Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,44 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
pipeline_tag: feature-extraction
|
| 4 |
+
tags:
|
| 5 |
+
- wav2vec2
|
| 6 |
+
- conformer
|
| 7 |
+
- speech
|
| 8 |
+
datasets:
|
| 9 |
+
- openslr/librispeech_asr
|
| 10 |
+
language:
|
| 11 |
+
- en
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# wav2vec2-conformer-base
|
| 15 |
+
|
| 16 |
+
## Dataset
|
| 17 |
+
- [librispeech](https://www.openslr.org/12)
|
| 18 |
+
|
| 19 |
+
## Framework
|
| 20 |
+
- [faiseq](https://github.com/facebookresearch/fairseq)
|
| 21 |
+
|
| 22 |
+
## Model Info
|
| 23 |
+
```
|
| 24 |
+
model:
|
| 25 |
+
_name: wav2vec2
|
| 26 |
+
quantize_targets: true
|
| 27 |
+
final_dim: 256
|
| 28 |
+
encoder_layerdrop: 0.05
|
| 29 |
+
dropout_input: 0.1
|
| 30 |
+
dropout_features: 0.1
|
| 31 |
+
feature_grad_mult: 0.1
|
| 32 |
+
|
| 33 |
+
encoder_layers: 12
|
| 34 |
+
encoder_embed_dim: 768
|
| 35 |
+
encoder_ffn_embed_dim: 3072
|
| 36 |
+
encoder_attention_heads: 12
|
| 37 |
+
|
| 38 |
+
layer_type: conformer
|
| 39 |
+
attn_type: espnet
|
| 40 |
+
pos_enc_type: rel_pos
|
| 41 |
+
```
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
This model is use in [emotion-conformer](https://github.com/poyu39/emotion-conformer)
|