A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 5, 14:45 UTC
main
full cudagraph for flex-attn (#36298)
shunting314 committed 7 months ago
[Model] Add Phi4ForCausalLMV for microsoft/Phi-4-reasoning-vision-15B (#38306)
Varun Sundar Rabindranath committed 7 months ago
[CI][ROCm] Add Qwen3.5-35B-A3B-MXFP4 model eval into CI (#38664)
Bowen Bao committed 7 months ago
[ROCm][Quantization][1/N] Refactor quark_moe w_mxfp4 w/ oracle (#38774)
Bowen Bao committed 7 months ago
refactor hard coded device string in test files under tests/v1 and tests/lora (#37566)
wliao2 committed 7 months ago
[Perf] Batch KV cache swap copies via cuMemcpyBatchAsync (#38460)
Itay Etelis committed 7 months ago
[Kernel] Fuse FP8 output quantization into merge_attn_states (#36518)
Carl Y committed 7 months ago
[mla] Support fused FP8/NVFP4 output quantization in MLA attention (#35792) (#36205)
Carl Y committed 7 months ago
[XPU] Initial support for GDN attention on Qwen3-next/Qwen3.5 (#33657)
Yan Ma committed 7 months ago
[CI] Fix `test_nixl_connector` (#38838)
Matthew Bonanni committed 7 months ago
[Bugfix] Fix NVFP4+MTP crash: force unquantized mtp.fc for Qwen3.5 (#38832)
Vadim Gimpelson committed 7 months ago
[New Model]: add support for telechat3 (#38510)
1096125073 committed 7 months ago
[Kernel] Mamba support different layout for Conv state (#37416)
Nicolò Lucchesi committed 7 months ago
[Bugfix]: Fix Gemma4ToolParser.__init__() missing `tools` parameter (#38847)
Michael committed 7 months ago
[CI] Fix: pass string cache_dtype in test_register_kv_caches (#38836)
zhanqiuhu committed 7 months ago
[CI] Add flashinfer.py to attention test source deps (#38792)
Stefano Castagnetta committed 7 months ago
Luciano Martins committed 7 months ago
Bump helion dependency from 0.3.2 to 0.3.3 (#38062)
Yanan Cao committed 7 months ago
[Bugfix] Fix test mocks after SM100 restriction in #38730 (#38791)
Stefano Castagnetta committed 7 months ago
[FA4] Update flash-attention to latest upstream FA4 (#38690)
Lucas Wilkinson committed 7 months ago