[CI/Build][BugFix][The Rock][AMD] Add spawn method in vision examples to avoid reinitialization (#47932)

rasmith committed 3 months ago

[ROCm] [Release] [Per-commit] Reenable per commit rocm wheel (#49044)

TJian committed 3 months ago

[Bugfix] Reject removed pooling parameters (#48984)

Taneem Ibrahim committed 3 months ago

[Refactor] Extract StructuredOutputsParams creation logic from Request.to_sampling_params (#49003)

yzong-rh committed 3 months ago

[Core] Add MRV2 virtual-batch PCP for MLA (#46570)

Lucas Wilkinson committed 3 months ago

[Multimodal] Automatically fallback to ViT DP when TP is unavailable (#49046)

Isotr0py committed 3 months ago

[ROCm][CI] Ensure sliding window tests release GPU memory (#49055)

Andreas Karatzas committed 3 months ago

[Bugfix] Qwen3-VL/Qwen-Omni: honor max_pixels/min_pixels for video prompts (#49015)

SYLAR committed 3 months ago

[Front-end] [Messages] Populate `num_cache_creation_tokens` (#48535)

yzong-rh committed 3 months ago

[ROCm][DSv3.2][Perf] Cap sparse MLA decode KV-splits with a work-per-split heuristic (#46832)

frida-andersson committed 3 months ago

[ci] Move 3 entrypoints tests to h200_35gb queue (#43164)

Kevin H. Luu committed 3 months ago

[Bugfix] Bump tml-fa4 for cutlass-dsl 4.6 API compatibility (#48988)

Michael Goin committed 3 months ago

Remove even more unnecessary `load_weights` methods (#48496)

Harry Mellor committed 3 months ago

[Bugfix] Re-sync parameter tp_rank after process_weights_after_loading (fix replicated / disable_tp weight reload) (#48025)

alexxu-roblox committed 3 months ago

[Perf][Hybrid] Vectorize _copy_mamba_state_block to uint64 for temporal (#48110)

Francesco Fusco committed 3 months ago

[ROCm][Bugfix] Fix GPT-OSS Quark MXFP4 MoE loading - emulation buffer not block-aligned (#43979)

xuebwang-amd committed 3 months ago

[LoRA] Optimize TrtLlmLoRAExperts (#48759)

Jee Jee Li committed 3 months ago

[Sampler] Stop upcasting logits to fp32 in apply_sampling_params (#48641)

Michael Goin committed 3 months ago

[Model] Use standard ModelOpt config for Inkling NVFP4 (#48990)

Michael Goin committed 3 months ago

[Frontend]Flatten beam-search beams with itertools.chain instead of sum (#48829)

Wang Xingda committed 3 months ago