update weights
Browse files- .ipynb_checkpoints/config-checkpoint.json +45 -0
- README.md +54 -0
.ipynb_checkpoints/config-checkpoint.json
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"activation_dropout": 0.15,
|
3 |
+
"activation_function": "relu",
|
4 |
+
"architectures": [
|
5 |
+
"Speech2TextTransformerForConditionalGeneration"
|
6 |
+
],
|
7 |
+
"attention_dropout": 0.15,
|
8 |
+
"bos_token_id": 0,
|
9 |
+
"classifier_dropout": 0.0,
|
10 |
+
"conv_channels": 1024,
|
11 |
+
"conv_kernel_sizes": [
|
12 |
+
5,
|
13 |
+
5
|
14 |
+
],
|
15 |
+
"d_model": 512,
|
16 |
+
"decoder_attention_heads": 8,
|
17 |
+
"decoder_ffn_dim": 2048,
|
18 |
+
"decoder_layerdrop": 0.0,
|
19 |
+
"decoder_layers": 6,
|
20 |
+
"decoder_start_token_id": 2,
|
21 |
+
"dropout": 0.15,
|
22 |
+
"early_stopping": true,
|
23 |
+
"encoder_attention_heads": 8,
|
24 |
+
"encoder_ffn_dim": 2048,
|
25 |
+
"encoder_layerdrop": 0.0,
|
26 |
+
"encoder_layers": 12,
|
27 |
+
"eos_token_id": 2,
|
28 |
+
"gradient_checkpointing": false,
|
29 |
+
"init_std": 0.02,
|
30 |
+
"input_channels": 1,
|
31 |
+
"input_feat_per_channel": 80,
|
32 |
+
"is_encoder_decoder": true,
|
33 |
+
"max_length": 200,
|
34 |
+
"max_source_positions": 6000,
|
35 |
+
"max_target_positions": 1024,
|
36 |
+
"model_type": "speech_to_text_transformer",
|
37 |
+
"num_beams": 5,
|
38 |
+
"num_conv_layers": 2,
|
39 |
+
"num_hidden_layers": 12,
|
40 |
+
"pad_token_id": 1,
|
41 |
+
"scale_embedding": true,
|
42 |
+
"transformers_version": "4.4.0.dev0",
|
43 |
+
"use_cache": true,
|
44 |
+
"vocab_size": 10000
|
45 |
+
}
|
README.md
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: en
|
3 |
+
datasets:
|
4 |
+
- librispeech_asr
|
5 |
+
tags:
|
6 |
+
- audio
|
7 |
+
- automatic-speech-recognition
|
8 |
+
license: apache-2.0
|
9 |
+
---
|
10 |
+
|
11 |
+
TODO: [To be filled]
|
12 |
+
|
13 |
+
|
14 |
+
## Evaluation on LibriSpeech Test
|
15 |
+
|
16 |
+
The following script shows how to evaluate this model on the [LibriSpeech](https://huggingface.co/datasets/librispeech_asr) *"clean"* and *"other"* test dataset.
|
17 |
+
|
18 |
+
```python
|
19 |
+
from datasets import load_dataset
|
20 |
+
from transformers import Speech2TextTransformerForConditionalGeneration, Speech2TextTransformerTokenizer
|
21 |
+
import soundfile as sf
|
22 |
+
from jiwer import wer
|
23 |
+
|
24 |
+
librispeech_eval = load_dataset("librispeech_asr", "clean", split="test") # change to "other" for other test dataset
|
25 |
+
|
26 |
+
model = Speech2TextTransformerForConditionalGeneration.from_pretrained("valhalla/s2t_librispeech_medium").to("cuda")
|
27 |
+
tokenizer = Speech2TextTransformerTokenizer.from_pretrained("valhalla/s2t_librispeech_medium", do_upper_case=True)
|
28 |
+
|
29 |
+
def map_to_array(batch):
|
30 |
+
speech, _ = sf.read(batch["file"])
|
31 |
+
batch["speech"] = speech
|
32 |
+
return batch
|
33 |
+
|
34 |
+
librispeech_eval = librispeech_eval.map(map_to_array)
|
35 |
+
|
36 |
+
def map_to_pred(batch):
|
37 |
+
features = tokenizer(batch["speech"], sample_rate=16000, padding=True, return_tensors="pt")
|
38 |
+
input_features = features.input_features.to("cuda")
|
39 |
+
attention_mask = features.attention_mask.to("cuda")
|
40 |
+
|
41 |
+
gen_tokens = model.generate(input_ids=input_features, attention_mask=attention_mask)
|
42 |
+
batch["transcription"] = tokenizer.batch_decode(gen_tokens, skip_special_tokens=True)
|
43 |
+
return batch
|
44 |
+
|
45 |
+
result = librispeech_eval.map(map_to_pred, batched=True, batch_size=8, remove_columns=["speech"])
|
46 |
+
|
47 |
+
print("WER:", wer(result["text"], result["transcription"]))
|
48 |
+
```
|
49 |
+
|
50 |
+
*Result (WER)*:
|
51 |
+
|
52 |
+
| "clean" | "other" |
|
53 |
+
|---|---|
|
54 |
+
| 3.5 | 7.8 |
|