'CosyVoiceModel' object has no attribute 'inference'

[doryashar]

trying to run training (CosyVoice/examples/libritts/cosyvoice/run.sh), while doing the inference step i get this error: 'CosyVoiceModel' object has no attribute 'inference' when looking inside inference.py i see: model = CosyVoiceModel(configs['llm'], configs['flow'], configs['hift']) while in cli/model.py there is the CosyVoiceModel class without inference method.

[osmankrblt]

Which model did you use?

[doryashar]

i don't see any specific model that you are referring to, if you are talking about the pretrained_model then i use the cosyvoice-300m

inference

if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then echo "Run inference. Please make sure utt in tts_text is in prompt_data" for mode in sft zero_shot; do python cosyvoice/bin/inference.py --mode $mode \ --gpu 0 \ --config conf/cosyvoice.yaml \ --prompt_data data/test-clean/parquet/data.list \ --prompt_utt2data data/test-clean/parquet/utt2data.list \ --tts_text pwd/tts_text.json \ --llm_model $pretrained_model_dir/llm.pt \ --flow_model $pretrained_model_dir/flow.pt \ --hifigan_model $pretrained_model_dir/hift.pt \ --result_dir pwd/exp/cosyvoice/test-clean/$mode done fi

[doryashar]

actually i don't even see the model.train function anywhere in this repo. is that expected?

[osmankrblt]

I couldn't find model.train too . I understand that the transition between train and inference modes is done differently. Would you try other variants of CosyVoice-300m for your problem?

[doryashar]

i would expect to see the backprop/train function inside the model definition.

[osmankrblt]

I think If I remember right. These are in executor.py

[aluminumbox]

change model.inference to model.tts, we will fix it

[github-actions[bot]]

This issue is stale because it has been open for 30 days with no activity.