v0.2.4

What's Changed

Fingerprint

• [Fingerprint] Unify fingerprint calculation by @jiazhihao in #171

Grace Hopper Support

• Grace Hopper: let users assign tasks to different warp groups by @xinhaoc in #165
• Set num_warp_groups and pipeline_stages with default value in generate_cuda_program() by @xinhaoc in #179
• Fix MMA Threadlayout issue by @xinhaoc in #197
• Hopper: Add bf16 and fix some corner cases by @xinhaoc in #198

QWen2.5 Demo

• DeepSeek / Qwen demo by @jiazhihao in #192
• [Demo] DeepSeek Demo Part 2 by @jiazhihao in #212
• [Demo] DeepSeek demo Part 3 by @jiazhihao in #215

New operators

• Added GeLU by @AMKCode in #180

Triton backend

• Triton warnings fix & triton kernel added by @NorthmanPKU in #181
• Triton kernel for RoPE by @NorthmanPKU in #218
• [Triton] Move Triton runtime under triton_transpiler/ instead of transpiler/ by @NorthmanPKU in #227

Others

• Fix build issue caused by the previous pr by @xinhaoc in #169
• CI Tests by @yzhou442 in #203
• Profiler by @xinhaoc in #199

New Contributors

• @olivia111 made their first contribution in #186
• @Theorem411 made their first contribution in #208

Full Changelog: v0.2.3...v0.2.4