xInfer v0.14.1

What's Changed

• Support new format AWQ/GPTQ MoE models by @guoqingbao in #451
• Fix MTP multirank issue on Blackwell & Logits projection sharding by @guoqingbao in #448
• Optimize safetensors model loading & fp8 model path by @guoqingbao in #453
• Improve native MLA kernel path & AWQ prefill precision by @guoqingbao in #454
• Add Metal AWQ MoE kernel path by @guoqingbao in #455

Full Changelog: v0.14.0...v0.14.1