v0.1.2

[LeiWang1999]

What's Changed

• [Dev] Add MLA and GQA decode examples by @chengyupku in [Dev] Add MLA and GQA decode examples #109
• [Example] Add Split-K and Stream-K Examples and move MLA from fld to mla by @LeiWang1999 in [Example] Add Split-K and Stream-K Examples and move MLA from fld to mla #110
• [Typo] Fix a typo in gemm splitk examples by @LeiWang1999 in [Typo] Fix a typo in gemm splitk examples #111
• [Typo] Fix links in installation instructions in README.md by @xwhzz in [Typo] Fix links in installation instructions in README.md #112
• [Typo] Fix formatting in installation instructions in README.md by @xwhzz in [Typo] Fix formatting in installation instructions in README.md #113
• [Benchmark] Add benchmark scripts for block sparse attention by @LeiWang1999 in [Benchmark] Add benchmark scripts for block sparse attention #114
• [Dev] Support vectorized value pack and atomicAdd for BFloat16 DType by @LeiWang1999 in [Dev] Support vectorized value pack and atomicAdd for BFloat16 DType #116
• [Bugfix] Bugfix of pass order for hopper by @chengyupku in [Bugfix] Bugfix of pass order for hopper #117
• [Dev] Update MLA decode kernel by @chengyupku in [Dev] Update MLA decode kernel #120
• [Example] Add GQA Example by @LeiWang1999 in [Example] Add GQA Example #118
• [Example] Implement TileLang Native Sparse Attention Kernel by @LeiWang1999 in [Example] Implement TileLang Native Sparse Attention Kernel #121
• [Doc] Update README.md with new example links for Flash MLA Decoding and Native Sparse Attention by @chengyupku in [Doc] Update README.md with new example links for Flash MLA Decoding and Native Sparse Attention #122
• [Example] Update GEMM FP8 Example by @LeiWang1999 in [Example] Update GEMM FP8 Example #123
• [Dev] Add RetNet Linear Attention example by @chengyupku in [Dev] Add RetNet Linear Attention example #124
• [JIT] Enhance cython/ctypes wrapper for tma descriptor by @LeiWang1999 in [JIT] Enhance cython/ctypes wrapper for tma descriptor #126
• [Dev][Bugfix] Fix bug in ThreadTagChecker; Add WgmmaSync rewriter and add MHA WGMMA pipelined example by @chengyupku in [Dev][Bugfix] Fix bug in ThreadTagChecker; Add WgmmaSync rewriter and add MHA WGMMA pipelined example #128
• [Dev] Remove buffer flatten when debug print a shared buffer by @LeiWang1999 in [Dev] Remove buffer flatten when debug print a shared buffer #129
• [Debug] Support T.print for fragment scope by @LeiWang1999 in [Debug] Support T.print for fragment scope #130
• [Example] Implememt FMHA Varlen Example by @LeiWang1999 in [Example] Implememt FMHA Varlen Example #131
• [Refactor] Set default log level from waning into info by @LeiWang1999 in [Refactor] Set default log level from waning into info #132
• [Kernel] Implement different SEQ Q/KV examples with block sparse by @LeiWang1999 in [Kernel] Implement different SEQ Q/KV examples with block sparse #133
• [Dev][Doc] Add DeepSeek MLA Decode Example with Documentation and Performance Benchmarks by @chengyupku in [Dev][Doc] Add DeepSeek MLA Decode Example with Documentation and Performance Benchmarks #134
• [Doc] Update MLA Documentation by @chengyupku in [Doc] Update MLA Documentation #135
• [Debug] Improve Memory Layout Plot by @LeiWang1999 in [Debug] Improve Memory Layout Plot #136
• [Doc] Add MLA Decoding Performance Benchmarks and Documentation by @chengyupku in [Doc] Add MLA Decoding Performance Benchmarks and Documentation #137
• [Bugfix] Add missing definition for AtomicAdd by @LeiWang1999 in [Bugfix] Add missing definition for AtomicAdd #138
• [Dev][Doc] Enhance Flash Attention Implementation in GQA Decoding Example and Fix Typo by @chengyupku in [Dev][Doc] Enhance Flash Attention Implementation in GQA Decoding Example and Fix Typo #139
• [Dev] Adjust computation logic to avoid precision loss when casting acc_s from float to float16 by @chengyupku in [Dev] Adjust computation logic to avoid precision loss when casting acc_s from float to float16 #141
• [Refactor] Rename gemm fp8 example as we currently lack T.gemm support for fp8 by @LeiWang1999 in [Refactor] Rename gemm fp8 example as we currently lack T.gemm support for fp8 #144
• [Enhancement] Support debug print for unsigned char datatype by @LeiWang1999 in [Enhancement] Support debug print for unsigned char datatype #145
• [Enhancement] Enable runtime tensor data type validation by @LeiWang1999 in [Enhancement] Enable runtime tensor data type validation #146
• [Refactor] Adapt Caver to benchmark by @LeiWang1999 in [Refactor] Adapt Caver to benchmark #148
• [Refactor] Remove BitBLAS Import Check in Benchmark by @SiriusNEO in [Refactor] Remove BitBLAS Import Check in Benchmark #150
• [Enhancement] Optimize TileLang install scripts with Dynamic CPU Cores by @LeiWang1999 in [Enhancement] Optimize TileLang install scripts with Dynamic CPU Cores #152
• [Carver] Enhance Carver Adaptation for MatMul Benchmarking by @LeiWang1999 in [Carver] Enhance Carver Adaptation for MatMul Benchmarking #153
• [Dev][Benchmark] Add MLA paged decoding example and benchmark script by @chengyupku in [Dev][Benchmark] Add MLA paged decoding example and benchmark script #158
• [Release] Bump Version to v0.1.2 by @LeiWang1999 in [Release] Bump Version to v0.1.2 #155

New Contributors

• @SiriusNEO made their first contribution in [Refactor] Remove BitBLAS Import Check in Benchmark #150

Full Changelog: v0.1.1...v0.1.2

---

This discussion was created from the release v0.1.2.