Search results

conditional time masking ratio can be reduced while training?

Status: Open.

#185 In k2-fsa/ZipVoice;

· Imesh7 opened on Apr 18, 2026

grad is not finite

Status: Open.

#184 In k2-fsa/ZipVoice;

· stormand opened on Apr 8, 2026

Absolute loss decreasing but reference loss increasing in stage-1of distillation training using custom dataset

Status: Open.

#183 In k2-fsa/ZipVoice;

· mukherjeesougata-eros opened on Mar 30, 2026

Tokenizer to be used for Multilingual finetuning on custom data

Status: Open.

#182 In k2-fsa/ZipVoice;

· mukherjeesougata-eros opened on Mar 27, 2026

When training on libritts, the speech generated by inference after training is garbled.

Status: Open.

#181 In k2-fsa/ZipVoice;

· 99-song opened on Mar 27, 2026

Training on a multi language

Status: Open.

#180 In k2-fsa/ZipVoice;

· marius-verdier opened on Mar 26, 2026

fixed speaker synthesis without prompt

Status: Open.

#179 In k2-fsa/ZipVoice;

· HYJ-creator opened on Mar 25, 2026

Finetuning after training from scratch

Status: Open.

#178 In k2-fsa/ZipVoice;

· Alekksander66 opened on Mar 6, 2026

Clarification on Checkpoint Averaging and ONNX Export in run_emilia.sh vs run_libritts.sh

Status: Open.

#177 In k2-fsa/ZipVoice;

· mukherjeesougata-eros opened on Feb 23, 2026

Experimental setting/setup guidance required for training distill models on 3.7k hours custom dataset

Status: Open.

#176 In k2-fsa/ZipVoice;

· mukherjeesougata-eros opened on Feb 18, 2026

Training restarts from epoch 1 on rerun even though previous epoch checkpoints exist

Status: Open.

#175 In k2-fsa/ZipVoice;

· mukherjeesougata-eros opened on Feb 10, 2026

Paralell Inference

Status: Open.

#174 In k2-fsa/ZipVoice;

· alpcansoydas opened on Feb 10, 2026