def _test8_serialization(self, cls, opt):
set_log_level(1)
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = cls(opt, data_opt=data_opt)
c.initialize()
c.train()
ret_a = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_a)
c.save('model.bin')
c.load('model.bin')
os.remove('model.bin')
ret_a = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_a)
def _test9_compact_serialization(self, cls, opt):
set_log_level(1)
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = cls(opt, data_opt=data_opt)
c.initialize()
c.train()
ret_a = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_a)
c.save('model.bin', with_userid_map=False)
c = cls(opt)
c.load('model.bin', data_fields=['Q', '_idmanager'])
ret_a = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_a)
self.assertFalse(hasattr(c, 'P'))
c.normalize(group='item')
ret_a = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_a)
def test10_fast_most_similar(self):
set_log_level(1)
opt = CFROption().get_default_option()
data_opt = StreamOptions().get_default_option()
data_opt.data.sppmi = {"windows": 5, "k": 10}
data_opt.data.internal_data_type = "matrix"
data_opt.input.main = self.ml_100k + 'stream'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = CFR(opt, data_opt=data_opt)
c.initialize()
c.train()
keys = [x for x, _ in c.most_similar('49.Star_Wars_(1977)', 10)]
start_t = time.time()
for i in range(100):
for key in keys:
c.most_similar(key)
elapsed_a = time.time() - start_t
c.normalize(group='item')
start_t = time.time()
for i in range(100):
for key in keys:
c.most_similar(key)
elapsed_b = time.time() - start_t
self.assertTrue(elapsed_a > elapsed_b)
def _test6_topk(self, cls, opt):
set_log_level(2)
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = cls(opt, data_opt=data_opt)
c.initialize()
c.train()
self.assertTrue(len(c.topk_recommendation('1', 10)), 10)
ret_a = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_a)
c.normalize()
ret_b = [
x for x, _ in c.most_similar('180.Return_of_the_Jedi_(1983)',
topk=100)
]
self.assertIn('49.Star_Wars_(1977)', ret_b)
self.assertEqual(ret_a[:10], ret_b[:10])
def test08_serialization(self):
opt = BPRMFOption().get_default_option()
opt.num_iters = 200
opt.d = 5
opt.validation = aux.Option({'topk': 10})
self._test8_serialization(BPRMF, opt)
def get_default_option(self) -> aux.Option:
opt = {
'type': 'stream',
'input': {
'main': '',
'uid': '', # if not set, row-id is used as userid.
'iid': '' # if not set, col-id is used as userid.
},
'data': {
'validation': {
'name': 'newest', # sample or oldest or newest
'p': 0.01, # if set oldest or newest, ignored
'n': 1, # if set sample, ignored
'max_samples': 500
},
'sppmi': {
# 'windows': 5,
# 'k': 1
},
'batch_mb': 1024,
'use_cache': False,
'tmp_dir': '/tmp/',
'path': './stream.h5py',
'internal_data_type':
'stream' # if set to 'matrix', internal data stored as like matrix market format
}
}
return aux.Option(opt)
def _test10_fast_most_similar(self, cls, opt):
set_log_level(1)
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = cls(opt, data_opt=data_opt)
c.initialize()
c.train()
keys = [x for x, _ in c.most_similar('49.Star_Wars_(1977)', topk=100)]
start_t = time.time()
for i in range(100):
for key in keys:
c.most_similar(key)
elapsed_a = time.time() - start_t
c.normalize(group='item')
start_t = time.time()
for i in range(100):
for key in keys:
c.most_similar(key)
elapsed_b = time.time() - start_t
self.assertTrue(elapsed_a > elapsed_b)
def test09_compact_serialization(self):
set_log_level(1)
opt = CFROption().get_default_option()
data_opt = StreamOptions().get_default_option()
