Skip to content

original training dataset and piepline for llm? #405

@sujitvasanth

Description

@sujitvasanth

Hi I think your diffusion models are amazing - especially the turbo versions that dont seem to need the audio codes to generate
But the Qwen omni LLM is poor at instruction following perhaps because its multifunctional
I was thinking of finetuning a different text only llm model for task of enhancing.embelishing prompt and lyric generation
and perhaps metadata generation
I couldnt find the training set you used...can you direct me please?
also i couldnt find the training pipeline for the llm acestep base model in this repo? (not LORA training)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions