BeardedJohn
commited on
Commit
•
5080428
1
Parent(s):
544c4d5
Added the results on the conll-endava mixed v2 dataset.
Browse files
README.md
CHANGED
@@ -25,6 +25,11 @@ This model is a fine-tuned version of [BERT](https://huggingface.co/bert-base-ca
|
|
25 |
- [NERPERDemo dataset](https://huggingface.co/datasets/BeardedJohn/NERPERDemo)
|
26 |
- 12000 instances of the [wikiann, english version dataset](https://huggingface.co/datasets/wikiann/viewer/en/train).
|
27 |
|
|
|
|
|
|
|
|
|
|
|
28 |
It achieves the following results on the NERPERDemo evaluation set:
|
29 |
- Train Loss: 0.0005
|
30 |
- Validation Loss: 0.0002
|
@@ -67,6 +72,15 @@ The following hyperparameters were used during training:
|
|
67 |
|
68 |
### Training results
|
69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
On NERPERDemo:
|
71 |
|
72 |
| Train Loss | Validation Loss | Epoch |
|
|
|
25 |
- [NERPERDemo dataset](https://huggingface.co/datasets/BeardedJohn/NERPERDemo)
|
26 |
- 12000 instances of the [wikiann, english version dataset](https://huggingface.co/datasets/wikiann/viewer/en/train).
|
27 |
|
28 |
+
It achieves the following results on the conll-endava mixed, second version evaluation set:
|
29 |
+
- Train Loss: 0.0190
|
30 |
+
- Validation Loss: 0.0310
|
31 |
+
- Epoch: 2
|
32 |
+
-
|
33 |
It achieves the following results on the NERPERDemo evaluation set:
|
34 |
- Train Loss: 0.0005
|
35 |
- Validation Loss: 0.0002
|
|
|
72 |
|
73 |
### Training results
|
74 |
|
75 |
+
On conll-endava mixed, second version:
|
76 |
+
|
77 |
+
| Train Loss | Validation Loss | Epoch |
|
78 |
+
|:----------:|:---------------:|:-----:|
|
79 |
+
| 0.2091 | 0.0391 | 0 |
|
80 |
+
| 0.0336 | 0.0322 | 1 |
|
81 |
+
| 0.0190 | 0.0310 | 2 |
|
82 |
+
|
83 |
+
|
84 |
On NERPERDemo:
|
85 |
|
86 |
| Train Loss | Validation Loss | Epoch |
|