data_opt.data.sppmi = {"windows": 5, "k": 10}
data_opt.data.internal_data_type = "matrix"
data_opt.input.main = self.ml_100k + 'stream'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = CFR(opt, data_opt=data_opt)
c.initialize()
c.train()
ret_a = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_a)
c.save('model.bin', with_userid_map=False)
c = CFR(opt)
c.load('model.bin', data_fields=['I', '_idmanager'])
ret_a = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_a)
self.assertFalse(hasattr(c, 'U'))
c.normalize(group='item')
ret_a = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_a)
def test11_gpu_validation(self):
if not inited_CUBPR:
return
np.random.seed(7)
opt = BPRMFOption().get_default_option()
opt.d = 100
opt.verify_neg = False
opt.accelerator = True
opt.lr = 0.01
opt.reg_b = 10.0
opt.num_iters = 500
opt.evaluation_period = 50
opt.random_seed = 777
opt.validation = aux.Option({'topk': 10})
opt.tensorboard = aux.Option({'root': './tb', 'name': 'bpr'})
self._test5_validation(BPRMF, opt, ndcg=0.03, map=0.02)
def __init__(self, *args, **kwargs):
self._idmanager = aux.Option({
'userid': [],
'userid_map': {},
'itemid': [],
'itemid_map': {},
'userid_mapped': False,
'itemid_mapped': False
})
def test11_train_ml_20m_on_gpu(self):
opt = ALSOption().get_default_option()
opt.num_workers = 8
opt.d = 100
opt.validation = aux.Option({'topk': 10})
opt.compute_loss_on_training = True
opt.accelerator = True
opt.num_cg_max_iters = 3
self._test7_train_ml_20m(ALS, opt)
def run(self, opt_path):
opt = aux.Option(opt_path)
als = _ALS(opt_path)
als.init_factors()
loss = als.train()
self.logger.info(f'ALS finished with loss({loss}).')
if opt.save_factors:
self.logger.info(f'Saving model to {opt.model_path}.')
als.save(opt.model_path)
def test4_optimize(self):
set_log_level(2)
opt = ALSOption().get_default_option()
opt.d = 5
opt.num_workers = 2
opt.model_path = 'als.bin'
opt.validation = aux.Option({'topk': 10})
optimize_option = aux.Option({
'loss': 'val_rmse',
'max_trials': 10,
'deployment': True,
'start_with_default_parameters': True,
'space': {
'd': ['randint', ['d', 10, 20]],
'reg_u': ['uniform', ['reg_u', 0.1, 0.3]],
'reg_i': ['uniform', ['reg_i', 0.1, 0.3]],
'alpha': ['randint', ['alpha', 8, 10]]
}
})
opt.optimize = optimize_option
opt.evaluation_period = 1
opt.tensorboard = aux.Option({'root': './tb',
'name': 'als'})
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
als = ALS(opt, data_opt=data_opt)
als.init_factors()
als.train()
default_result = als.get_validation_results()
als.optimize()
base_loss = default_result['rmse'] # val_rmse
optimize_loss = als.get_optimization_data()['best']['val_rmse']
self.assertTrue(base_loss > optimize_loss)
als.load('als.bin')
loss = als.get_validation_results()
self.assertAlmostEqual(loss['rmse'], optimize_loss)
os.remove('als.bin')
def test00_tensorboard(self):
set_log_level(2)
opt = ALSOption().get_default_option()
opt.d = 5
opt.validation = aux.Option({'topk': 10})
opt.tensorboard = aux.Option({'root': './tb', 'name': 'als'})
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
als = ALS(opt, data_opt=data_opt)
als.initialize()
als.train()
results = als.get_validation_results()
self.assertTrue(results['ndcg'] > 0.025)
self.assertTrue(results['map'] > 0.015)
def test5_validation(self, ndcg=0.06, map=0.04):
set_log_level(3)
opt = CFROption().get_default_option()
opt.validation = aux.Option({'topk': 10})
opt.tensorboard = aux.Option({'root': './tb', 'name': 'cfr'})
data_opt = StreamOptions().get_default_option()
data_opt.data.validation.name = "sample"
