Uses preprocess_cache.py + train_lora_cached.py - pre-computes encodings once. Much faster training - no VAE/text encoding during training Easy to experiment with different hyperparameters Cached data ...
def filter_by_session_len(df, min_session_len=2): df_session_item_count = (df.groupby(SESSION_ID_FIELD, as_index=False)[ITEM_ID_FILED].count().rename(columns={ITEM_ID ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results