Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: facebook/wav2vec2-base
|
| 4 |
+
tags:
|
| 5 |
+
- generated_from_trainer
|
| 6 |
+
metrics:
|
| 7 |
+
- accuracy
|
| 8 |
+
- f1
|
| 9 |
+
model-index:
|
| 10 |
+
- name: wav2vec2_capstone
|
| 11 |
+
results: []
|
| 12 |
+
datasets:
|
| 13 |
+
- mozilla-foundation/common_voice_16_1
|
| 14 |
+
language:
|
| 15 |
+
- en
|
| 16 |
+
- ca
|
| 17 |
+
- rw
|
| 18 |
+
- be
|
| 19 |
+
- eo
|
| 20 |
+
- de
|
| 21 |
+
- fr
|
| 22 |
+
- ka
|
| 23 |
+
- es
|
| 24 |
+
- lg
|
| 25 |
+
- sw
|
| 26 |
+
- fa
|
| 27 |
+
- it
|
| 28 |
+
- mh
|
| 29 |
+
- zh
|
| 30 |
+
- ba
|
| 31 |
+
- ta
|
| 32 |
+
- ru
|
| 33 |
+
- eu
|
| 34 |
+
- th
|
| 35 |
+
- pt
|
| 36 |
+
- pl
|
| 37 |
+
- ja
|
| 38 |
+
pipeline_tag: audio-classification
|
| 39 |
+
---
|
| 40 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 41 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 42 |
+
|
| 43 |
+
# WAV2VEC2_CAPSTONE_MODEL2
|
| 44 |
+
|
| 45 |
+
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on a subset of the common_voice_16_1 dataset.
|
| 46 |
+
It achieves the following results on the evaluation set:
|
| 47 |
+
- Loss: 0.3952
|
| 48 |
+
- Accuracy: 0.9098
|
| 49 |
+
- F1 score: 0.9097
|
| 50 |
+
## Model description
|
| 51 |
+
More information needed
|
| 52 |
+
## Intended uses & limitations
|
| 53 |
+
More information needed
|
| 54 |
+
## Training and evaluation data
|
| 55 |
+
More information needed
|
| 56 |
+
## Training procedure
|
| 57 |
+
### Training hyperparameters
|
| 58 |
+
The following hyperparameters were used during training:
|
| 59 |
+
- learning_rate: 0.0003
|
| 60 |
+
- train_batch_size: 7
|
| 61 |
+
- eval_batch_size: 7
|
| 62 |
+
- seed: 42
|
| 63 |
+
- gradient_accumulation_steps: 11
|
| 64 |
+
- total_train_batch_size: 77
|
| 65 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 66 |
+
- lr_scheduler_type: linear
|
| 67 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 68 |
+
- num_epochs: 7
|
| 69 |
+
- mixed_precision_training: Native AMP
|
| 70 |
+
### Training results
|
| 71 |
+
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 score |
|
| 72 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------:|:--------:|
|
| 73 |
+
| 1.5222 | 1.0 | 513 | 1.4214 | 0.5477 | 0.5374 |
|
| 74 |
+
| 0.9554 | 2.0 | 1027 | 1.0182 | 0.6892 | 0.6826 |
|
| 75 |
+
| 0.6128 | 3.0 | 1541 | 0.6117 | 0.8205 | 0.8203 |
|
| 76 |
+
| 0.4882 | 4.0 | 2055 | 0.5092 | 0.8474 | 0.8478 |
|
| 77 |
+
| 0.2305 | 5.0 | 2569 | 0.5244 | 0.8643 | 0.8639 |
|
| 78 |
+
| 0.1219 | 6.0 | 3082 | 0.4169 | 0.8966 | 0.8966 |
|
| 79 |
+
| 0.0672 | 6.99 | 3591 | 0.3952 | 0.9098 | 0.9097 |
|
| 80 |
+
### Framework versions
|
| 81 |
+
- Transformers 4.36.2
|
| 82 |
+
- Pytorch 2.1.2+cu121
|
| 83 |
+
- Datasets 2.16.0
|
| 84 |
+
- Tokenizers 0.15.0
|