Commit
•
6625434
1
Parent(s):
ead7d39
Update README.md
Browse files
README.md
CHANGED
|
@@ -2,8 +2,14 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
base_model: openai/whisper-large-v3
|
| 4 |
tags:
|
|
|
|
|
|
|
|
|
|
| 5 |
- generated_from_trainer
|
|
|
|
|
|
|
| 6 |
metrics:
|
|
|
|
| 7 |
- wer
|
| 8 |
model-index:
|
| 9 |
- name: whisper-large-v3-translate-zh-v0.1-lt
|
|
@@ -16,7 +22,7 @@ This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingf
|
|
| 16 |
|
| 17 |
## Model description
|
| 18 |
|
| 19 |
-
3500
|
| 20 |
|
| 21 |
## Usage
|
| 22 |
|
|
@@ -32,14 +38,14 @@ The following hyperparameters were used during training:
|
|
| 32 |
- train_batch_size: 32
|
| 33 |
- eval_batch_size: 32
|
| 34 |
- seed: 42
|
| 35 |
-
- distributed_type: multi-GPU
|
| 36 |
- gradient_accumulation_steps: 2
|
| 37 |
- total_train_batch_size: 64
|
| 38 |
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
|
| 39 |
- lr_scheduler_type: linear
|
| 40 |
- lr_scheduler_warmup_steps: 4000
|
| 41 |
-
-
|
| 42 |
-
-
|
|
|
|
| 43 |
|
| 44 |
### Training results
|
| 45 |
|
|
@@ -92,4 +98,4 @@ The following hyperparameters were used during training:
|
|
| 92 |
- Transformers 4.41.1
|
| 93 |
- Pytorch 2.3.1+cu121
|
| 94 |
- Datasets 2.19.1
|
| 95 |
-
- Tokenizers 0.19.1
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
base_model: openai/whisper-large-v3
|
| 4 |
tags:
|
| 5 |
+
- audio
|
| 6 |
+
- automatic-speech-recognition
|
| 7 |
+
- translate
|
| 8 |
- generated_from_trainer
|
| 9 |
+
language:
|
| 10 |
+
- zh
|
| 11 |
metrics:
|
| 12 |
+
- cer
|
| 13 |
- wer
|
| 14 |
model-index:
|
| 15 |
- name: whisper-large-v3-translate-zh-v0.1-lt
|
|
|
|
| 22 |
|
| 23 |
## Model description
|
| 24 |
|
| 25 |
+
3500小时数据微调, 翻译模式直出中文
|
| 26 |
|
| 27 |
## Usage
|
| 28 |
|
|
|
|
| 38 |
- train_batch_size: 32
|
| 39 |
- eval_batch_size: 32
|
| 40 |
- seed: 42
|
|
|
|
| 41 |
- gradient_accumulation_steps: 2
|
| 42 |
- total_train_batch_size: 64
|
| 43 |
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
|
| 44 |
- lr_scheduler_type: linear
|
| 45 |
- lr_scheduler_warmup_steps: 4000
|
| 46 |
+
- dropout: 0.1
|
| 47 |
+
- mask_time_prob: 0.05
|
| 48 |
+
- mask_feature_prob: 0.2
|
| 49 |
|
| 50 |
### Training results
|
| 51 |
|
|
|
|
| 98 |
- Transformers 4.41.1
|
| 99 |
- Pytorch 2.3.1+cu121
|
| 100 |
- Datasets 2.19.1
|
| 101 |
+
- Tokenizers 0.19.1
|