copied w2v fb
Browse files- README.md +63 -0
- config.json +68 -0
- flax_model.msgpack +3 -0
- local_disk0/w2v-sub/.gitattributes +17 -0
- preprocessor_config.json +8 -0
- pytorch_model.bin +3 -0
- special_tokens_map.json +1 -0
- tokenizer_config.json +1 -0
- vocab.json +1 -0
README.md
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language: nl
|
3 |
+
datasets:
|
4 |
+
- common_voice
|
5 |
+
tags:
|
6 |
+
- speech
|
7 |
+
- audio
|
8 |
+
- automatic-speech-recognition
|
9 |
+
license: apache-2.0
|
10 |
+
---
|
11 |
+
|
12 |
+
## Evaluation on Common Voice NL Test
|
13 |
+
|
14 |
+
```python
|
15 |
+
import torchaudio
|
16 |
+
from datasets import load_dataset, load_metric
|
17 |
+
from transformers import (
|
18 |
+
Wav2Vec2ForCTC,
|
19 |
+
Wav2Vec2Processor,
|
20 |
+
)
|
21 |
+
import torch
|
22 |
+
import re
|
23 |
+
import sys
|
24 |
+
|
25 |
+
model_name = "facebook/wav2vec2-large-xlsr-53-dutch"
|
26 |
+
device = "cuda"
|
27 |
+
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"]' # noqa: W605
|
28 |
+
|
29 |
+
model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
|
30 |
+
processor = Wav2Vec2Processor.from_pretrained(model_name)
|
31 |
+
|
32 |
+
ds = load_dataset("common_voice", "nl", split="test", data_dir="./cv-corpus-6.1-2020-12-11")
|
33 |
+
|
34 |
+
resampler = torchaudio.transforms.Resample(orig_freq=48_000, new_freq=16_000)
|
35 |
+
|
36 |
+
def map_to_array(batch):
|
37 |
+
speech, _ = torchaudio.load(batch["path"])
|
38 |
+
batch["speech"] = resampler.forward(speech.squeeze(0)).numpy()
|
39 |
+
batch["sampling_rate"] = resampler.new_freq
|
40 |
+
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
|
41 |
+
return batch
|
42 |
+
|
43 |
+
ds = ds.map(map_to_array)
|
44 |
+
|
45 |
+
|
46 |
+
def map_to_pred(batch):
|
47 |
+
features = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0], padding=True, return_tensors="pt")
|
48 |
+
input_values = features.input_values.to(device)
|
49 |
+
attention_mask = features.attention_mask.to(device)
|
50 |
+
with torch.no_grad():
|
51 |
+
logits = model(input_values, attention_mask=attention_mask).logits
|
52 |
+
pred_ids = torch.argmax(logits, dim=-1)
|
53 |
+
batch["predicted"] = processor.batch_decode(pred_ids)
|
54 |
+
batch["target"] = batch["sentence"]
|
55 |
+
return batch
|
56 |
+
|
57 |
+
result = ds.map(map_to_pred, batched=True, batch_size=16, remove_columns=list(ds.features.keys()))
|
58 |
+
|
59 |
+
wer = load_metric("wer")
|
60 |
+
print(wer.compute(predictions=result["predicted"], references=result["target"]))
|
61 |
+
```
|
62 |
+
|
63 |
+
**Result**: 21.1 %
|
config.json
ADDED
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"activation_dropout": 0.1,
|
3 |
+
"apply_spec_augment": true,
|
4 |
+
"architectures": [
|
5 |
+
"Wav2Vec2ForCTC"
|
6 |
+
],
|
7 |
+
"attention_dropout": 0.1,
|
8 |
+
"bos_token_id": 1,
|
9 |
+
"conv_bias": true,
|
10 |
+
"conv_dim": [
|
11 |
+
512,
|
12 |
+
512,
|
13 |
+
512,
|
14 |
+
512,
|
15 |
+
512,
|
16 |
+
512,
|
17 |
+
512
|
18 |
+
],
|
19 |
+
"conv_kernel": [
|
20 |
+
10,
|
21 |
+
3,
|
22 |
+
3,
|
23 |
+
3,
|
24 |
+
3,
|
25 |
+
2,
|
26 |
+
2
|
27 |
+
],
|
28 |
+
"conv_stride": [
|
29 |
+
5,
|
30 |
+
2,
|
31 |
+
2,
|
32 |
+
2,
|
33 |
+
2,
|
34 |
+
2,
|
35 |
+
2
|
36 |
+
],
|
37 |
+
"ctc_loss_reduction": "sum",
|
38 |
+
"ctc_zero_infinity": false,
|
39 |
+
"do_stable_layer_norm": true,
|
40 |
+
"eos_token_id": 2,
|
41 |
+
"feat_extract_activation": "gelu",
|
42 |
+
"feat_extract_dropout": 0.0,
|
43 |
+
"feat_extract_norm": "layer",
|
44 |
+
"feat_proj_dropout": 0.1,
|
45 |
+
"final_dropout": 0.1,
|
46 |
+
"gradient_checkpointing": false,
|
47 |
+
"hidden_act": "gelu",
|
48 |
+
"hidden_dropout": 0.1,
|
49 |
+
"hidden_dropout_prob": 0.1,
|
50 |
+
"hidden_size": 1024,
|
51 |
+
"initializer_range": 0.02,
|
52 |
+
"intermediate_size": 4096,
|
53 |
+
"layer_norm_eps": 1e-05,
|
54 |
+
"layerdrop": 0.1,
|
55 |
+
"mask_feature_length": 10,
|
56 |
+
"mask_feature_prob": 0.0,
|
57 |
+
"mask_time_length": 10,
|
58 |
+
"mask_time_prob": 0.05,
|
59 |
+
"model_type": "wav2vec2",
|
60 |
+
"num_attention_heads": 16,
|
61 |
+
"num_conv_pos_embedding_groups": 16,
|
62 |
+
"num_conv_pos_embeddings": 128,
|
63 |
+
"num_feat_extract_layers": 7,
|
64 |
+
"num_hidden_layers": 24,
|
65 |
+
"pad_token_id": 0,
|
66 |
+
"transformers_version": "4.4.0.dev0",
|
67 |
+
"vocab_size": 50
|
68 |
+
}
|
flax_model.msgpack
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9974b52afaf933d231d2ada715ff43a40a97c8d2ced1c36aafece145d6fcab8b
|
3 |
+
size 1261975272
|
local_disk0/w2v-sub/.gitattributes
ADDED
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.bin.* filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
5 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
6 |
+
*.tar.gz filter=lfs diff=lfs merge=lfs -text
|
7 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
8 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
9 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
10 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
11 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
14 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
15 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
16 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
17 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
preprocessor_config.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"do_normalize": true,
|
3 |
+
"feature_size": 1,
|
4 |
+
"padding_side": "right",
|
5 |
+
"padding_value": 0.0,
|
6 |
+
"return_attention_mask": true,
|
7 |
+
"sampling_rate": 16000
|
8 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c803055f87112ce0b28b3169f6c0960a792ba43c301cb6b9eea8de2179e2d8b0
|
3 |
+
size 1262132951
|
special_tokens_map.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}
|
tokenizer_config.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "<pad>", "do_lower_case": true, "word_delimiter_token": "|", "special_tokens_map_file": "/home/patrick/hugging_face/add_wav2vec/hf/wav2vec2-large-xlsr-53-dutch/special_tokens_map.json", "tokenizer_file": null, "name_or_path": "/home/patrick/hugging_face/add_wav2vec/hf/wav2vec2-large-xlsr-53-dutch"}
|
vocab.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "E": 4, "N": 5, "A": 6, "I": 7, "T": 8, "O": 9, "D": 10, "R": 11, "|": 12, "L": 13, "S": 14, "H": 15, "G": 16, "M": 17, "K": 18, "V": 19, "J": 20, "W": 21, "Z": 22, "U": 23, "B": 24, "C": 25, "P": 26, "F": 27, "Y": 28, "É": 29, "'": 30, "X": 31, "Ë": 32, "Q": 33, "-": 34, "Ê": 35, "À": 36, "Ä": 37, "È": 38, "Ï": 39, "Â": 40, "Û": 41, "Ö": 42, "Ô": 43, "Ü": 44, "Î": 45, "Ç": 46, "Æ": 47, "Ù": 48, "Œ": 49}
|