Update mimicit_format.md

docs/mimicit_format.md

@@ -1,5 +1,7 @@
 # Breaking Down the MIMIC-IT Format
 
+❗❗❗We changed previous `images.json` to `images.parquet`. They are all containing multiple `key:base64` pairs but the later one would consume far less CPU memory and faster during loading with `pandas.Dataframe`. It enables us to train with larger datasets more conviently.
+
 We mainly use one integrate dataset format and we refer it to MIMIC-IT format since.
 
 The mimic-it format contains the following data yaml file. Within this data yaml file, you could assign the path of the instruction json file and the image parquet file, and also the number of samples you want to use. The number of samples within each group will be uniformly sampled, and the `number_samples / total_numbers`` will decide sampling ratio of each dataset.
@@ -86,4 +88,4 @@ parquet_file_path = os.path.join(
     parquet_root_path, os.path.basename(json_file_path).split(".")[0].replace("_image", "") + ".parquet"
 )
 df.to_parquet(parquet_file_path, engine="pyarrow")
-```
+```