Repository Issue Activity (beta feature)

#236 Feature Request: Support sm_120 ( 5090 and blackwell 6000 pro )

#317 DeepSeek-V4 on SM120: tf32_hc_prenorm_gemm and paged_mqa_logits kernels missing

#305 need support sm120

#306 About the version of DeepEP for mega_moe_kernel

#315 [Question] Do you have any plan to impl mega_mla_kernel or mega_mla_with_indexer_kernel?

#313 [Question] Is there a simpler implementation for the UTCCP transposer?

#309 fix: operator precedence bug in pack_ue8m0_to_int assertion (mantissa check always passes)

#281 可不可以不要在春节前或春节期间出v4啊

#204 How to use DeepGEMM Indexer?

#296 [Question] How to autotune DeepGemm FP8 to maximize this baseline performance?

#289 Question about the scale factor layout

#222 [QUESTION] Can fp8_paged_mqa_logits support next_n > 2?

#301 [Bug] JIT cache lacks cross-process synchronization, can cause failures under multi-process parallelism

#177 Create Git tags and releases for versioning

#300 CUDA kernel errors

#297 [RFC] Using tvm-ffi binding instead of python api to remove the dependency for python binding

#295 [Bug]: NVRTC JIT compilation fails on CUDA 12.8 (smxx_clean_logits.cuh: expression must have a constant value)

#290 [QUESTION]Why does fp8_mqa_logits use four MATH WGs for computation?

#194 Wheel is not supported in the platform

#221 Can fp8_paged_mqa_logits support non-contiguous layout?

#291 [question] sm100_fp8_gemm_1d1d_impl: UTCCP transposer for scale factor is strange

#283 weight gradient kernels for dense and MoE backward 没有grad的反向算子

#209 [QUESTION] dsv32 mqa_logits kernel not considering causal masking?

#239 [Feature Request] Maybe we can add env var to control whether use min sms?

#189 关于Wgrad 计算中 fp8_gemm_nt 输入量化方式的疑问与实验观察