Help message improvement (#1090)

tools/who_what_benchmark/whowhatbench/wwb.py

@@ -199,61 +199,61 @@ def parse_args():
     parser.add_argument(
         "--base-model",
         default=None,
-        help="Model to ground truth generation.",
+        help="Model for ground truth generation.",
     )
     parser.add_argument(
         "--target-model",
         default=None,
-        help="Model to comparison with base_model. Usually it is compressed, quantized version of base_model.",
+        help="Model to compare against the base_model. Usually it is compressed, quantized version of base_model.",
     )
     parser.add_argument(
         "--tokenizer",
         default=None,
-        help="Tokenizer for divergency metric. If not defined then will be load from base_model or target_model.",
+        help="Tokenizer for divergency metric. If not provided, it will be load from base_model or target_model.",
     )
 
     parser.add_argument(
         "--gt-data",
         default=None,
-        help="CSV file with base_model generation. If defined and exists then base_model will not used."
-        "I defined and not exists them will be generated by base_model evaluation.",
+        help="CSV file containing GT outputs from base_model. If defined and exists then base_model will not used."
+        " If the files does not exist, it will be generated by base_model evaluation.",
     )
     parser.add_argument(
         "--model-type",
         type=str,
         choices=["text", "text-to-image"],
         default="text",
-        help="Indicated the model type, e.g. 'text' - for causal text generation, 'text-to-image' - for image generation.",
+        help="Indicated the model type: 'text' - for causal text generation, 'text-to-image' - for image generation.",
     )
     parser.add_argument(
         "--data-encoder",
         type=str,
         default="sentence-transformers/all-mpnet-base-v2",
         help="Model for measurement of similarity between base_model and target_model."
-        "By default it is sentence-transformers/all-mpnet-base-v2,"
-        "but for Chinese LLMs better to use sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2.",
+        " By default it is sentence-transformers/all-mpnet-base-v2,"
+        " but for Chinese LLMs, better to use sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2.",
     )
     parser.add_argument(
         "--dataset",
         type=str,
         default=None,
         help="Name of the dataset with prompts. The interface for dataset is load_dataset from datasets library."
-        "Please provide this argument in format path,name (for example wikitext,wikitext-2-v1)."
-        "If None then internal list of prompts will be used.",
+        " Please provide this argument in format path,name (for example wikitext,wikitext-2-v1)."
+        " If None then internal list of prompts will be used.",
     )
     parser.add_argument(
         "--dataset-field",
         type=str,
         default="text",
         help="The name of field in dataset for prompts. For example question or context in squad."
-        "Will be used only if dataset is defined.",
+        " Will be used only if dataset is defined.",
     )
     parser.add_argument(
         "--split",
         type=str,
         default=None,
         help="Split of prompts from dataset (for example train, validation, train[:32])."
-        "Will be used only if dataset is defined.",
+        " Will be used only if dataset is defined.",
     )
     parser.add_argument(
         "--output",