benchmarks/models/sasrec/eval.py

import logging
import pathlib as Path
import random

import click
import numpy as np
import polars as pl
import torch
from model import SASRecEncoder
from torch.utils.data import DataLoader

from data import Data, EvalDataset, collate_fn, preprocess
from yambda.evaluation.metrics import calc_metrics
from yambda.evaluation.ranking import Embeddings, Targets, rank_items


logging.basicConfig(
    level=logging.DEBUG, format='[%(asctime)s] [%(levelname)s]: %(message)s', datefmt='%Y-%m-%d %H:%M:%S'
)
logger = logging.getLogger(__name__)


def infer_users(eval_dataloader: DataLoader, model: torch.nn.Module, device: str):
    user_ids = []
    user_embeddings = []

    model.eval()
    for batch in eval_dataloader:
        for key in batch.keys():
            batch[key] = batch[key].to(device)

        user_ids.append(batch['user.ids'])  # (batch_size)
        user_embeddings.append(model(batch))  # (batch_size, embedding_dim)

    return torch.cat(user_ids, dim=0), torch.cat(user_embeddings, dim=0)


def infer_items(model: SASRecEncoder):
    return model.item_embeddings.weight.data


@click.command()
@click.option('--exp_name', required=True, type=str)
@click.option('--data_dir', required=True, type=str, default='../../data/', show_default=True)
@click.option(
    '--size',
    required=True,
    type=click.Choice(['50m', '500m', '5b']),
    default='50m',
    show_default=True,
)
@click.option(
    '--interaction',
    required=True,
    type=click.Choice(['likes', 'listens']),
    default='likes',
    show_default=True,
)
@click.option('--batch_size', required=True, type=int, default=256, show_default=True)
@click.option('--max_seq_len', required=False, type=int, default=200, show_default=True)
@click.option('--seed', required=False, type=int, default=42, show_default=True)
@click.option('--device', required=True, type=str, default='cuda:0', show_default=True)
def main(
    exp_name: str,
    data_dir: str,
    size: str,
    interaction: str,
    batch_size: int,
    max_seq_len: int,
    seed: int,
    device: str,
):
    random.seed(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    torch.set_float32_matmul_precision('high')

    path = Path.Path(data_dir) / 'sequential' / size / interaction
    df = pl.scan_parquet(path.with_suffix('.parquet'))

    logger.debug('Preprocessing data...')
    data: Data = preprocess(df, interaction, val_size=0, max_seq_len=max_seq_len)
    train_df = data.train.collect(engine="streaming")
    eval_df = data.test.collect(engine="streaming")
    logger.debug('Preprocessing data has finished!')

    eval_df = train_df.join(eval_df, on='uid', how='inner', suffix='_valid').select(
        pl.col('uid'), pl.col('item_id').alias('item_id_train'), pl.col('item_id_valid')
    )
    eval_dataset = EvalDataset(dataset=eval_df, max_seq_len=max_seq_len)

    eval_dataloader = DataLoader(
        dataset=eval_dataset,
        batch_size=batch_size,
        collate_fn=collate_fn,
        drop_last=False,
        shuffle=True,
    )

    model = torch.load(f'./checkpoints/{exp_name}_best_state.pth', weights_only=False).to(device)
    model.eval()
    with torch.inference_mode():
        user_ids, user_embeddings = infer_users(eval_dataloader=eval_dataloader, model=model, device=device)

        item_embeddings = infer_items(model=model)

    item_embeddings = Embeddings(
        ids=torch.arange(start=0, end=item_embeddings.shape[0], device=device), embeddings=item_embeddings
    )
    user_embeddings = Embeddings(ids=user_ids, embeddings=user_embeddings)

    df_user_ids = torch.tensor(eval_df['uid'].to_list(), dtype=torch.long, device=device)
    df_target_ids = [
        torch.tensor(item_ids, dtype=torch.long, device=device) for item_ids in eval_df['item_id_valid'].to_list()
    ]
    targets = Targets(user_ids=df_user_ids, item_ids=df_target_ids)
    with torch.no_grad():
        ranked = rank_items(users=user_embeddings, items=item_embeddings, num_items=100)

    metric_names = [f'{name}@{k}' for name in ["recall", "ndcg", "coverage"] for k in [10, 50, 100]]
    metrics = calc_metrics(ranked, targets, metrics=metric_names)
    print(metrics)


if __name__ == '__main__':
    main()