adbailey1
commented
4 years ago Hi! Thanks for checking out the repo and for finding this bug. I actually fixed this issue in one of my other repos LINK, before I get round to updating and closing this issue, the following code should be enough:
if not os.path.exists(config.COMP_DATASET_PATH):
if not os.path.exists(config.FULL_TRAIN_SPLIT_PATH):
utilities.merge_csv(config.TRAIN_SPLIT_PATH, config.DEV_SPLIT_PATH, config.FULL_TRAIN_SPLIT_PATH)
utilities.merge_csv(config.FULL_TRAIN_SPLIT_PATH, config.TEST_SPLIT_PATH, config.COMP_DATASET_PATH)where config.FULL_TRAIN_SPLIT_PATH and config.COMP_DATASET_PATH are:
FULL_TRAIN_SPLIT_PATH = os.path.join(DATASET, 'full_train_split_Depression_AVEC2017.csv')
COMP_DATASET_PATH = os.path.join(DATASET, 'complete_Depression_AVEC2017.csv')and where utilities.merge_csv() is:
def merge_csv(path, path2, filename):
a = pd.read_csv(path)
b = pd.read_csv(path2)
columnsa = list(a)
columnsb = list(b)
# The test split doesn't have the same columns as the train/dev
# Create these extra columns and fill them with -1
if len(columnsa) > len(columnsb):
difference = len(columnsa) - len(columnsb)
names = columnsa[-difference:]
h, _ = b.shape
zeros = [-1] * h
for i in range(difference):
b[names[i]] = zeros
columnsb = list(b)
# This checks that the column headers are the same in the two CSV files
for i in range(len(columnsa)):
# If headers are different, re-name
if columnsa[i] != columnsb[i]:
b = b.rename(columns={columnsb[i]: columnsa[i]})
# Create a single dataframe from the 2 CSV files and save
dataframes = [a, b]
c = pd.concat(dataframes)
c = c.sort_values(by=['Participant_ID'])
c.to_csv(filename, index=False)```
If this doesn't work or there are any other problems let me know. I'll hopefully get around to updating this during the week. A cleaner and simpler repro than this one is [HERE](https://github.com/adbailey1/DepAudioNet), again I need to clean up the code but this is the most recent version I have been working with for experimentation. 
pengsiyu2019102942
Hello, I recently read the code you wrote. I want to ask you a question. The data set I downloaded is a bit different from yours. I am missing the full_train_split_Depression_AVEC2017.csv file and the complete_Depression_AVEC2017.csv file. I don’t know if you are willing to share yours. Thank you for the link to the network disk of the data set! I wish you all the best!