drakoshaandMatthew Bonanni 6472131298 [Bugfix] Set kv_quant_mode on the generic MLA KV-cache spec (#48379) ...

Signed-off-by: Mikhail Kostryukov <mike@triptrack.net>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>

2026-07-15 03:36:52 +00:00

..

attention

[Bugfix] Set kv_quant_mode on the generic MLA KV-cache spec (#48379)

2026-07-15 03:36:52 +00:00

fla

[Perf][ROCm] Fix GDN KKT warmup regression on RDNA by avoiding fp32 tl.dot (#45000)

2026-07-13 20:48:54 -07:00

fused_moe

[LoRA][1/N] Integrate flashinfer MoE LoRA for BF16 model (#48632)

2026-07-15 10:54:00 +08:00

fusion

Fix docs build on main (#45536)

2026-06-13 08:53:23 -07:00

hpc

[Attention Backend] HPC_ATTN backend support mtp and dynamic scheduled attention (#47433)

2026-07-05 18:18:25 -07:00

mamba

[ROCm] Retune MI355 selective_state_update float32 config on the unified effective_batch grid (#48373)

2026-07-14 18:26:35 +00:00

minimax_rms_norm

[ROCm][MiniMax-M2] Dispatch fused QK-norm + AllReduce via AITER (#44849)

2026-07-14 03:10:16 +00:00

pooler

[Core] Ensure memory is pinned prior to async h2d copy (#45424)

2026-06-20 20:02:24 -07:00

quantization

[Quant] Enable humming w[2-7]a[4,8] inference with compressed-tensors (#46390)

2026-07-14 20:22:31 -06:00

rotary_embedding

fix: guard flash-attn rotary import (#42679)

2026-06-11 08:43:31 -04:00

__init__.py

…

activation.py

[Kernel] Implement CUDA kernel for ReLUSquaredActivation (relu^2) (#39058)

2026-07-12 19:18:03 -07:00

attention_layer_base.py

DCP supports hybrid attention (#40996)

2026-07-09 21:34:45 -07:00

batch_invariant.py

[Feature] Support batch invariant rms norm with residual (#42453)

2026-06-03 15:22:01 -04:00

conv.py

[Misc] Always use forward_mulmat for Conv3d on newer versions of torch. (#38487)

2026-03-30 05:39:41 +00:00

fused_allreduce_gemma_rms_norm.py

[Model] Add MiniMax M3 support (#45381)

2026-06-16 01:01:25 +08:00

fused_qk_norm_rope.py

[Perf] fuse qk rmsnorm rope gate for qwen3.5 (#44176)

2026-06-09 22:12:17 +08:00

layernorm.py

[Kernel] Add weightless RMSNorm CUDA kernels for has_weight=False (#41430) (#44109)

2026-06-16 23:45:55 -07:00

lightning_attn.py

[Kernel][Test] Extend lightning_attn and awq_triton kernel tests to XPU (#43307)

2026-06-04 14:25:59 -04:00

linear.py

[CI] Fix Transformers modeling backend LoRA test (#47832)

2026-07-07 11:40:00 +00:00

logits_processor.py

[Core] Support fp32 lm_head for generation models via head_dtype (RFC #48305 §3.6) (#48390)

2026-07-13 16:43:34 +08:00

mhc.py

[ROCm][DSV4] Use aiter mHC pre/post as the default ROCm path (#43950)

2026-07-01 16:27:42 +08:00

mla.py

[bugfix]Indexer init skip and MTP TopK share for iteration (#45895)

2026-06-19 09:57:51 +08:00

resampler.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

sparse_attn_indexer.py

fix: skip cooperative top-K on SM120 (#47164)

2026-06-30 21:32:54 -07:00

utils.py

[CPU][Spec Decode] Support DFlash speculative decoding for GDN models on CPU (#46090)

2026-07-13 04:16:18 +00:00

vocab_parallel_embedding.py

[Bugfix] Transformers backend: apply learned lm_head.bias for tied-embedding models (#46835)

2026-06-29 08:59:15 +00:00