data_opt.data.sppmi = {"windows": 5, "k": 10}
data_opt.data.internal_data_type = "matrix"
data_opt.input.main = self.ml_100k + 'stream'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = CFR(opt, data_opt=data_opt)
c.initialize()
c.train()
results = c.get_validation_results()
self.assertTrue(results['ndcg'] > ndcg)
self.assertTrue(results['map'] > map)
def test12_gpu_train_ml_20m(self):
if not inited_CUBPR:
return
opt = BPRMFOption().get_default_option()
opt.accelerator = True
opt.d = 100
opt.verify_neg = False
opt.num_iters = 30
opt.evaluation_period = 5
opt.validation = aux.Option({'topk': 10})
self._test7_train_ml_20m(BPRMF, opt)
def _get_initial_tensorboard_data(self):
tb = aux.Option({'summary_writer': None,
'name': None,
'metrics': {},
'feed_dict': {},
'merged_summary_op': None,
'session': None,
'pbar': None,
'data_root': None,
'step': 1})
return tb
def _test4_train(self, cls, opt):
set_log_level(3)
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = cls(opt, data_opt=data_opt)
c.initialize()
c.train()
self.assertTrue(True)
def __init__(self, opt, *args, **kwargs):
self.opt = aux.Option(opt)
self.tmp_root = opt.data.tmp_dir
if not os.path.isdir(self.tmp_root):
os.makedirs(self.tmp_root)
self.handle = None
self.header = None
self.prepro = prepro.PreProcess(self.opt.data)
self.value_prepro = self.prepro
if self.opt.data.value_prepro:
self.prepro = getattr(prepro, self.opt.data.value_prepro.name)(self.opt.data.value_prepro)
self.value_prepro = self.prepro
self.data_type = None
def test6_topk(self):
set_log_level(1)
opt = CFROption().get_default_option()
opt.validation = aux.Option({'topk': 10})
data_opt = StreamOptions().get_default_option()
data_opt.data.validation.name = "sample"
data_opt.data.sppmi = {"windows": 5, "k": 10}
data_opt.data.internal_data_type = "matrix"
data_opt.input.main = self.ml_100k + 'stream'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = CFR(opt, data_opt=data_opt)
c.initialize()
c.train()
self.assertTrue(len(c.topk_recommendation('1', 10)), 10)
ret_a = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_a)
c.normalize()
ret_b = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_b)
self.assertEqual(ret_a, ret_b)
def test4_train(self):
set_log_level(3)
opt = CFROption().get_default_option()
data_opt = StreamOptions().get_default_option()
data_opt.data.sppmi = {"windows": 5, "k": 10}
data_opt.data.internal_data_type = "matrix"
data_opt.input.main = self.ml_100k + 'stream'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = CFR(opt, data_opt=data_opt)
c.initialize()
c.train()
self.assertTrue(True)
def load(opt):
from buffalo.data.mm import MatrixMarket
from buffalo.data.stream import Stream
if isinstance(opt, (str, )):
opt = aux.Option(opt)
assert isinstance(
opt, (dict, aux.Option
)), 'opt must be either str, or dict/aux.Option but {}'.format(
type(opt))
if opt['type'] == 'matrix_market':
return MatrixMarket(opt)
if opt['type'] == 'stream':
return Stream(opt)
raise RuntimeError('Unexpected data.type: {}'.format(opt['type']))
def _test5_validation(self, cls, opt, ndcg=0.06, map=0.04):
set_log_level(2)
data_opt = MatrixMarketOptions().get_default_option()
data_opt.input.main = self.ml_100k + 'main'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = cls(opt, data_opt=data_opt)
c.initialize()
c.train()
results = c.get_validation_results()
self.assertTrue(results['ndcg'] > ndcg, msg='NDCG Test')
self.assertTrue(results['map'] > map, msg='MAP Test')
def get_default_option(self) -> aux.Option:
opt = {
'type': 'matrix_market',
'input': {
'main': '',
'uid': '', # if not set, row-id is used as userid.
