Update examples_autotm_fit_predict.py

examples/examples_autotm_fit_predict.py

@@ -15,11 +15,16 @@
 )
 logger = logging.getLogger()
 
+# "base" configuration is an example of default configuration of the algorithm
+# "base_en" is showing how to specify dataset parameters and pass it directly to the algorithm
+# "static_chromosome" is a previous version of optimization algorithm, may work faster comparing to the "pipeline" version, but results are slightly worse
+# "surrogate" is showing how to start training with surrogates option (increase optimization speed)
+# "llm" is an example of how to run optimization with gpt-base quality function
+# "bayes" is an option with Bayesian Optimization
 CONFIGURATIONS = {
  "base": {
  "alg_name": "ga",
- "num_iterations": 2,
- "num_individuals": 2,
+ "num_iterations": 50,
  "use_pipeline": True
  },
  "base_en": {
@@ -29,33 +34,29 @@
  "dataset_path": "data/sample_corpora/imdb_100.csv",
  "dataset_name": "imdb_100"
  },
- "num_iterations": 2,
- "num_individuals": 2,
+ "num_iterations": 20,
  "use_pipeline": True
  },
  "static_chromosome": {
  "alg_name": "ga",
- "num_iterations": 2,
- "num_individuals": 2,
+ "num_iterations": 20,
  "use_pipeline": False
  },
  "surrogate": {
  "alg_name": "ga",
- "num_iterations": 2,
- "num_individuals": 2,
+ "num_iterations": 20,
  "use_pipeline": True,
  "surrogate_name": "random-forest-regressor"
  },
  "llm": {
  "alg_name": "ga",
- "num_iterations": 2,
- "num_individuals": 2,
+ "num_iterations": 20,
  "use_pipeline": True,
  "individual_type": "llm"
  },
  "bayes": {
  "alg_name": "bayes",
- "num_evaluations": 5,
+ "num_evaluations": 150,
  }
 }
 
@@ -78,7 +79,6 @@ def run(alg_name: str, alg_params: Dict[str, Any], dataset: Optional[Dict[str, A
  topic_count=20,
  preprocessing_params={
  "lang": dataset['lang'],
- "min_tokens_count": 3
  },
  alg_name=alg_name,
  alg_params=alg_params,