[CI/Build] Allow hermetic builds (#18064)

Fabien Dupont committed over 1 year ago

[Bugfix] Fix some narrowing conversion warnings (#20141)

Tyler Michael Smith committed over 1 year ago

[Bugfix] Mark 'hidden_states' as mutable in moe_forward registration. (#20152)

bnellnm committed over 1 year ago

[Fix][torch.compile] Enable custom ops by default when Inductor off (#20102)

Luka Govedič committed over 1 year ago

[Fix][ROCm] Remove unused variables to fix build error on GFX11/12 (#19891)

Hosang committed over 1 year ago

[Perf] Improved perf for resolve_chat_template_content_format (#20065)

Ilya Lavrenov committed over 1 year ago

[Bugfix] Fix flaky failure when getting DP ports (#20151)

Michael Goin committed over 1 year ago

Gemma3n (Text-only) (#20134)

Robert Shaw committed over 1 year ago

Quick Fix by adding conditional import for flash_attn_varlen_func in flash_attn (#20143)

Chendi.Xue committed over 1 year ago

[Perf][Frontend] eliminate api_key and x_request_id headers middleware overhead (#19946)

Yazan Sharaya committed over 1 year ago

[Model][1/N] Automatic conversion of CrossEncoding model (#20012)

wang.yuqi committed over 1 year ago

[Fix] Fix gemma CI test failing on main (#20124)

Thomas Parnell committed over 1 year ago

[CI] Sync test dependency with test.in for torch nightly (#19632)

Yang Wang committed over 1 year ago

[Feature] add quick all reduce (#19744)

li haoyang committed over 1 year ago

[Bugfix] Build moe_data for both sm100 and sm90 (#20086)

Michael Goin committed over 1 year ago

[Perf] SM100 FP8 GEMM Optimizations after cutlass_profiler (#20071)

Ilya Markov committed over 1 year ago

[Quantization] Bump to use latest `compressed-tensors` (#20033)

Dipika Sikka committed over 1 year ago

[CI Failure] Fix OOM with test_oot_registration_embedding (#20144)

Michael Goin committed over 1 year ago

[Feature] Expert Parallelism Load Balancer (EPLB) (#18343)

Bowen Wang committed over 1 year ago

[Doc] correct LoRA capitalization (#20135)

Kyle Yu committed over 1 year ago