sentiment-analysis-using-multimodal-AI-Model/train_sagemaker.py at Sentiment-Analysis · GeneralSubhra/sentiment-analysis-using-multimodal-AI-Model · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
from sagemaker.pytorch import PyTorch
from sagemaker.debugger import TensorBoardOutputConfig


def start_training():
    tensorboard_config = TensorBoardOutputConfig(
        s3_output_path="s3://your-bucket-name/tensorboard",
        container_local_output_path="/opt/ml/output/tensorboard"
    )

    estimator = PyTorch(
        entry_point="train.py",
        source_dir="training",
        role="your-execution-role",
        framework_version="2.5.1",
        py_version="py311",
        instance_count=1,
        instance_type="ml.g5.xlarge",
        hyperparameters={
            "batch-size": 32,
            "epochs": 25
        },
        tensorboard_config=tensorboard_config
    )

    # Start training
    estimator.fit({
        "training": "s3://your-bucket-name/dataset/train",
        "validation": "s3://your-bucket-name/dataset/dev",
        "test": "s3://your-bucket-name/dataset/test"
    })


if __name__ == "__main__":
    start_training()