FastVideo/LingBot-Video-MoE-30B-A3B-Diffusers

1---2license: apache-2.03library_name: fastvideo4pipeline_tag: text-to-video5base_model: robbyant/lingbot-video-moe-30b-a3b6tags:7  - fastvideo8  - video-generation9  - text-to-video10  - mixture-of-experts11  - lingbot-video12---13 14# LingBot-Video MoE 30B-A3B for FastVideo15 16This repository contains the released LingBot-Video MoE 30B-A3B base model and17its second-stage refiner in the component layout loaded natively by FastVideo.18 19## Usage20 21Use a FastVideo revision containing22[LingBot-Video support](https://github.com/hao-ai-lab/FastVideo/pull/1595):23 24```python25from fastvideo import VideoGenerator26 27generator = VideoGenerator.from_pretrained(28    "FastVideo/LingBot-Video-MoE-30B-A3B-Diffusers",29    num_gpus=8,30    sp_size=8,31    use_fsdp_inference=True,32    refine_enabled=True,33    text_encoder_cpu_offload=True,34)35result = generator.generate({36    "prompt": "A red fox runs through fresh snow at sunrise.",37    "output": {"output_path": "outputs/lingbot-video", "save_video": True},38})39generator.shutdown()40```41 42The released preset generates a 121-frame 832×480 base video and optionally43refines it to 1920×1088. Set `refine_enabled=False` when loading the model to run44the base MoE transformer without the refiner.45 46## Provenance47 48- Original model: https://huggingface.co/robbyant/lingbot-video-moe-30b-a3b49- Original code: https://github.com/robbyant/lingbot-video50- FastVideo port: https://github.com/hao-ai-lab/FastVideo/pull/159551 52The original model and this converted package are distributed under the53Apache-2.0 license declared by the released checkpoint.54