LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generation

Unknown Date

AI Summary

This work addresses the computational and memory bottlenecks in long video generation during both training and inference by proposing the first end-to-end efficient generation framework supporting NVFP4 precision. By integrating sequence-parallel autoregressive training, chunked VAE encoding, quantized KV caching, asynchronous streaming VAE decoding, and LoRA fine-tuning, the method eliminates conventional ODE initialization and distillation pipelines, directly adapting diffusion models into multi-shot interactive autoregressive generators. Evaluated on LongLive-2.0-5B, the approach achieves a 2.15× speedup in training and a 1.84× acceleration in inference, enabling real-time video synthesis at 45.7 FPS while preserving high generation quality.

Abstract

We present LongLive-2.0, an NVFP4-based parallel infrastructure throughout the full training and inference workflow of long video generation, addressing speed and memory bottlenecks. For training, we introduce sequence-parallel autoregressive (AR) training, instantiated as Balanced SP, which co-designs the efficient teacher-forcing layout with SP execution by pairing clean-history and noisy-target temporal chunks on each rank, enabling a natural teacher-forcing mask with SP-aware chunked VAE encoding. Combined with NVFP4 precision, it reduces GPU memory cost and accelerates GEMM computation during training, the proportion of which increases as video length grows. Moreover, we show that a high-quality infrastructure and dataset enable a remarkably clean training pipeline. Unlike existing Self-Forcing series methods that rely on ODE initialization and subsequent distribution matching distillation (DMD), LongLive-2.0 directly tunes a diffusion model into a long, multi-shot, interactive auto-regressive (AR) diffusion model. It can be further converted to real-time generation (4 to 2 denoising steps) with standalone LoRA weights. For inference on Blackwell GPUs, we enable W4A4 NVFP4 inference, quantize KV cache into NVFP4 for memory savings, and boost end-to-end throughput with asynchronous streaming VAE decoding. On non-Blackwell GPU architectures, we deploy SP inference to match the speed on Blackwell GPUs, while the quantized KV cache can lower inter-GPU communication of SP. Experiments show up to 2.15x speedup in training, and 1.84x in inference. LongLive-2.0-5B achieves 45.7 FPS inference while attaining strong performance on benchmarks. To our knowledge, LongLive-2.0 is the first NVFP4 training and inference system for long video generation.

Problem

• long video generation
• training bottleneck
• inference speed
• memory efficiency
• NVFP4

Innovation

• NVFP4
• sequence-parallel autoregressive training
• long video generation
• quantized KV cache
• real-time diffusion AR model