Parameter Optimization

[tripledoubleE]

I just started using the Recbole library. I want to do hyper parameter optimization on AutoInt which is a context-aware model. I did a tutorial before and I didn't keep the parameters in a .yaml file, but instead with the dictionary I named parameter_dict = {...} as follows:

parameter_dict = {

'data_path': 'bla bla',
'USER_ID_FIELD': 'bla bla',
'ITEM_ID_FIELD': 'bla bla',
'TIME_FIELD': 'bla bla',
'load_col': {'inter': [bla bla], 
            'user': [bla bla],
            'item': [bla bla]
},
'epochs': 50,
'eval_args': {
    'split': {'RS': [4, 1, 1]},
    'group_by': 'user',
    'order': 'TO',
    'mode': 'full'},
'metrics': ['Recall', 'MRR', 'NDCG', 'Hit', 'Precision', 'MAP'],
'valid_metric': 'MAP@12',
'field_separator':',',
"mlp_hidden_size": "[128,128]", # Default
"train_neg_sample_args":{"uniform": 1},
"user_inter_num_interval": "[2,inf)", 
"item_inter_num_interval": "[2,inf)"

}

I created Config by writing config = Config(model='AutoInt', dataset='recbole_format', config_dict=parameter_dict) code and I achieved to train, evaluate and predict from the model.

Similarly, I want to make parameter tuning using the parameter dictionary I created as follows:

Note: I have also model.hyper file wich includes:
learning_rate loguniform -8,0
embedding_size choice [64,96,128]
mlp_hidden_size choice ['[64,64,64]','[128,128]']

from recbole.trainer import HyperTuning
from recbole.quick_start import objective_function

hp = HyperTuning(objective_function=objective_function, algo='exhaustive', early_stop=10,
max_evals=100, params_file='model.hyper', fixed_config_file_list= parameter_dict)

hp.run()

But I got this error:

---->
hp.run()

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/recbole/trainer/hyper_tuning.py:411, in HyperTuning.run(self)
408 r"""begin to search the best parameters"""
409 from hyperopt import fmin
--> 411 fmin(
412 self.trial,
413 self.space,
414 algo=self.algo,
415 max_evals=self.max_evals,
416 early_stop_fn=self.early_stop_fn,
417 )
418 if self.display_file is not None:
419 self.plot_hyper()

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/hyperopt/fmin.py:553, in fmin(fn, space, algo, max_evals, timeout, loss_threshold, trials, rstate, allow_trials_fmin, pass_expr_memo_ctrl, catch_eval_exceptions, verbose, return_argmin, points_to_evaluate, max_queue_len, show_progressbar, early_stop_fn, trials_save_file)
550 rval.catch_eval_exceptions = catch_eval_exceptions
552 # next line is where the fmin is actually executed
--> 553 rval.exhaust()
555 if return_argmin:
556 if len(trials.trials) == 0:

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/hyperopt/fmin.py:356, in FMinIter.exhaust(self)
354 def exhaust(self):
355 n_done = len(self.trials)
--> 356 self.run(self.max_evals - n_done, block_until_done=self.asynchronous)
357 self.trials.refresh()
358 return self

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/hyperopt/fmin.py:292, in FMinIter.run(self, N, block_until_done)
289 time.sleep(self.poll_interval_secs)
290 else:
291 # -- loop over trials and do the jobs directly
--> 292 self.serial_evaluate()
294 self.trials.refresh()
295 if self.trials_save_file != "":

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/hyperopt/fmin.py:170, in FMinIter.serial_evaluate(self, N)
168 ctrl = base.Ctrl(self.trials, current_trial=trial)
169 try:
--> 170 result = self.domain.evaluate(spec, ctrl)
171 except Exception as e:
172 logger.error("job exception: %s" % str(e))

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/hyperopt/base.py:907, in Domain.evaluate(self, config, ctrl, attach_attachments)
898 else:
899 # -- the "work" of evaluating config can be written
900 # either into the pyll part (self.expr)
901 # or the normal Python part (self.fn)
902 pyll_rval = pyll.rec_eval(
903 self.expr,
904 memo=memo,
905 print_node_on_error=self.rec_eval_print_node_on_error,
906 )
--> 907 rval = self.fn(pyll_rval)
909 if isinstance(rval, (float, int, np.number)):
910 dict_rval = {"loss": float(rval), "status": STATUS_OK}

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/recbole/trainer/hyper_tuning.py:346, in HyperTuning.trial(self, params)
344 self.params_list.append(params_str)
345 print("running parameters:", config_dict)
--> 346 result_dict = self.objective_function(config_dict, self.fixed_config_file_list)
347 self.params2result[params_str] = result_dict
348 model, score, bigger = (
349 result_dict["model"],
350 result_dict["best_valid_score"],
351 result_dict["valid_score_bigger"],
352 )

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/recbole/quick_start/quick_start.py:133, in objective_function(config_dict, config_file_list, saved)
124 def objective_function(config_dict=None, config_file_list=None, saved=True):
125 r"""The default objective_function used in HyperTuning
126
127 Args:
(...)
130 saved (bool, optional): Whether to save the model. Defaults to True.
131 """
--> 133 config = Config(config_dict=config_dict, config_file_list=config_file_list)
134 init_seed(config["seed"], config["reproducibility"])
135 logger = getLogger()

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/recbole/config/configurator.py:81, in Config.init(self, model, dataset, config_file_list, config_dict)
79 self._init_parameters_category()
80 self.yaml_loader = self._build_yaml_loader()
---> 81 self.file_config_dict = self._load_config_files(config_file_list)
82 self.variable_config_dict = self._load_variable_config_dict(config_dict)
83 self.cmd_config_dict = self._load_cmd_line()

File ~/anaconda3/envs/RecBole/lib/python3.8/site-packages/recbole/config/configurator.py:150, in Config._load_config_files(self, file_list)
148 if file_list:
149 for file in file_list:
--> 150 with open(file, "r", encoding="utf-8") as f:
151 file_config_dict.update(
152 yaml.load(f.read(), Loader=self.yaml_loader)
153 )
154 return file_config_dict

FileNotFoundError: [Errno 2] No such file or directory: 'data_path'

Even if I have data_path in my parameter dictionary, do I need ho have a file or dictionary named "data_path"

I would be very grateful if you could help with the issue :)