def do_prepare(opts):
name = opts['-d']
ml = MovieLens(f'data/{name}')
train, test = next(sample_users(ml.ratings, 1, 10000, SampleN(5)))
test.to_parquet(f'data/{name}-test.parquet', index=False)
_log.info('getting popular recs')
pop = Popular()
pop.fit(train)
pop_recs = recommend(pop, test['user'].unique(), 100)
_log.info('getting ALS recs')
als = ImplicitMF(20, iterations=10)
als = Recommender.adapt(als)
als.fit(train.drop(columns=['rating']))
als_recs = recommend(als, test['user'].unique(), 100)
_log.info('merging recs')
recs = pd.concat({
'Popular': pop_recs,
'ALS': als_recs
},
names=['Algorithm'])
recs.reset_index('Algorithm', inplace=True)
recs.to_parquet(f'data/{name}-recs.parquet', index=False)
def test_pop_recommend(ml20m, rng, n_jobs):
users = rng.choice(ml20m['user'].unique(), 10000, replace=False)
algo = Popular()
_log.info('training %s', algo)
algo.fit(ml20m)
_log.info('recommending with %s', algo)
recs = batch.recommend(algo, users, 10, n_jobs=n_jobs)
assert recs['user'].nunique() == 10000
def test_store_save(store_cls):
algo = Popular()
algo.fit(lktu.ml_test.ratings)
with store_cls() as store:
k = store.put_model(algo)
a2 = store.get_model(k)
assert a2 is not algo
assert a2.item_pop_ is not algo.item_pop_
assert all(a2.item_pop_ == algo.item_pop_)
del a2
def test_store_client_pickle(store_cls):
algo = Popular()
algo.fit(lktu.ml_test.ratings)
with store_cls() as store:
k = store.put_model(algo)
client = store.client()
client = pickle.loads(pickle.dumps(client))
k = pickle.loads(pickle.dumps(k))
a2 = client.get_model(k)
assert a2 is not algo
assert a2.item_pop_ is not algo.item_pop_
assert all(a2.item_pop_ == algo.item_pop_)
del a2
def test_sweep_filenames(tmp_path):
tmp_path = norm_path(tmp_path)
work = pathlib.Path(tmp_path)
sweep = batch.MultiEval(tmp_path)
ratings = ml_pandas.renamed.ratings
folds = []
for part, (train, test) in enumerate(xf.partition_users(ratings, 2, xf.SampleN(5))):
trfn = work / 'p{}-train.csv'.format(part)
tefn = work / 'p{}-test.csv'.format(part)
train.to_csv(trfn)
test.to_csv(tefn)
folds.append((trfn, tefn))
sweep.add_datasets(folds, DataSet='ml-small')
sweep.add_algorithms([Bias(damping=0), Bias(damping=5), Bias(damping=10)],
attrs=['damping'])
sweep.add_algorithms(Popular())
try:
sweep.run()
finally:
if (work / 'runs.csv').exists():
runs = pd.read_csv(work / 'runs.csv')
print(runs)
assert (work / 'runs.csv').exists()
assert (work / 'runs.parquet').exists()
assert (work / 'predictions.parquet').exists()
assert (work / 'recommendations.parquet').exists()
runs = pd.read_parquet(work / 'runs.parquet')
# 4 algorithms by 2 partitions
assert len(runs) == 8
def test_sweep_norecs(tmp_path):
work = pathlib.Path(tmp_path)
sweep = batch.MultiEval(tmp_path, recommend=None)
ratings = ml_test.ratings
folds = xf.partition_users(ratings, 5, xf.SampleN(5))
sweep.add_datasets(folds, DataSet='ml-small')
sweep.add_algorithms([Bias(damping=0), Bias(damping=5), Bias(damping=10)],
attrs=['damping'])
sweep.add_algorithms(Popular())
try:
sweep.run()
finally:
if (work / 'runs.csv').exists():
runs = pd.read_csv(work / 'runs.csv')
print(runs)
assert (work / 'runs.csv').exists()
