Search results

[Feature]: Preparing a v2 to address the comments from the SGLang community

stale

Status: Open.

#94 In bytedance-iaas/vllm;

· JackChuang opened on Mar 31, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Feature]: Generate Triton code using torch.compile and compare the performance with it

low priority

stale

Status: Open.

#85 In bytedance-iaas/vllm;

· JackChuang opened on Mar 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Feature] Get performance numbers on H20 for the SGLang PR

stale

Status: Open.

#84 In bytedance-iaas/vllm;

· JackChuang opened on Mar 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Feature]: Support Triton-Based Non-Attention Operators in SGLang

stale

Status: Open.

#83 In bytedance-iaas/vllm;

· JackChuang opened on Mar 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Feature]: Introduce a Triton-only Transformer Execution Path in SGLang

stale

Status: Open.

#82 In bytedance-iaas/vllm;

· JackChuang opened on Mar 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Bug]: No attn_metadata tensor found in kwargs

bugSomething isn't workingSomething isn't working

stale

Status: Open.

#77 In bytedance-iaas/vllm;

· CharlesJhonson opened on Mar 11, 2025

[Bug]: HFValidation

bugSomething isn't workingSomething isn't working

stale

Status: Open.

#76 In bytedance-iaas/vllm;

· CharlesJhonson opened on Mar 6, 2025

[Feature]: Add Triton Multihead Attention (MHA) Backend in vLLM

low priority

stale

Status: Open.

#75 In bytedance-iaas/vllm;

· JackChuang opened on Feb 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Feature]: Introduce a Triton-only Transformer Execution Path in vLLM

low priority

stale

Status: Open.

#73 In bytedance-iaas/vllm;

· JackChuang opened on Feb 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Doc]: (Blog) Enabling CUDA-Free Support for vLLM Inference Using Triton Kernels

documentationImprovements or additions to documentationImprovements or additions to documentation

stale

Status: Open.

#65 In bytedance-iaas/vllm;

· JackChuang opened on Jan 17, 2025

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance

[Feature]: End-to-end support on AMD

low priority

stale

Status: Open.

#50 In bytedance-iaas/vllm;

· chizhang118 opened on Dec 18, 2024

 ·  

Triton milestone 2 - Run E2E vLLM on NPU and improve performance