Commit
•
c16a918
1
Parent(s):
affa7ed
Update README.md
Browse files
README.md
CHANGED
|
@@ -19,56 +19,10 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 19 |
# git-base-instagram-cap
|
| 20 |
|
| 21 |
This model is a fine-tuned version of [microsoft/git-base](https://huggingface.co/microsoft/git-base) on an unknown dataset.
|
| 22 |
-
It achieves the following results on the evaluation set:
|
| 23 |
-
- Loss: 0.2581
|
| 24 |
-
- Wer Score: 3.1462
|
| 25 |
-
|
| 26 |
-
## Model description
|
| 27 |
-
|
| 28 |
-
More information needed
|
| 29 |
-
|
| 30 |
-
## Intended uses & limitations
|
| 31 |
-
|
| 32 |
-
More information needed
|
| 33 |
-
|
| 34 |
-
## Training and evaluation data
|
| 35 |
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
## Training procedure
|
| 39 |
-
|
| 40 |
-
### Training hyperparameters
|
| 41 |
-
|
| 42 |
-
The following hyperparameters were used during training:
|
| 43 |
-
- learning_rate: 5e-05
|
| 44 |
-
- train_batch_size: 32
|
| 45 |
-
- eval_batch_size: 32
|
| 46 |
-
- seed: 42
|
| 47 |
-
- gradient_accumulation_steps: 2
|
| 48 |
-
- total_train_batch_size: 64
|
| 49 |
-
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 50 |
-
- lr_scheduler_type: linear
|
| 51 |
-
- num_epochs: 50
|
| 52 |
-
- mixed_precision_training: Native AMP
|
| 53 |
-
|
| 54 |
-
### Training results
|
| 55 |
|
| 56 |
-
| Training Loss | Epoch | Step | Validation Loss | Wer Score |
|
| 57 |
-
|:-------------:|:-----:|:----:|:---------------:|:---------:|
|
| 58 |
-
| 7.3639 | 3.45 | 50 | 4.5936 | 0.9495 |
|
| 59 |
-
| 2.389 | 6.9 | 100 | 0.6192 | 0.9387 |
|
| 60 |
-
| 0.2242 | 10.34 | 150 | 0.2274 | 0.9036 |
|
| 61 |
-
| 0.0883 | 13.79 | 200 | 0.2271 | 0.9084 |
|
| 62 |
-
| 0.0643 | 17.24 | 250 | 0.2319 | 0.9387 |
|
| 63 |
-
| 0.0493 | 20.69 | 300 | 0.2388 | 0.9501 |
|
| 64 |
-
| 0.0409 | 24.14 | 350 | 0.2435 | 0.9522 |
|
| 65 |
-
| 0.0344 | 27.59 | 400 | 0.2458 | 0.9616 |
|
| 66 |
-
| 0.0311 | 31.03 | 450 | 0.2492 | 1.0027 |
|
| 67 |
-
| 0.028 | 34.48 | 500 | 0.2519 | 1.3565 |
|
| 68 |
-
| 0.0255 | 37.93 | 550 | 0.2540 | 2.0640 |
|
| 69 |
-
| 0.0227 | 41.38 | 600 | 0.2554 | 2.1813 |
|
| 70 |
-
| 0.0189 | 44.83 | 650 | 0.2575 | 2.8416 |
|
| 71 |
-
| 0.0154 | 48.28 | 700 | 0.2581 | 3.1462 |
|
| 72 |
|
| 73 |
|
| 74 |
### Framework versions
|
|
|
|
| 19 |
# git-base-instagram-cap
|
| 20 |
|
| 21 |
This model is a fine-tuned version of [microsoft/git-base](https://huggingface.co/microsoft/git-base) on an unknown dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
+
It achieves the following results on the evaluation set:
|
| 24 |
+
- Loss: 0.0093
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
|
| 28 |
### Framework versions
|