fix: preserve parameter attrs during weight reload (FP8 block + MXFP4 MoE)#2

Open

fergusfinn wants to merge 32 commits intomainfrom

fix/reload-preserve-param-attrs

Commits on Jan 26, 2026

[Bugfix] Fix Dtypes for Pynccl Wrapper (#33030)

authored and
khluu
committed

Commits on Jan 28, 2026

[CI] Fix AssertionError: MCP tool call not found in output_messages (#33093)

chaunceyjiang
authored and
khluu
committed
[Models] Kimi-K2.5 (#33131)

authored and
khluu
committed
[Bugfix] Disable CG for Whisper+FA2 (#33164)

NickLucche
authored and
khluu
committed
[Bugfix] Fix DeepseekV32 `AssertionError: num_kv_heads == 1` (#33090)

NickLucche
authored and
khluu
committed
[BugFix] Fix P/D with non-MoE DP (#33037)

njhill
authored and
khluu
committed
Relax protobuf library version constraints (#33202)

jeffreywang-anyscale
authored and
khluu
committed
Revert "Enable Cross layers KV cache layout at NIXL Connector (#30207)" (#33241)

orozery
and
khluu
committed

Commits on Jan 29, 2026

[Release] [CI] Optim release pipeline (#33156)

tjtanaa
authored and
khluu
committed

Commits on Feb 2, 2026

[Bugfix] Register fp8 cutlass_group_gemm as supported for only SM90+SM100 (#33285)

mgoin
authored and
khluu
committed
[Bugfix] Fix Qwen3-VL-Reranker load. (#33298)

authored and
khluu
committed
[Bugfix] Enable Triton MoE for FP8 per-tensor dynamic (#33300)

mgoin
authored and
khluu
committed
[release] Minor fixes to release annotation and wheel upload (#33129)
khluu
committed
[Bugfix][ROCm] Fixing the skinny gemm dispatch logic from #32831 (#33366)

gshtras
authored and
khluu
committed
[BugFix] Fix whisper FA2 + full cudagraphs (#33360)

authored and
khluu
committed
[fix][torch.compile] Fix cold-start compilation time increase by adding kv cache update to splitting ops (#33441)

authored and
khluu
committed
fix: Add SM120 (RTX Blackwell) support for FlashInfer CUTLASS NVFP4 MoE kernels (#33417)

authored and
khluu
committed
pin LMCache to v0.3.9 or greater with vLLM v0.15.0 (#33440)

authored and
khluu
committed
[Models] Step-3.5-Flash (#33523)

authored and
khluu
committed
[Nightly CI] Remove CT Model (#33530)

authored and
khluu
committed
[Fix] prefix cache hit rate == 0 bug with gpt-oss style models (#33524)

ivanium
authored and
khluu
committed
[Release] patch step3p5 attention class in v0.15.1 release (#33602)

zhewenl
and
zhewenli
authored

Commits on Feb 3, 2026

[Release] Fix format and cherry-pick (#33618)

zhewenl
and
zhewenli
authored
Patch aiohttp for CVE-2025-69223 (#33621)

zaristei2
and
zaristei
authored
Patch Protobuf for CVE 2026-0994 (#33619)

zaristei2
and
zaristei
authored
[Misc][Build] Lazy load cv2 in nemotron_parse.py (#33189)

kiersten-stokes
authored and
khluu
committed
[torch.compile] Speed up MOE handling in forward_context (#33184)

zou3519
authored and
khluu
committed
[torch.compile] Don't do the fast moe cold start optimization if there is speculative decoding (#33624)

authored and
khluu
committed

Commits on Feb 4, 2026

[Bugfix] Disable RoutingMethodType.[Renormalize,RenormalizeNaive] TRTLLM per-tensor FP8 MoE (#33620)

mgoin
authored and
Robert Shaw
committed
cherry pick

mgoin
authored and
Robert Shaw
committed
[BugFix][Spec Decoding] Fix negative accepted tokens metric crash (#33729)

njhill
authored and
Robert Shaw
committed

Commits on Feb 11, 2026

fix: preserve all parameter attributes during weight reload

fergusfinn
and
claude
committed