Robert Shaw 6874638bc4 [Model Bash] DeepSeek R1 BF16 Min Latency QKV A GEMM (0.5% E2E Speedup) (#34758) ...

Signed-off-by: Robert Shaw <robshaw@redhat.com>
Co-authored-by: Robert Shaw <robshaw@redhat.com>

2026-02-18 07:42:36 -08:00

..

attention

…

core

…

cpu

[Bugfix][CPU] Fix llama4 inference on CPU (#34321)

2026-02-11 19:07:23 +08:00

cutlass_extensions

[Kernel] Add enable_sm120_or_later for SM121 (DGX Spark) CUTLASS support (#33517)

2026-02-06 20:28:01 -08:00

mamba/mamba_ssm

…

moe

[Refactor] Remove align block size logic in moe_permute (#33449)

2026-02-06 10:57:06 -08:00

quantization

[Bugfix] Fix quant RMS norm fusion for quantization with TMA-aligned scales (#33255)

2026-02-17 23:35:04 -08:00

quickreduce

…

rocm

small adjustment to wvSplitKrc (#34410)

2026-02-12 20:26:36 +00:00

sparse/cutlass

[Bugfix] Fix Sparse24 Compressed Tensors models (#33446)

2026-02-11 23:15:16 -08:00

activation_kernels.cu

[Kernel] Apply 256bit LDG/STG To Activation Kernels (#33022)

2026-02-10 19:31:51 -08:00

cache_kernels_fused.cu

Fuse RoPE and MLA KV-cache write (#25774)

2026-01-09 19:18:37 -08:00

cache_kernels.cu

[BugFix] [Build] fix string literals comparison in indexer_k_quant_and_cache calling site (#34653)

2026-02-17 19:19:41 -08:00

cache.h

OffloadingConnector: Support kernel_block_size != block_size (#30692)

2026-01-22 12:30:04 +00:00

cub_helpers.h

…

cuda_compat.h

…

cuda_utils_kernels.cu

…

cuda_utils.h

…

cuda_view.cu

[Feature] Support CPU Offloading without Pytorch Pinned Memory that leads to doubled allocation (#32993)

2026-02-13 08:11:26 -08:00

cumem_allocator_compat.h

…

cumem_allocator.cpp

[Feature][Core] Support Fabric detection to adapt the MNNVL protocol for the GB series (#33540)

2026-02-02 22:55:46 +08:00

custom_all_reduce_test.cu

…

custom_all_reduce.cu

…

custom_all_reduce.cuh

…

custom_quickreduce.cu

…

dispatch_utils.h

…

dsv3_fused_a_gemm.cu

[Model Bash] DeepSeek R1 BF16 Min Latency QKV A GEMM (0.5% E2E Speedup) (#34758)

2026-02-18 07:42:36 -08:00

fused_qknorm_rope_kernel.cu

[ROCm] [Critical]: Remove unused variable (#31156)

2025-12-22 08:28:22 -08:00

launch_bounds_utils.h

…

layernorm_kernels.cu

…

layernorm_quant_kernels.cu

…

ops.h

[Model Bash] DeepSeek R1 BF16 Min Latency QKV A GEMM (0.5% E2E Speedup) (#34758)

2026-02-18 07:42:36 -08:00

permute_cols.cu

…

pos_encoding_kernels.cu

…

sampler.cu

[Perf][Kernel] Add faster topKperRow decode kernel for DeepSeek-V3.2 sparse attention (#33680)

2026-02-10 10:29:52 -05:00

topk.cu

[Bugfix] Replace c10::optional with std::optional in topk kernel (#34467)

2026-02-13 08:30:23 -08:00

torch_bindings.cpp

[Model Bash] DeepSeek R1 BF16 Min Latency QKV A GEMM (0.5% E2E Speedup) (#34758)

2026-02-18 07:42:36 -08:00

type_convert.cuh

…