the test infer using opencpop dataset isnot working
I just follow all the steps in docs SVS-opencpop-e2e.md. All the data are as following: . |--data |--raw |--opencpop |--segments |--transcriptions.txt |--wavs |--checkpoints |--MY_DS_EXP_NAME (optional) |--0109_hifigan_bigpopcs_hop128 (vocoder) |--model_ckpt_steps_1512000.ckpt |--config.yaml
- the first error is "Run the following scripts to pack the dataset for training/inference." . the error log shows that the popcs folder is missing? but the test dataset i used is opencpop, so i manualy set the "raw_data_dir" by "opencpop data folder" in binarize.py and modified the "test_prefixes" to ['2044', 2086] in 0228_opencpop_ds100_rel's config.yaml. Then the packing command ran successfully.
- the second error is "Inference from packed test set". The error log shows that the test set batch num is zero. I want to debug the dataloader, and see where the test folder is. However the code is too complex.
So I wanna konw if i should download the popcs dataset to infer with SVS-opencpop-e2e? if i want to use the opencpop test data to infer, which configs should be modified?
Maybe you could try to confirm whether the raw folder is inside the data folder, because the raw_data_dir in binarize.py will not be synced to the whole project, so other parts of the program may not be able to access the folder wanted.