'iid': '' # if not set, col-id is used as itemid.
},
'data': {
'internal_data_type': 'matrix',
'validation': {
'name': 'sample',
'p': 0.01,
'max_samples': 500
},
'batch_mb': 1024,
'use_cache': False,
'tmp_dir': '/tmp/',
'path': './mm.h5py'
}
}
return aux.Option(opt)
def test8_serialization(self):
set_log_level(1)
opt = CFROption().get_default_option()
data_opt = StreamOptions().get_default_option()
data_opt.data.sppmi = {"windows": 5, "k": 10}
data_opt.data.internal_data_type = "matrix"
data_opt.input.main = self.ml_100k + 'stream'
data_opt.input.uid = self.ml_100k + 'uid'
data_opt.input.iid = self.ml_100k + 'iid'
data_opt.data.value_prepro = aux.Option({'name': 'OneBased'})
c = CFR(opt, data_opt=data_opt)
c.initialize()
c.train()
ret_a = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_a)
c.save('model.bin')
c.load('model.bin')
os.remove('model.bin')
ret_a = [x for x, _ in c.most_similar('49.Star_Wars_(1977)')]
self.assertIn('180.Return_of_the_Jedi_(1983)', ret_a)
def get_default_option(self) -> aux.Option:
opt = {
'type': 'matrix_market',
'input': {
'main': '', # str or numpy-kind data
'uid': '', # if not set, row-id is used as userid. It is okay to pass list or 1d dence array as a id list information.
'iid': '' # if not set, col-id is used as itemid. It is okay to pass list or 1d dence array as a id list information.
},
'data': {
'internal_data_type': 'matrix',
'validation': {
'name': 'sample',
'p': 0.01,
'max_samples': 500
},
'batch_mb': 1024,
'use_cache': False,
'tmp_dir': '/tmp/',
'path': './mm.h5py'
}
}
return aux.Option(opt)
def test0_onebased(self):
opt = MatrixMarketOptions().get_default_option()
opt.input.main = self.mm_path
opt.input.uid = self.uid_path
opt.input.iid = self.iid_path
opt.data.value_prepro = aux.Option({'name': 'OneBased'})
mm = MatrixMarket(opt)
mm.create()
self.temp_files.append(opt.data.path)
self.assertTrue(True)
db = mm.handle
self.assertEqual(sorted(db.keys()),
sorted(['vali', 'idmap', 'rowwise', 'colwise']))
header = mm.get_header()
self.assertEqual(header['num_nnz'], 5)
self.assertEqual(header['num_users'], 5)
self.assertEqual(header['num_items'], 3)
data = [(u, kk, vv) for u, kk, vv in mm.iterate()]
self.assertEqual(len(data), 5)
self.assertEqual([int(kk) for _, kk, _ in data], [0, 0, 2, 1, 1])
self.assertEqual([int(vv) for _, _, vv in data], [1, 1, 1, 1, 1])
self.assertEqual(data[2], (2, 2, 1.0))
def test2_implicit_als(self):
opt = MatrixMarketOptions().get_default_option()
opt.input.main = self.mm_path
opt.input.uid = self.uid_path
opt.input.iid = self.iid_path
opt.data.value_prepro = aux.Option({
'name': 'ImplicitALS',
'epsilon': 0.5
})
mm = MatrixMarket(opt)
mm.create()
self.assertTrue(True)
db = mm.handle
self.assertEqual(sorted(db.keys()),
sorted(['vali', 'idmap', 'rowwise', 'colwise']))
header = mm.get_header()
self.assertEqual(header['num_nnz'], 5)
self.assertEqual(header['num_users'], 5)
self.assertEqual(header['num_items'], 3)
data = [(u, kk, vv) for u, kk, vv in mm.iterate()]
self.assertEqual(len(data), 5)
