# MiniMax H3: unified text/video/audio encoder stage + diffusion generation stage. pipeline: minimax_h3_disaggregated async_chunk: false trust_remote_code: true distributed_executor_backend: mp stages: - stage_id: 0 devices: "0,1" max_num_seqs: 1 gpu_memory_utilization: 0.6 tensor_parallel_size: 2 hf_overrides: minimax_h3_encoder_components: text_encoder: parallel_mode: tp video_vae: parallel_mode: patch audio_vae: parallel_mode: leader default_sampling_params: max_tokens: 1 temperature: 0.0 detokenize: false - stage_id: 1 devices: "2,3,4,5" max_num_seqs: 1 model_loaded: text_encoder: false vae_encoder: false parallel_config: tensor_parallel_size: 1 ulysses_degree: 4 vae_patch_parallel_size: 4 default_sampling_params: seed: 42 num_inference_steps: 50