assert (work / 'runs.parquet').exists()
assert (work / 'predictions.parquet').exists()
assert not (work / 'recommendations.parquet').exists()
runs = pd.read_parquet(work / 'runs.parquet')
# 4 algorithms by 5 partitions
assert len(runs) == 20
assert all(np.sort(runs.AlgoClass.unique()) == ['Bias', 'Popular'])
bias_runs = runs[runs.AlgoClass == 'Bias']
assert all(bias_runs.damping.notna())
pop_runs = runs[runs.AlgoClass == 'Popular']
assert all(pop_runs.damping.isna())
preds = pd.read_parquet(work / 'predictions.parquet')
assert all(preds.RunId.isin(bias_runs.RunId))
def test_sweep_nopreds(tmp_path):
work = pathlib.Path(tmp_path)
sweep = batch.MultiEval(tmp_path, eval_n_jobs=1)
ratings = ml_test.ratings
folds = [(train, test.drop(columns=['rating']))
for (train, test) in xf.partition_users(ratings, 5, xf.SampleN(5))
]
sweep.add_datasets(folds, DataSet='ml-small')
sweep.add_algorithms(Popular())
sweep.add_algorithms(Bias(damping=0))
try:
sweep.run()
finally:
if (work / 'runs.csv').exists():
runs = pd.read_csv(work / 'runs.csv')
print(runs)
assert (work / 'runs.csv').exists()
assert (work / 'runs.parquet').exists()
assert not (work / 'predictions.parquet').exists()
assert (work / 'recommendations.parquet').exists()
runs = pd.read_parquet(work / 'runs.parquet')
# 2 algorithms by 5 partitions
assert len(runs) == 10
assert all(np.sort(runs.AlgoClass.unique()) == ['Bias', 'Popular'])
bias_runs = runs[runs.AlgoClass == 'Bias']
recs = pd.read_parquet(work / 'recommendations.parquet')
assert all(recs.RunId.isin(runs.RunId))
assert recs['score'].dtype == np.float64
def test_sweep_combine(tmp_path):
tmp_path = norm_path(tmp_path)
work = pathlib.Path(tmp_path)
sweep = batch.MultiEval(tmp_path, combine=False)
ratings = ml_pandas.renamed.ratings
sweep.add_datasets(lambda: xf.partition_users(ratings, 5, xf.SampleN(5)), name='ml-small')
sweep.add_algorithms([Bias(damping=0), Bias(damping=5)],
attrs=['damping'])
sweep.add_algorithms(Popular())
sweep.persist_data()
for i in range(1, 6):
assert (work / 'ds{}-train.parquet'.format(i)).exists()
assert (work / 'ds{}-test.parquet'.format(i)).exists()
for ds, cf, dsa in sweep.datasets:
assert isinstance(ds, tuple)
train, test = ds
assert isinstance(train, pathlib.Path)
assert isinstance(test, pathlib.Path)
assert sweep.run_count() == 5 * 3
try:
sweep.run()
finally:
if (work / 'runs.csv').exists():
runs = pd.read_csv(work / 'runs.csv')
print(runs)
assert not (work / 'runs.csv').exists()
assert not (work / 'runs.parquet').exists()
assert not (work / 'predictions.parquet').exists()
assert not (work / 'recommendations.parquet').exists()
for i, (ds, a) in enumerate(sweep._flat_runs()):
run = i + 1
assert (work / 'run-{}.json'.format(run)).exists()
if isinstance(a.algorithm, Predictor):
assert (work / 'predictions-{}.parquet'.format(run)).exists()
assert (work / 'recommendations-{}.parquet'.format(run)).exists()
sweep.collect_results()
assert (work / 'runs.csv').exists()
assert (work / 'runs.parquet').exists()
assert (work / 'predictions.parquet').exists()
assert (work / 'recommendations.parquet').exists()
runs = pd.read_parquet(work / 'runs.parquet')
assert len(runs) == 5 * 3
