MiniMax-M3.1-Flash Vision
Extracted from MiniMaxAI/MiniMax-M3.1-Flash at 54e4e69dfc452b1e94a426c24a4fdeeaa8e3ad5e.
Copyright MiniMax. Distributed under the accompanying MiniMax Community License.
Tower and Projector tensors are preserved exactly. State-dictionary keys are adapted for native Transformers classes.
This is a standalone vision extraction, not the complete language model.
projector.py and examples/inference.py are original vision-tower-bench helper code.
Copyright (c) 2026 Immanuel Peter. MIT; see LICENSE-CODE.