NTU-yiwen/awm-minimax-h3-new1344-lora-checkpoints

CWM MiniMax-H3 1344×768 LoRA Checkpoints

This repository publicly releases five selected CWM LoRA adapters from our MiniMax-H3 1344×768 q4 System-Chat training runs.

Contents

• q4_system_chat/: five selected System-Chat LoRA adapters.
• manifest.json: exact filenames, byte sizes, SHA-256 checksums, training steps, and checkpoint status.

The repository contains five adapters totaling 11,927,917,704 bytes. Each adapter contains 600 LoRA tensors for 200 target modules. The official MiniMax-H3 Ref2VA base model is not included.

Usage

These checkpoints support inference and LoRA weight warm-start with the matching MiniMax-H3 Ref2VA base model. They do not support optimizer-level exact training resumption because optimizer, scheduler, dataloader, sampler, and RNG states are not included.

Only the selected final or latest completed adapter from each training run is included. A filename without a step suffix denotes the final adapter from a completed run. A filename with a step suffix denotes the latest retained periodic checkpoint from that run.