def run(self, strategy_context: RecommenderAlgorithmStrategyContext
) -> np.ndarray:
data_set_source = strategy_context.data_set_source
data_frame_reader: DataFrameReaderStrategy = self.data_frame_reader_factory.create(
data_set_source)
data_set: DataFrame = data_frame_reader.parse(
DataFrameReaderStrategyContext(data_set_source))
partition = list(
partition_users(data=data_set,
partitions=1,
method=crossfold.SampleFrac(0.2)))[0]
test, train = partition.test, partition.train
number_of_recommendations = strategy_context.number_of_recommendations
algorithm = Popular()
trained_algorithm = algorithm.fit(train)
recommendations = lenskit.batch.recommend(trained_algorithm,
test['user'].unique(),
number_of_recommendations)
return recommendations.groupby('user')['item'].apply(
lambda x: x).to_numpy().reshape((-1, number_of_recommendations))
def test_save_models(tmp_path, format):
tmp_path = norm_path(tmp_path)
work = pathlib.Path(tmp_path)
sweep = batch.MultiEval(tmp_path, save_models=format)
sweep.add_algorithms(Bias(5))
sweep.add_algorithms(Popular())
ratings = ml_test.ratings
sweep.add_datasets(lambda: xf.sample_users(ratings, 2, 100, xf.SampleN(5)),
name='ml-small')
sweep.run()
runs = pd.read_parquet(fspath(tmp_path / 'runs.parquet'))
runs = runs.set_index('RunId')
for i in range(4):
run_id = i + 1
fn = work / 'model-{}'.format(run_id)
if format is True:
fn = fn.with_suffix('.pkl')
assert fn.exists()
with fn.open('rb') as f:
algo = pickle.load(f)
elif format == 'gzip':
fn = fn.with_suffix('.pkl.gz')
assert fn.exists()
with gzip.open(fspath(fn), 'rb') as f:
algo = pickle.load(f)
elif format == 'joblib':
fn = fn.with_suffix('.jlpkl')
assert fn.exists()
algo = joblib.load(fn)
else:
assert False
assert algo is not None
algo_class = algo.__class__.__name__
if isinstance(algo, TopN):
algo_class = algo.predictor.__class__.__name__
assert algo_class == runs.loc[run_id, 'AlgoClass']
def test_sweep_persist(tmp_path):
tmp_path = norm_path(tmp_path)
work = pathlib.Path(tmp_path)
sweep = batch.MultiEval(tmp_path)
ratings = ml_pandas.renamed.ratings
sweep.add_datasets(lambda: xf.partition_users(ratings, 5, xf.SampleN(5)),
name='ml-small')
sweep.persist_data()
for i in range(1, 6):
assert (work / 'ds{}-train.parquet'.format(i)).exists()
assert (work / 'ds{}-test.parquet'.format(i)).exists()
for ds, cf, dsa in sweep.datasets:
assert isinstance(ds, tuple)
train, test = ds
assert isinstance(train, pathlib.Path)
assert isinstance(test, pathlib.Path)
sweep.add_algorithms(
[Bias(damping=0), Bias(damping=5),
Bias(damping=10)],
attrs=['damping'])
sweep.add_algorithms(Popular())
try:
sweep.run()
finally:
if (work / 'runs.csv').exists():
runs = pd.read_csv(work / 'runs.csv')
print(runs)
assert (work / 'runs.csv').exists()
assert (work / 'runs.parquet').exists()
assert (work / 'predictions.parquet').exists()
assert (work / 'recommendations.parquet').exists()
runs = pd.read_parquet(work / 'runs.parquet')
# 4 algorithms by 5 partitions
assert len(runs) == 20
def test_pop_batch_recommend(ml_folds: MLFolds, ncpus):
algo = Popular()
recs = ml_folds.eval_all(algo, nprocs=ncpus)
ml_folds.check_positive_ndcg(recs)
