper-Small-Inbrowser-Proctor
|
| 15 |
+
results:
|
| 16 |
+
- task:
|
| 17 |
+
name: Automatic Speech Recognition
|
| 18 |
+
type: automatic-speech-recognition
|
| 19 |
+
dataset:
|
| 20 |
+
name: Inbrowser Procotor Dataset
|
| 21 |
+
type: lord-reso/inbrowser-proctor-dataset
|
| 22 |
+
args: 'config: en, split: test'
|
| 23 |
+
metrics:
|
| 24 |
+
- name: Wer
|
| 25 |
+
type: wer
|
| 26 |
+
value: 17.075501752150366
|
| 27 |
---
|
| 28 |
|
| 29 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 33 |
|
| 34 |
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the Inbrowser Procotor Dataset dataset.
|
| 35 |
It achieves the following results on the evaluation set:
|
| 36 |
+
- Loss: 0.3099
|
| 37 |
+
- Wer: 17.0755
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
|
| 39 |
## Model description
|
| 40 |
|
|
|
|
| 54 |
|
| 55 |
The following hyperparameters were used during training:
|
| 56 |
- learning_rate: 5e-06
|
| 57 |
+
- train_batch_size: 8
|
| 58 |
- eval_batch_size: 8
|
| 59 |
- seed: 42
|
| 60 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 61 |
- lr_scheduler_type: linear
|
| 62 |
+
- lr_scheduler_warmup_steps: 25
|
| 63 |
+
- training_steps: 250
|
| 64 |
- mixed_precision_training: Native AMP
|
| 65 |
|
| 66 |
+
### Training results
|
| 67 |
+
|
| 68 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 69 |
+
|:-------------:|:------:|:----:|:---------------:|:-------:|
|
| 70 |
+
| 0.3461 | 0.4545 | 25 | 0.4545 | 26.0433 |
|
| 71 |
+
| 0.1902 | 0.9091 | 50 | 0.3309 | 17.4419 |
|
| 72 |
+
| 0.1184 | 1.3636 | 75 | 0.3120 | 14.6543 |
|
| 73 |
+
| 0.0944 | 1.8182 | 100 | 0.3066 | 16.7251 |
|
| 74 |
+
| 0.0632 | 2.2727 | 125 | 0.3046 | 14.8455 |
|
| 75 |
+
| 0.0688 | 2.7273 | 150 | 0.3060 | 14.8933 |
|
| 76 |
+
| 0.0479 | 3.1818 | 175 | 0.3063 | 17.1074 |
|
| 77 |
+
| 0.0515 | 3.6364 | 200 | 0.3081 | 15.4986 |
|
| 78 |
+
| 0.0296 | 4.0909 | 225 | 0.3096 | 17.2507 |
|
| 79 |
+
| 0.0348 | 4.5455 | 250 | 0.3099 | 17.0755 |
|
| 80 |
+
|
| 81 |
+
|
| 82 |
### Framework versions
|
| 83 |
|
| 84 |
- Transformers 4.44.2
|