Commit
•
337b315
1
Parent(s):
ed79b14
End of training
Browse files
README.md
CHANGED
|
@@ -3,6 +3,8 @@ license: apache-2.0
|
|
| 3 |
base_model: facebook/wav2vec2-base
|
| 4 |
tags:
|
| 5 |
- generated_from_trainer
|
|
|
|
|
|
|
| 6 |
model-index:
|
| 7 |
- name: my_awesome_asr_mind_model
|
| 8 |
results: []
|
|
@@ -14,6 +16,9 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 14 |
# my_awesome_asr_mind_model
|
| 15 |
|
| 16 |
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the None dataset.
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
## Model description
|
| 19 |
|
|
@@ -36,13 +41,38 @@ The following hyperparameters were used during training:
|
|
| 36 |
- train_batch_size: 4
|
| 37 |
- eval_batch_size: 8
|
| 38 |
- seed: 42
|
|
|
|
|
|
|
| 39 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 40 |
- lr_scheduler_type: linear
|
| 41 |
-
-
|
|
|
|
| 42 |
- mixed_precision_training: Native AMP
|
| 43 |
|
| 44 |
### Training results
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
|
| 48 |
### Framework versions
|
|
|
|
| 3 |
base_model: facebook/wav2vec2-base
|
| 4 |
tags:
|
| 5 |
- generated_from_trainer
|
| 6 |
+
metrics:
|
| 7 |
+
- wer
|
| 8 |
model-index:
|
| 9 |
- name: my_awesome_asr_mind_model
|
| 10 |
results: []
|
|
|
|
| 16 |
# my_awesome_asr_mind_model
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 1.8925
|
| 21 |
+
- Wer: 0.4558
|
| 22 |
|
| 23 |
## Model description
|
| 24 |
|
|
|
|
| 41 |
- train_batch_size: 4
|
| 42 |
- eval_batch_size: 8
|
| 43 |
- seed: 42
|
| 44 |
+
- gradient_accumulation_steps: 2
|
| 45 |
+
- total_train_batch_size: 8
|
| 46 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 47 |
- lr_scheduler_type: linear
|
| 48 |
+
- lr_scheduler_warmup_steps: 500
|
| 49 |
+
- training_steps: 2000
|
| 50 |
- mixed_precision_training: Native AMP
|
| 51 |
|
| 52 |
### Training results
|
| 53 |
|
| 54 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 55 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 56 |
+
| 4.5119 | 1.77 | 100 | 4.1083 | 1.0 |
|
| 57 |
+
| 3.287 | 3.54 | 200 | 3.2437 | 1.0 |
|
| 58 |
+
| 3.1513 | 5.31 | 300 | 3.1230 | 1.0 |
|
| 59 |
+
| 3.0487 | 7.08 | 400 | 3.0786 | 1.0 |
|
| 60 |
+
| 3.0241 | 8.85 | 500 | 3.0934 | 1.0 |
|
| 61 |
+
| 2.9968 | 10.62 | 600 | 2.9948 | 1.0 |
|
| 62 |
+
| 2.9601 | 12.39 | 700 | 2.9549 | 1.0 |
|
| 63 |
+
| 2.9061 | 14.16 | 800 | 2.8990 | 1.0 |
|
| 64 |
+
| 2.3543 | 15.93 | 900 | 2.2582 | 0.9272 |
|
| 65 |
+
| 1.3794 | 17.7 | 1000 | 1.7532 | 0.8179 |
|
| 66 |
+
| 0.8947 | 19.47 | 1100 | 1.2148 | 0.6710 |
|
| 67 |
+
| 0.5989 | 21.24 | 1200 | 1.3229 | 0.5579 |
|
| 68 |
+
| 0.5861 | 23.01 | 1300 | 1.4233 | 0.5267 |
|
| 69 |
+
| 0.4311 | 24.78 | 1400 | 1.5458 | 0.5104 |
|
| 70 |
+
| 0.3286 | 26.55 | 1500 | 1.6509 | 0.5039 |
|
| 71 |
+
| 0.2765 | 28.32 | 1600 | 1.6818 | 0.4948 |
|
| 72 |
+
| 0.2541 | 30.09 | 1700 | 1.7650 | 0.4629 |
|
| 73 |
+
| 0.2151 | 31.86 | 1800 | 1.7185 | 0.4460 |
|
| 74 |
+
| 0.1959 | 33.63 | 1900 | 1.9164 | 0.4577 |
|
| 75 |
+
| 0.1909 | 35.4 | 2000 | 1.8925 | 0.4558 |
|
| 76 |
|
| 77 |
|
| 78 |
### Framework versions
|