Huawei open-sources MACaQ, a quantization tool for multimodal local inference
What happened
Huawei released its internal quantization tool MACaQ under the Apache License 2.0, according to a report on the Hacker News front page on November 12. The tool supports quantization and local inference for text, image, video and audio models, and works with llama.cpp, vLLM, Diffusers and PyTorch. The reported performance figure says MACaQ can quantize a 122B-parameter model in 64 minutes using 32.5 GB of device memory. The report covers the release, the license, framework support and quantization performance.
Written by AI from the coverage · updated 6 hours ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageHuawei releases MACaQ, its internal quantizer for local inference
华为发布内部量化工具 MACaQ，支持文本、图像、视频和音频模型的量化与本地推理，以 Apache License 2.0 开源。工具支持 llama.cpp、vLLM、Diffusers 和 PyTorch，可使用 32.5 GB 设备内存在 64 分钟内量化一个 122B 参数模型。
Heat over time
Heat now 9·Comparable peak 10（Nov 12 06:00）·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.