Upload train.py with huggingface_hub
Browse files
train.py
CHANGED
@@ -1,28 +1,47 @@
|
|
|
|
|
|
1 |
from datasets import load_dataset
|
2 |
from transformers import TrainingArguments
|
3 |
-
from span_marker import SpanMarkerModel, Trainer
|
|
|
|
|
|
|
|
|
4 |
|
5 |
|
6 |
def main() -> None:
|
7 |
# Load the dataset, ensure "tokens" and "ner_tags" columns, and get a list of labels
|
8 |
-
|
|
|
|
|
9 |
labels = dataset["train"].features["ner_tags"].feature.names
|
10 |
|
11 |
# Initialize a SpanMarker model using a pretrained BERT-style encoder
|
12 |
-
|
|
|
13 |
model = SpanMarkerModel.from_pretrained(
|
14 |
-
|
15 |
labels=labels,
|
16 |
# SpanMarker hyperparameters:
|
17 |
model_max_length=256,
|
18 |
marker_max_length=128,
|
19 |
entity_max_length=8,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
)
|
21 |
|
22 |
# Prepare the 🤗 transformers training arguments
|
|
|
23 |
args = TrainingArguments(
|
24 |
-
output_dir=
|
25 |
-
run_name=
|
26 |
# Training Hyperparameters:
|
27 |
learning_rate=5e-5,
|
28 |
per_device_train_batch_size=32,
|
@@ -49,12 +68,13 @@ def main() -> None:
|
|
49 |
eval_dataset=dataset["validation"],
|
50 |
)
|
51 |
trainer.train()
|
52 |
-
trainer.save_model(f"models/span_marker_bert_base_acronyms/checkpoint-final")
|
53 |
|
54 |
# Compute & save the metrics on the test set
|
55 |
-
metrics = trainer.evaluate()
|
56 |
trainer.save_metrics("validation", metrics)
|
57 |
-
|
|
|
|
|
58 |
|
59 |
|
60 |
if __name__ == "__main__":
|
|
|
1 |
+
from pathlib import Path
|
2 |
+
import shutil
|
3 |
from datasets import load_dataset
|
4 |
from transformers import TrainingArguments
|
5 |
+
from span_marker import SpanMarkerModel, Trainer, SpanMarkerModelCardData
|
6 |
+
|
7 |
+
import os
|
8 |
+
|
9 |
+
os.environ["CODECARBON_LOG_LEVEL"] = "error"
|
10 |
|
11 |
|
12 |
def main() -> None:
|
13 |
# Load the dataset, ensure "tokens" and "ner_tags" columns, and get a list of labels
|
14 |
+
dataset_name = "Acronym Identification"
|
15 |
+
dataset_id = "acronym_identification"
|
16 |
+
dataset = load_dataset(dataset_id).rename_column("labels", "ner_tags")
|
17 |
labels = dataset["train"].features["ner_tags"].feature.names
|
18 |
|
19 |
# Initialize a SpanMarker model using a pretrained BERT-style encoder
|
20 |
+
encoder_id = "bert-base-cased"
|
21 |
+
model_id = "tomaarsen/span-marker-bert-base-acronyms"
|
22 |
model = SpanMarkerModel.from_pretrained(
|
23 |
+
encoder_id,
|
24 |
labels=labels,
|
25 |
# SpanMarker hyperparameters:
|
26 |
model_max_length=256,
|
27 |
marker_max_length=128,
|
28 |
entity_max_length=8,
|
29 |
+
# Model card variables
|
30 |
+
model_card_data=SpanMarkerModelCardData(
|
31 |
+
model_id=model_id,
|
32 |
+
encoder_id=encoder_id,
|
33 |
+
dataset_name=dataset_name,
|
34 |
+
dataset_id=dataset_id,
|
35 |
+
license="apache-2.0",
|
36 |
+
language="en",
|
37 |
+
),
|
38 |
)
|
39 |
|
40 |
# Prepare the 🤗 transformers training arguments
|
41 |
+
output_dir = Path("models") / model_id
|
42 |
args = TrainingArguments(
|
43 |
+
output_dir=output_dir,
|
44 |
+
run_name=model_id,
|
45 |
# Training Hyperparameters:
|
46 |
learning_rate=5e-5,
|
47 |
per_device_train_batch_size=32,
|
|
|
68 |
eval_dataset=dataset["validation"],
|
69 |
)
|
70 |
trainer.train()
|
|
|
71 |
|
72 |
# Compute & save the metrics on the test set
|
73 |
+
metrics = trainer.evaluate(metric_key_prefix="validation")
|
74 |
trainer.save_metrics("validation", metrics)
|
75 |
+
|
76 |
+
trainer.save_model(output_dir / "checkpoint-final")
|
77 |
+
shutil.copy2(__file__, output_dir / "checkpoint-final" / "train.py")
|
78 |
|
79 |
|
80 |
if __name__ == "__main__":
|