InstantX/MiniMax-H3-Turbo-Lora-Diffusers

1# MiniMax-H3 inference (infer.py / convert.py)2# Versions pinned to the working docker env (CUDA 12.6).3#4# Install torch first from the CUDA wheel index, then the rest:5#   pip install torch==2.9.1 torchvision==0.24.1 --index-url https://download.pytorch.org/whl/cu1266#   pip install -r requirements.txt7#8# MiniMax-H3 is not in a released diffusers build yet — install the local checkout9# (or the PR) instead of the PyPI package:10#   pip install -e ./diffusers11#   # or: pip install git+https://github.com/huggingface/diffusers.git@refs/pull/14355/head12 13# --- PyTorch (install via the CUDA index above; listed here for reference) ---14# torch==2.9.1+cu12615# torchvision==0.24.1+cu12616 17torch==2.9.118torchvision==0.24.119 20# --- Hugging Face stack ---21transformers==5.14.122accelerate==1.14.023huggingface-hub==1.26.024safetensors==0.8.025peft==0.20.026tokenizers==0.22.227sentencepiece==0.2.028protobuf==4.24.429 30# --- Diffusers extras used by infer.py ---31kernels==0.16.032 33# --- Numerics / IO ---34numpy==1.24.435scipy==1.12.036Pillow==10.2.037einops==0.7.038ftfy==6.2.039opencv-python==4.7.040av==17.1.041soundfile==0.12.142imageio==2.37.443imageio-ffmpeg==0.6.044