Abhishek Kumar Tiwari

icon indicating a website link https://discord.gg/QMcKPdYQ

icon company ModelsLab icon location Kharagpur, West Bengal icon location Building a collaborative project building space

Results 6 comments of Abhishek Kumar Tiwari

Low Model Accuracy After Extended Training with Unicode-Based Language

Hey are you doing lora finetuning or full finetuning?

How to continue training with finetune? What are the steps to continue training

I am facing the same issue can somebody help? I have passed the merged model directory in config.yaml ,```pretrained_ckpt_path: /workspace/FISHSPEECH/lora_weights```. @Stardust-minus @HimanshuRepozitory

Fine-tuning result gradually becoming noise

Hey there I have been finetuning the v1.5 model on english dataset just to see how it behaves and my loss curve doesnt go below 8. Any reasons why it...

Fine-tuning result gradually becoming noise

![image](https://github.com/user-attachments/assets/75a71700-d83f-45e3-b162-76daaecf16f1) ![image](https://github.com/user-attachments/assets/a453bd09-8d6a-49a6-89f4-91b2ef48f911) Are the loss curves going fairly well? I do not what should be the minimum loss when the generated samples would sound well.

[Feature] Add other languages

Has anyone got good results on finetuning the v1.5 model in any language?

[Feature] Add other languages

> Tamil language support. Hey @SmartManoj Could you make any progress in finetuining for indic language? Needed some help as i am trying for hindi language lora finetuning.