End of training
Browse files
README.md
CHANGED
@@ -1,10 +1,10 @@
|
|
1 |
---
|
2 |
-
base_model: microsoft/speecht5_tts
|
3 |
-
datasets:
|
4 |
-
- voxpopuli
|
5 |
license: mit
|
|
|
6 |
tags:
|
7 |
- generated_from_trainer
|
|
|
|
|
8 |
model-index:
|
9 |
- name: speecht5_tts_voxpopuli_it_v2
|
10 |
results: []
|
@@ -17,7 +17,7 @@ should probably proofread and complete it, then remove this comment. -->
|
|
17 |
|
18 |
This model is a fine-tuned version of [microsoft/speecht5_tts](https://huggingface.co/microsoft/speecht5_tts) on the voxpopuli dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
-
- Loss: 0.
|
21 |
|
22 |
## Model description
|
23 |
|
@@ -36,37 +36,47 @@ More information needed
|
|
36 |
### Training hyperparameters
|
37 |
|
38 |
The following hyperparameters were used during training:
|
39 |
-
- learning_rate:
|
40 |
-
- train_batch_size:
|
41 |
-
- eval_batch_size:
|
42 |
- seed: 42
|
43 |
-
- gradient_accumulation_steps:
|
44 |
- total_train_batch_size: 32
|
45 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
46 |
- lr_scheduler_type: linear
|
47 |
-
- lr_scheduler_warmup_steps:
|
48 |
-
- training_steps:
|
49 |
- mixed_precision_training: Native AMP
|
50 |
|
51 |
### Training results
|
52 |
|
53 |
| Training Loss | Epoch | Step | Validation Loss |
|
54 |
|:-------------:|:------:|:----:|:---------------:|
|
55 |
-
| 0.
|
56 |
-
| 0.
|
57 |
-
| 0.
|
58 |
-
| 0.
|
59 |
-
| 0.
|
60 |
-
| 0.
|
61 |
-
| 0.
|
62 |
-
| 0.
|
63 |
-
| 0.
|
64 |
-
| 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
65 |
|
66 |
|
67 |
### Framework versions
|
68 |
|
69 |
- Transformers 4.43.1
|
70 |
-
- Pytorch 2.
|
71 |
-
- Datasets 3.0.
|
72 |
- Tokenizers 0.19.1
|
|
|
1 |
---
|
|
|
|
|
|
|
2 |
license: mit
|
3 |
+
base_model: microsoft/speecht5_tts
|
4 |
tags:
|
5 |
- generated_from_trainer
|
6 |
+
datasets:
|
7 |
+
- voxpopuli
|
8 |
model-index:
|
9 |
- name: speecht5_tts_voxpopuli_it_v2
|
10 |
results: []
|
|
|
17 |
|
18 |
This model is a fine-tuned version of [microsoft/speecht5_tts](https://huggingface.co/microsoft/speecht5_tts) on the voxpopuli dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
+
- Loss: 0.4484
|
21 |
|
22 |
## Model description
|
23 |
|
|
|
36 |
### Training hyperparameters
|
37 |
|
38 |
The following hyperparameters were used during training:
|
39 |
+
- learning_rate: 0.0001
|
40 |
+
- train_batch_size: 8
|
41 |
+
- eval_batch_size: 4
|
42 |
- seed: 42
|
43 |
+
- gradient_accumulation_steps: 4
|
44 |
- total_train_batch_size: 32
|
45 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
46 |
- lr_scheduler_type: linear
|
47 |
+
- lr_scheduler_warmup_steps: 200
|
48 |
+
- training_steps: 2000
|
49 |
- mixed_precision_training: Native AMP
|
50 |
|
51 |
### Training results
|
52 |
|
53 |
| Training Loss | Epoch | Step | Validation Loss |
|
54 |
|:-------------:|:------:|:----:|:---------------:|
|
55 |
+
| 0.5707 | 0.2358 | 100 | 0.5183 |
|
56 |
+
| 0.5452 | 0.4717 | 200 | 0.5096 |
|
57 |
+
| 0.5313 | 0.7075 | 300 | 0.4890 |
|
58 |
+
| 0.5229 | 0.9434 | 400 | 0.4807 |
|
59 |
+
| 0.5119 | 1.1792 | 500 | 0.4802 |
|
60 |
+
| 0.5121 | 1.4151 | 600 | 0.4681 |
|
61 |
+
| 0.5037 | 1.6509 | 700 | 0.4719 |
|
62 |
+
| 0.4996 | 1.8868 | 800 | 0.4691 |
|
63 |
+
| 0.4931 | 2.1226 | 900 | 0.4621 |
|
64 |
+
| 0.4903 | 2.3585 | 1000 | 0.4620 |
|
65 |
+
| 0.4949 | 2.5943 | 1100 | 0.4573 |
|
66 |
+
| 0.4853 | 2.8302 | 1200 | 0.4579 |
|
67 |
+
| 0.4826 | 3.0660 | 1300 | 0.4547 |
|
68 |
+
| 0.4827 | 3.3019 | 1400 | 0.4535 |
|
69 |
+
| 0.4835 | 3.5377 | 1500 | 0.4523 |
|
70 |
+
| 0.4802 | 3.7736 | 1600 | 0.4514 |
|
71 |
+
| 0.4777 | 4.0094 | 1700 | 0.4503 |
|
72 |
+
| 0.4792 | 4.2453 | 1800 | 0.4499 |
|
73 |
+
| 0.4779 | 4.4811 | 1900 | 0.4491 |
|
74 |
+
| 0.4755 | 4.7170 | 2000 | 0.4484 |
|
75 |
|
76 |
|
77 |
### Framework versions
|
78 |
|
79 |
- Transformers 4.43.1
|
80 |
+
- Pytorch 2.2.0
|
81 |
+
- Datasets 3.0.1
|
82 |
- Tokenizers 0.19.1
|