End of training
Browse files
README.md
CHANGED
|
@@ -14,8 +14,13 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 14 |
|
| 15 |
This model is a fine-tuned version of [Ransaka/sinhala-ocr-model](https://huggingface.co/Ransaka/sinhala-ocr-model) on an unknown dataset.
|
| 16 |
It achieves the following results on the evaluation set:
|
| 17 |
-
-
|
| 18 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
## Model description
|
| 21 |
|
|
@@ -34,35 +39,17 @@ More information needed
|
|
| 34 |
### Training hyperparameters
|
| 35 |
|
| 36 |
The following hyperparameters were used during training:
|
| 37 |
-
- learning_rate:
|
| 38 |
- train_batch_size: 4
|
| 39 |
- eval_batch_size: 4
|
| 40 |
- seed: 42
|
| 41 |
-
- gradient_accumulation_steps:
|
| 42 |
-
- total_train_batch_size:
|
| 43 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 44 |
- lr_scheduler_type: linear
|
| 45 |
- training_steps: 6000
|
| 46 |
- mixed_precision_training: Native AMP
|
| 47 |
|
| 48 |
-
### Training results
|
| 49 |
-
|
| 50 |
-
| Training Loss | Epoch | Step | Validation Loss | Cer |
|
| 51 |
-
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 52 |
-
| 4.543 | 3.27 | 500 | 6.2682 | 0.7086 |
|
| 53 |
-
| 2.6146 | 6.54 | 1000 | 5.8348 | 0.6390 |
|
| 54 |
-
| 1.8448 | 9.8 | 1500 | 5.8076 | 0.6166 |
|
| 55 |
-
| 1.3887 | 13.07 | 2000 | 6.0250 | 0.6072 |
|
| 56 |
-
| 1.0271 | 16.34 | 2500 | 5.9971 | 0.5707 |
|
| 57 |
-
| 0.8891 | 19.61 | 3000 | 5.9803 | 0.5630 |
|
| 58 |
-
| 0.6548 | 22.88 | 3500 | 6.0045 | 0.5542 |
|
| 59 |
-
| 0.4939 | 26.14 | 4000 | 6.0223 | 0.5354 |
|
| 60 |
-
| 0.322 | 29.41 | 4500 | 6.1360 | 0.5233 |
|
| 61 |
-
| 0.2459 | 32.68 | 5000 | 6.1166 | 0.5220 |
|
| 62 |
-
| 0.123 | 35.95 | 5500 | 6.1740 | 0.5162 |
|
| 63 |
-
| 0.1575 | 39.22 | 6000 | 6.2306 | 0.5161 |
|
| 64 |
-
|
| 65 |
-
|
| 66 |
### Framework versions
|
| 67 |
|
| 68 |
- Transformers 4.35.2
|
|
|
|
| 14 |
|
| 15 |
This model is a fine-tuned version of [Ransaka/sinhala-ocr-model](https://huggingface.co/Ransaka/sinhala-ocr-model) on an unknown dataset.
|
| 16 |
It achieves the following results on the evaluation set:
|
| 17 |
+
- eval_loss: 4.8494
|
| 18 |
+
- eval_cer: 0.4227
|
| 19 |
+
- eval_runtime: 229.7041
|
| 20 |
+
- eval_samples_per_second: 1.776
|
| 21 |
+
- eval_steps_per_second: 0.444
|
| 22 |
+
- epoch: 5.23
|
| 23 |
+
- step: 400
|
| 24 |
|
| 25 |
## Model description
|
| 26 |
|
|
|
|
| 39 |
### Training hyperparameters
|
| 40 |
|
| 41 |
The following hyperparameters were used during training:
|
| 42 |
+
- learning_rate: 1e-05
|
| 43 |
- train_batch_size: 4
|
| 44 |
- eval_batch_size: 4
|
| 45 |
- seed: 42
|
| 46 |
+
- gradient_accumulation_steps: 4
|
| 47 |
+
- total_train_batch_size: 16
|
| 48 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 49 |
- lr_scheduler_type: linear
|
| 50 |
- training_steps: 6000
|
| 51 |
- mixed_precision_training: Native AMP
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
### Framework versions
|
| 54 |
|
| 55 |
- Transformers 4.35.2
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1260933520
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c9d409c533ee4e3ed6cb94d3e71aefb587b699a7efbeaaf08ee089a86133d7c1
|
| 3 |
size 1260933520
|
runs/Jan05_03-20-25_beb7e32edffa/events.out.tfevents.1704424826.beb7e32edffa.42.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d22e0022c92a123a7d92c81964f58f9a7d542c32a45909900aafc488a5135c4a
|
| 3 |
+
size 12251
|