Wan2.2-TI2V-5B-VedioQuant

Wan2.2 + TurboQuant cache compression = 10x less VRAM for video generation

This model packages Wan2.2-TI2V-5B (the current best open-source video model family) with VedioQuant cache compression pre-configured. Just load and use — no extra setup needed.

What's Different From Base Wan2.2?

	Base Wan2.2-5B	This Model (+ VedioQuant)
Model weights	Unchanged	Identical (same weights)
Video quality	Baseline	0.983 cosine similarity (< 2% loss)
720P/81f 30-layer cache	25.96 GB	2.44 GB (10.6× smaller)
VRAM saved	—	23.5 GB
Extra code needed	N/A	None (auto-enabled)

The model weights are identical to the original Wan2.2. VedioQuant only compresses the inference cache (not the weights) using TurboQuant's random rotation + quantization technique, originally developed for LLM KV Cache compression.

Quick Start

import torch
from pipeline_vedioquant import VedioQuantPipeline

# Load — VedioQuant is auto-enabled
pipe = VedioQuantPipeline.from_pretrained(
    "viberobin/Wan2.2-TI2V-5B-VedioQuant",
    torch_dtype=torch.float16,
)
pipe.to("cuda")  # or "mps" for Mac

# Generate video — same API as WanPipeline
video = pipe(
    prompt="a cat sitting on a sofa, cinematic lighting",
    num_frames=17,
    height=480,
    width=832,
).frames[0]

# Check compression stats
print(pipe.get_vedioquant_stats())
# → {'steps': 50, 'cache_hits': 32, 'hit_rate': '64%',
#    'bits': 3, 'compression_ratio': '10.7x'}

Configuration

# Adjust compression (default: 3-bit, recommended)
pipe.enable_vedioquant(bits=3, threshold=0.05)

# More aggressive compression (16x, slightly lower quality)
pipe.enable_vedioquant(bits=2)

# More conservative (8x, near-lossless)
pipe.enable_vedioquant(bits=4)

# Disable entirely (same as base Wan2.2)
pipe.disable_vedioquant()

Benchmark (Real Measurement on Wan2.2-TI2V-5B)

Tested with real Wan2.2-5B attention features (hidden_dim=3072, 30 layers):

Config	Compression	Cosine Sim	Min Cosine	Compress	Decompress
2-bit	16.0×	0.9396	0.9363	53ms	16ms
3-bit	10.6×	0.9827	0.9811	14ms	14ms
4-bit	8.0×	0.9953	0.9946	14ms	12ms

Gaussianization Verification

Property	Before Rotation	After Rotation	Theory (Gaussian)
Kurtosis	6.0	0.1	0.0
Std dev	—	0.018040	0.018042

Wan2.2 features are even more Gaussian-friendly than Wan2.1 (kurtosis 6 vs 15), resulting in excellent compression quality.

Memory Savings (720P, 81 frames, 30-layer full cache)

	fp32	3-bit VedioQuant	Saved
Wan2.2-5B	25.96 GB	2.44 GB	23.52 GB
Wan2.1-1.3B	12.98 GB	1.22 GB	11.76 GB

The 5B model has hidden_dim=3072 (vs 1536 for 1.3B), making cache compression even more impactful — saving 23.5 GB of VRAM.

How It Works

VedioQuant applies TurboQuant (Google Research, arXiv:2504.19874) to compress the feature cache used by TeaCache (arXiv:2411.19108) during video diffusion inference:

1. Random orthogonal rotation transforms feature coordinates into a Gaussian distribution
2. Precomputed Lloyd-Max codebook quantizes each coordinate from 32 bits to 3 bits
3. Cache memory drops by 10.6× with only 1.7% quality loss (cosine similarity 0.983)

This is a cross-domain transfer: TurboQuant was designed for LLM KV Cache, but we discovered that video model attention features actually compress better than LLM features (kurtosis 6~15 vs 900).

Citation

@misc{vedioquant2025,
  title={VedioQuant: Extreme Cache Compression for Video Diffusion Model Inference},
  author={Peng Han},
  year={2025},
  url={https://github.com/robin-ph/vedioquant}
}

References

• TurboQuant: arXiv:2504.19874 (Google Research)
• TeaCache: arXiv:2411.19108 (CVPR 2025)
• Wan2.2: GitHub (Alibaba Cloud)
• VedioQuant: GitHub