chengyansen-ai avatar

chengyansen-ai

Community

@chengyansen-ai

1Followers
|
7Public Repos
|
1Published Skills

Auditable MiniMax H3 LoRA training pipeline covering dataset planning, configuration, evaluation, and packaging across image, video, audio, and motion modalities.

Skills Distribution
DomainAI Models & ...Multimodal Model F.. (45%)Training Dataset E.. (25%)Model Evaluation &.. (20%)Reference-Conditio.. (10%)

Agent Skills by chengyansen-ai

Showing 1 vetted skills indexed across 1 GitHub repositories.

Frequently Asked Questions About chengyansen-ai

FAQPage Schema
What tasks can I accomplish with the h3-multimodal-lora-training skill?

You can plan, audit, configure, run when authorized, evaluate, and package MiniMax H3 LoRA training across images, video, stereo audio, motion, and Ref2VA reference conditioning, producing an auditable dataset and training run.

Who should use this H3 LoRA training skill?

Machine learning engineers and researchers fine-tuning MiniMax H3 models who need auditable LoRA datasets and training runs across multimodal inputs including images, video, stereo audio, and motion data.

When should I NOT use this skill?

Do not use it for prompt-only Hailuo generation or for LoRA training on unrelated image models. It is scoped strictly to MiniMax H3 LoRA dataset preparation, training, evaluation, and packaging.

What does the end-to-end H3 LoRA training process look like?

The process covers dataset planning and auditing, training configuration, execution once authorized, post-training evaluation, and final packaging of the LoRA artifact, with Ref2VA reference conditioning supported throughout.

What modalities does H3 LoRA training support?

It supports five modalities: images, video, stereo audio, motion data, and true Ref2VA reference conditioning, all within a single auditable MiniMax H3 LoRA training pipeline.