ivangtorre
commited on
Commit
•
d35ad98
1
Parent(s):
cd3386c
Add model files
Browse files- README.md +3 -3
- pytorch_model.bin +1 -1
- trainer_state.json +12 -12
- training_args.bin +1 -1
- vocab.json +1 -1
README.md
CHANGED
@@ -21,7 +21,7 @@ model-index:
|
|
21 |
metrics:
|
22 |
- name: Test WER
|
23 |
type: wer
|
24 |
-
value:
|
25 |
---
|
26 |
|
27 |
# Wav2Vec2-Large-XLSR-53-euskera
|
@@ -84,7 +84,7 @@ processor = Wav2Vec2Processor.from_pretrained("ivangtorre/wav2vec2-large-xlsr-53
|
|
84 |
model = Wav2Vec2ForCTC.from_pretrained("ivangtorre/wav2vec2-large-xlsr-53-basque")
|
85 |
model.to("cuda")
|
86 |
|
87 |
-
chars_to_ignore_regex = '
|
88 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
89 |
|
90 |
# Preprocessing the datasets.
|
@@ -114,7 +114,7 @@ result = test_dataset.map(evaluate, batched=True, batch_size=8)
|
|
114 |
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
|
115 |
```
|
116 |
|
117 |
-
**Test Result**:
|
118 |
|
119 |
|
120 |
## Training
|
21 |
metrics:
|
22 |
- name: Test WER
|
23 |
type: wer
|
24 |
+
value: 44.48
|
25 |
---
|
26 |
|
27 |
# Wav2Vec2-Large-XLSR-53-euskera
|
84 |
model = Wav2Vec2ForCTC.from_pretrained("ivangtorre/wav2vec2-large-xlsr-53-basque")
|
85 |
model.to("cuda")
|
86 |
|
87 |
+
chars_to_ignore_regex = '[",", "?", ".", "!", "-", ";", ":", '""', "%", "'", '"', "�"]'
|
88 |
resampler = torchaudio.transforms.Resample(48_000, 16_000)
|
89 |
|
90 |
# Preprocessing the datasets.
|
114 |
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
|
115 |
```
|
116 |
|
117 |
+
**Test Result**: 44.48 %
|
118 |
|
119 |
|
120 |
## Training
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1262055847
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:509588dea56f1c5e67678b30228db8d4070585445da2cfab49c13ee8fc134dc5
|
3 |
size 1262055847
|
trainer_state.json
CHANGED
@@ -1,30 +1,30 @@
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
-
"epoch":
|
5 |
"global_step": 500,
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
9 |
"log_history": [
|
10 |
{
|
11 |
-
"epoch":
|
12 |
-
"learning_rate": 0.
|
13 |
-
"loss": 1.
|
14 |
"step": 500
|
15 |
},
|
16 |
{
|
17 |
-
"epoch":
|
18 |
-
"eval_loss": 0.
|
19 |
-
"eval_runtime":
|
20 |
-
"eval_samples_per_second":
|
21 |
-
"eval_wer": 0.
|
22 |
"step": 500
|
23 |
}
|
24 |
],
|
25 |
-
"max_steps":
|
26 |
-
"num_train_epochs":
|
27 |
-
"total_flos":
|
28 |
"trial_name": null,
|
29 |
"trial_params": null
|
30 |
}
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 3.785894206549118,
|
5 |
"global_step": 500,
|
6 |
"is_hyper_param_search": false,
|
7 |
"is_local_process_zero": true,
|
8 |
"is_world_process_zero": true,
|
9 |
"log_history": [
|
10 |
{
|
11 |
+
"epoch": 3.79,
|
12 |
+
"learning_rate": 0.00039436231366470835,
|
13 |
+
"loss": 1.5121,
|
14 |
"step": 500
|
15 |
},
|
16 |
{
|
17 |
+
"epoch": 3.79,
|
18 |
+
"eval_loss": 0.186998650431633,
|
19 |
+
"eval_runtime": 714.138,
|
20 |
+
"eval_samples_per_second": 7.242,
|
21 |
+
"eval_wer": 0.39502654080485317,
|
22 |
"step": 500
|
23 |
}
|
24 |
],
|
25 |
+
"max_steps": 6600,
|
26 |
+
"num_train_epochs": 50,
|
27 |
+
"total_flos": 7.843188479774457e+18,
|
28 |
"trial_name": null,
|
29 |
"trial_params": null
|
30 |
}
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 2287
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:38479a919eda7dd28d3ae5f9cab083881c8741ef35462dbc9ab73b870d8c6d48
|
3 |
size 2287
|
vocab.json
CHANGED
@@ -1 +1 @@
|
|
1 |
-
{"
|
1 |
+
{"m": 0, "d": 1, "c": 2, "l": 3, "b": 4, "i": 5, "y": 6, "w": 7, "k": 8, "n": 10, "\u00f1": 11, "o": 12, "a": 13, "x": 14, "z": 15, "u": 16, "r": 17, "q": 18, "h": 19, "t": 20, "g": 21, "p": 22, "v": 23, "e": 24, "j": 25, "f": 26, "s": 27, "\u00ed": 28, "|": 9, "[UNK]": 29, "[PAD]": 30}
|