self.assertEqual([int(kk) for _, kk, _ in data], [0, 0, 2, 1, 1])
self.assertAlmostEqual(data[2][2], math.log(1 + 1.0 / 0.5))
def __init__(self, opt_path=None, *args, **kwargs):
Algo.__init__(self, *args, **kwargs)
W2VOption.__init__(self, *args, **kwargs)
Evaluable.__init__(self, *args, **kwargs)
Serializable.__init__(self, *args, **kwargs)
Optimizable.__init__(self, *args, **kwargs)
if opt_path is None:
opt_path = W2VOption().get_default_option()
self.logger = log.get_logger('W2V')
self.opt, self.opt_path = self.get_option(opt_path)
self.obj = CyW2V()
assert self.obj.init(bytes(
self.opt_path, 'utf-8')), 'cannot parse option file: %s' % opt_path
self.data = None
data = kwargs.get('data')
data_opt = self.opt.get('data_opt')
data_opt = kwargs.get('data_opt', data_opt)
if data_opt:
self.data = buffalo.data.load(data_opt)
assert self.data.data_type == 'stream'
self.data.create()
elif isinstance(data, Data):
self.data = data
self.logger.info('W2V(%s)' % json.dumps(self.opt, indent=2))
if self.data:
self.logger.info(self.data.show_info())
assert self.data.data_type in ['stream']
self._vocab = aux.Option({
'size': 0,
'index': None,
'inv_index': None,
'scale': None,
'dist': None,
'total_word_count': 0
})
def example1():
log.set_log_level(log.DEBUG)
als_option = ALSOption().get_default_option()
als_option.validation = aux.Option({'topk': 10})
data_option = MatrixMarketOptions().get_default_option()
data_option.input.main = '../tests/ext/ml-100k/main'
data_option.input.iid = '../tests/ext/ml-100k/iid'
als = ALS(als_option, data_opt=data_option)
als.initialize()
als.train()
print('MovieLens 100k metrics for validations\n%s' %
json.dumps(als.get_validation_results(), indent=2))
print('Similar movies to Star_Wars_(1977)')
for rank, (movie_name,
score) in enumerate(als.most_similar('49.Star_Wars_(1977)')):
print(f'{rank + 1:02d}. {score:.3f} {movie_name}')
print('Run hyper parameter optimization for val_ndcg...')
als.opt.num_workers = 4
als.opt.evaluation_period = 10
als.opt.optimize = aux.Option({
'loss': 'val_ndcg',
'max_trials': 100,
'deployment': True,
'start_with_default_parameters': True,
'space': {
'd': ['randint', ['d', 10, 128]],
'reg_u': ['uniform', ['reg_u', 0.1, 1.0]],
'reg_i': ['uniform', ['reg_i', 0.1, 1.0]],
'alpha': ['randint', ['alpha', 1, 10]],
}
})
log.set_log_level(log.INFO)
als.opt.model_path = './example1.ml100k.als.optimize.bin'
print(
json.dumps(
{
'alpha': als.opt.alpha,
'd': als.opt.d,
'reg_u': als.opt.reg_u,
'reg_i': als.opt.reg_i
},
indent=2))
als.optimize()
als.load('./example1.ml100k.als.optimize.bin')
print('Similar movies to Star_Wars_(1977)')
for rank, (movie_name,
score) in enumerate(als.most_similar('49.Star_Wars_(1977)')):
print(f'{rank + 1:02d}. {score:.3f} {movie_name}')
optimization_res = als.get_optimization_data()
best_parameters = optimization_res['best_parameters']
print(json.dumps(optimization_res['best'], indent=2))
print(
json.dumps(
{
'alpha': best_parameters['alpha'],
'd': best_parameters['d'],
'reg_u': best_parameters['reg_u'],
'reg_i': best_parameters['reg_i']
},
indent=2))