search

yan-gao-GY / Flower-SpeechBrain

11 stars 7 forks source link

Error When Starting the client #3

Open mnabihali opened 1 year ago

mnabihali commented 1 year ago

Dear thanks for sharing your code. I am trying to run one experiment based on your readme instructions as follows:

python server.py --data_path /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/ --tr_path /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/train.tsv --test_path /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/test.tsv --weight_strategy num

Then

python3 client.py --cid 0 --server_address 127.0.1.1 --data_path /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/ --tr_path /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/train.tsv --dev_path /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/dev.tsv I got this error

FileNotFoundError: /home/mnabih/Flower-SpeechBrain/it/cv-corpus-6.1-2020-12-11/it/train.tsvtrain_0.tsv doesn't exist, verify your dataset!

Can you help why this could be happen?

yan-gao-GY commented 1 year ago

Hi, it seems that the .tsv files for clients are missing. Before training, we need to do FL data partitioning to create the .tsv file for each client. The original .tsv file provided by Common Voice dataset can only be used for centralised training.

mnabihali commented 1 year ago

So I should do it manually split the data between all clients?

yan-gao-GY commented 1 year ago

You could write a script to randomly split the original train.tsv file into separate partitions based on your preferred FL settings.