TTS working but model would give error while voice conversion

[tieincred]

Hello, Please answer this I have been stuck for a while.
When I am trying to inference from my trained model the inference is being done and audio is saved as it should be, but when I am trying to do voice conversion with following command:
tts --model_path output_multilingual//checkpoint_20000.pth --config_path output_multilingual/config.json --reference_wav target_content/asura_10secs.wav --language_idx hi --speaker_idx yayati
I am receiving the following error.
RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int; but got torch.FloatTensor instead (while checking arguments for embedding)
I tried to check what is going wrong but since I was able to inference for text to speech with the same model and config I am unable to figure out what might be causing the issue.
I have used speaker_embedding and not the d_vectors
Below I have attached full error log.