yangzhuxinyzx 74ea7cf330 [Kernel] Add source-sized IQ2_S gated-pair reader ...

Signed-off-by: yangzhuxinyzx <153831768+yangzhuxinyzx@users.noreply.github.com>

2026-10-05 17:52:28 +08:00

..

attention

[Kernel] Add FP32 grouped verification for SM70 FP16 KV caches

2026-10-04 19:44:54 +08:00

core

…

cpu

…

cutlass_extensions

…

libtorch_stable

[Kernel] Package native GGUF formats and Volta prefill routing

2026-10-03 09:45:41 +08:00

moe

[Kernel][SM70] Index Qwen3.8 NVFP4 prefill rows

2026-08-28 16:17:53 +08:00

quantization

[Kernel] Package native GGUF formats and Volta prefill routing

2026-10-03 09:45:41 +08:00

quickreduce

…

rocm

…

sm70_turbomind

[Kernel] Add source-sized IQ2_S gated-pair reader

2026-10-05 17:52:28 +08:00

async_util.cuh

…

cache.h

…

cub_helpers.h

…

cuda_compat.h

…

cuda_utils_kernels.cu

…

cuda_utils.h

…

cuda_view.cu

…

cumem_allocator_compat.h

…

cumem_allocator.cpp

…

custom_all_reduce_test.cu

…

custom_all_reduce.cu

[Test] Retain the matched fused norm reference in the source artifact

2026-10-05 12:45:08 +08:00

custom_all_reduce.cuh

[Test] Retain the matched fused norm reference in the source artifact

2026-10-05 12:45:08 +08:00

custom_quickreduce.cu

…

dispatch_utils.h

…

fused_deepseek_v4_qnorm_rope_kv_insert_kernel.cu

…

launch_bounds_utils.h

…

minimax_reduce_rms_kernel.cu

…

minimax_reduce_rms_kernel.h

…

ops.h

Merge commit '1c0a0e3718' into agent/v100-gguf-iq3-iq4-pair-20261005-050952

2026-10-05 14:30:31 +08:00

persistent_topk.cuh

…

ple_disk_rows.cpp

[Kernel] Prefetch and gather disk-mapped PLE rows on CPU (#885)

2026-10-05 12:18:47 +08:00

qsa_lexicographic_topk.cuh

[Merge] Integrate MTP kernels without changing concurrent arithmetic

2026-09-28 00:43:59 +08:00

sm70_qwen38_hc_batch.cuh

[Kernel] Reuse replicated HC MMA when TP transport costs exceed local savings

2026-10-04 19:43:57 +08:00

sm70_ring_collectives.cu

[Kernel] Register small SM70 ring collectives by peer capability

2026-10-04 13:25:43 +08:00

sm70_tile_runtime_signal.cuh

…

spinloop.cpp

…

torch_bindings.cpp

Merge commit '1c0a0e3718' into agent/v100-gguf-iq3-iq4-pair-20261005-050952

2026-10-05 14:30:31 +08:00

torch_utils.h

…

type_convert.cuh

…