Update README.md
Browse files
README.md
CHANGED
@@ -15,20 +15,20 @@ should probably proofread and complete it, then remove this comment. -->
|
|
15 |
|
16 |
This model is a fine-tuned version of [](https://huggingface.co/) on the id_liputan6 dataset.
|
17 |
It achieves the following results on the evaluation set:
|
18 |
-
- Loss:
|
19 |
-
- R1 Precision: 0.
|
20 |
-
- R1 Recall: 0.
|
21 |
-
- R1 Fmeasure: 0.
|
22 |
-
- R2 Precision: 0.
|
23 |
-
- R2 Recall: 0.
|
24 |
-
- R2 Fmeasure: 0.
|
25 |
-
- Rl Precision: 0.
|
26 |
-
- Rl Recall: 0.
|
27 |
-
- Rl Fmeasure: 0.
|
28 |
|
29 |
## Model description
|
30 |
|
31 |
-
|
32 |
|
33 |
## Intended uses & limitations
|
34 |
|
@@ -39,27 +39,21 @@ More information needed
|
|
39 |
More information needed
|
40 |
|
41 |
## Training procedure
|
|
|
|
|
42 |
|
43 |
### Training hyperparameters
|
44 |
|
45 |
The following hyperparameters were used during training:
|
46 |
- learning_rate: 5e-05
|
47 |
-
- train_batch_size:
|
48 |
-
- eval_batch_size:
|
49 |
- seed: 42
|
50 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
51 |
- lr_scheduler_type: linear
|
52 |
-
- num_epochs:
|
53 |
- mixed_precision_training: Native AMP
|
54 |
|
55 |
-
### Training results
|
56 |
-
|
57 |
-
| Training Loss | Epoch | Step | Validation Loss | R1 Precision | R1 Recall | R1 Fmeasure | R2 Precision | R2 Recall | R2 Fmeasure | Rl Precision | Rl Recall | Rl Fmeasure |
|
58 |
-
|:-------------:|:-----:|:----:|:---------------:|:------------:|:---------:|:-----------:|:------------:|:---------:|:-----------:|:------------:|:---------:|:-----------:|
|
59 |
-
| No log | 1.0 | 8 | 8.2130 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
|
60 |
-
| No log | 2.0 | 16 | 8.0750 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
|
61 |
-
|
62 |
-
|
63 |
### Framework versions
|
64 |
|
65 |
- Transformers 4.38.2
|
|
|
15 |
|
16 |
This model is a fine-tuned version of [](https://huggingface.co/) on the id_liputan6 dataset.
|
17 |
It achieves the following results on the evaluation set:
|
18 |
+
- Loss: 2.876
|
19 |
+
- R1 Precision: 0.2907
|
20 |
+
- R1 Recall: 0.3519
|
21 |
+
- R1 Fmeasure: 0.3149
|
22 |
+
- R2 Precision: 0.1166
|
23 |
+
- R2 Recall: 0.1408
|
24 |
+
- R2 Fmeasure: 0.1261
|
25 |
+
- Rl Precision: 0.2301
|
26 |
+
- Rl Recall: 0.2787
|
27 |
+
- Rl Fmeasure: 0.2493
|
28 |
|
29 |
## Model description
|
30 |
|
31 |
+
Encoder Decoder Model using IndoBERT (indobert-base-uncased) as Encoder and Decoder
|
32 |
|
33 |
## Intended uses & limitations
|
34 |
|
|
|
39 |
More information needed
|
40 |
|
41 |
## Training procedure
|
42 |
+
- Max length generation: 80
|
43 |
+
- Min length generation: 10
|
44 |
|
45 |
### Training hyperparameters
|
46 |
|
47 |
The following hyperparameters were used during training:
|
48 |
- learning_rate: 5e-05
|
49 |
+
- train_batch_size: 18
|
50 |
+
- eval_batch_size: 18
|
51 |
- seed: 42
|
52 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
53 |
- lr_scheduler_type: linear
|
54 |
+
- num_epochs: 8
|
55 |
- mixed_precision_training: Native AMP
|
56 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
### Framework versions
|
58 |
|
59 |
- Transformers 4.38.2
|