Feature/wandb logging

[asawczyn]

additionally close #152, but there is still a problem with the flair hps pipelines e.g. embeddings/pipeline/flair_hps_pipeline.py:139

an example hps logging output:
https://wandb.ai/albert__/test-hps-run-3
for source code:

from embeddings.hyperparameter_search.lighting_configspace import (
    LightingTextClassificationConfigSpace,
)
from embeddings.hyperparameter_search.parameters import ConstantParameter
from embeddings.pipeline.lightning_classification import LightningClassificationPipeline
from embeddings.pipeline.lightning_hps_pipeline import OptimizedLightingClassificationPipeline
from embeddings.utils.loggers import LightningLoggingConfig

LightningClassificationPipeline.DEFAULT_DATAMODULE_KWARGS.update(
    {
        "downsample_train": 0.03,
        "downsample_val": 0.03,
        "downsample_test": 0.03,
    }
)
WANDB_PROJECT = "test-hps-run-3"

for embedding_name in ["allegro/herbert-base-cased", "hf-internal-testing/tiny-albert"]:
    config_space = LightingTextClassificationConfigSpace(
        embedding_name_or_path=embedding_name,
        finetune_last_n_layers=ConstantParameter("finetune_last_n_layers", 0),
        max_epochs=ConstantParameter("max_epochs", 5),
    )
    pipeline = OptimizedLightingClassificationPipeline(
        config_space=config_space,
        dataset_name_or_path="clarin-pl/polemo2-official",
        input_column_name="text",
        target_column_name="target",
        logging_config=LightningLoggingConfig.from_flags(
            wandb=True,
            tracking_project_name=WANDB_PROJECT,
        ),
        n_trials=3,
    ).persisting(
        best_params_path="best_params.yaml",
        log_path="hps_log.pickle",
    )
    df, metadata = pipeline.run(run_name=f"search-{embedding_name}")

    pipeline = LightningClassificationPipeline(
        output_path=".",
        logging_config=LightningLoggingConfig.from_flags(
            wandb=True,
            tracking_project_name=WANDB_PROJECT,
        ),
        **metadata,
    )
    pipeline.run(run_name=f"best-params-retrain-{embedding_name}")

[netlify]

✅ Deploy Preview for embeddingsclarinpl canceled.

Name	Link
🔨 Latest commit	01f5b41
🔍 Latest deploy log	https://app.netlify.com/sites/embeddingsclarinpl/deploys/624d87a14d32410008b75f5d