Update README.md
Browse files
README.md
CHANGED
@@ -11,25 +11,19 @@ probably proofread and complete it, then remove this comment. -->
|
|
11 |
|
12 |
# GeoBERT
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
- Train Loss: 0.0660
|
17 |
-
- Validation Loss: 0.0470
|
18 |
-
- Epoch: 0
|
19 |
|
20 |
-
## Model description
|
21 |
|
22 |
-
|
23 |
|
24 |
-
|
25 |
|
26 |
-
|
|
|
|
|
|
|
27 |
|
28 |
-
## Training and evaluation data
|
29 |
-
|
30 |
-
More information needed
|
31 |
-
|
32 |
-
## Training procedure
|
33 |
|
34 |
### Training hyperparameters
|
35 |
|
@@ -37,12 +31,6 @@ The following hyperparameters were used during training:
|
|
37 |
- optimizer: {'name': 'AdamWeightDecay', 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 14000, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False, 'weight_decay_rate': 0.01}
|
38 |
- training_precision: mixed_float16
|
39 |
|
40 |
-
### Training results
|
41 |
-
|
42 |
-
| Train Loss | Validation Loss | Epoch |
|
43 |
-
|:----------:|:---------------:|:-----:|
|
44 |
-
| 0.0660 | 0.0470 | 0 |
|
45 |
-
|
46 |
|
47 |
### Framework versions
|
48 |
|
@@ -50,3 +38,30 @@ The following hyperparameters were used during training:
|
|
50 |
- TensorFlow 2.10.0
|
51 |
- Datasets 2.4.0
|
52 |
- Tokenizers 0.12.1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
|
12 |
# GeoBERT
|
13 |
|
14 |
+
GeoBERT is a NER model that was fine-tuned from SciBERT on the Geoscientific Corpus dataset.
|
15 |
+
The model was trained on the Labeled Geoscientific Corpus dataset (~1 million sentences).
|
|
|
|
|
|
|
16 |
|
|
|
17 |
|
18 |
+
## Intended uses
|
19 |
|
20 |
+
The NER product in this model has a goal to identify four main semantic types or categories related to Geosciences.
|
21 |
|
22 |
+
1. GeoPetro for any entities that belong to all terms in Geosciences
|
23 |
+
2. GeoMeth for all tools or methods associated with Geosciences
|
24 |
+
3. GeoLoc to identify geological locations
|
25 |
+
4. GeoTime for identifying the geological time scale entities
|
26 |
|
|
|
|
|
|
|
|
|
|
|
27 |
|
28 |
### Training hyperparameters
|
29 |
|
|
|
31 |
- optimizer: {'name': 'AdamWeightDecay', 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 14000, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False, 'weight_decay_rate': 0.01}
|
32 |
- training_precision: mixed_float16
|
33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
### Framework versions
|
36 |
|
|
|
38 |
- TensorFlow 2.10.0
|
39 |
- Datasets 2.4.0
|
40 |
- Tokenizers 0.12.1
|
41 |
+
|
42 |
+
## Model performances (metric: seqeval)
|
43 |
+
|
44 |
+
Overall
|
45 |
+
|
46 |
+
precision|recall|f1
|
47 |
+
-|-|-
|
48 |
+
0.8859|0.8971|0.8914
|
49 |
+
|
50 |
+
By entity
|
51 |
+
|
52 |
+
entity|precision|recall|f1
|
53 |
+
-|-|-|-
|
54 |
+
GeoLoc |0.9727|0.9591|0.9658
|
55 |
+
GeoMeth |0.9433|0.9447|0.9445
|
56 |
+
GeoPetro|0.9767|0.9745|0.9756
|
57 |
+
GeoTime |0.9695|0.9666|0.9680
|
58 |
+
|
59 |
+
## How to use GeoBERT with HuggingFace
|
60 |
+
|
61 |
+
##### Load GeoBERT and its sub-word tokenizer :
|
62 |
+
|
63 |
+
```python
|
64 |
+
from transformers import AutoTokenizer, AutoModelForTokenClassification
|
65 |
+
|
66 |
+
tokenizer = AutoTokenizer.from_pretrained("botryan96/GeoBERT")
|
67 |
+
model = AutoModelForTokenClassification.from_pretrained("botryan96/GeoBERT")
|