Files

EX Engine 002f9879b2 ref(upstream): FULL TREE — Deep-Spark xllm (1470) + ds_vllm csrc/models (703)

Replaces cherry-picked upstream_ref with complete source trees.

xllm/ — Iluvatar official C++ inference engine (15MB, 1470 files)
  Complete: kernels → layers → models → runtime → scheduler → api
  Excluded: .git, binary images, third_party submodule checkouts

ds_vllm/ — Iluvatar official vllm fork (8MB, 703 files)
  Included: csrc/ (ALL CUDA kernels), fused_moe/, qwen3_5 model, _custom_ops
  Excluded: tests, benchmarks, docs, examples (not needed for reference)

Critical call chains now fully traceable:
  MoE: moe_topk_softmax_kernels.cuh → ixformer.h → fused_moe.cpp → layer
  GDN: qwen3_gated_delta_net_base.cpp → qwen3_5_gated_delta_net.cpp
  Attention: ixformer.h → xllm_paged_attention → attention.cpp

2026-08-10 02:54:03 +00:00

1.7 KiB

Raw Blame History

English | 中文

Contribute to xLLM

Write / translate / fix our documentation
Raise questions / Answer questions
Provide demos, examples or test cases
Give suggestions or other comments
Paticipate in issues or discussions
Pull requests
Sharing related research / application
Any other ways to improve xLLM

For developers who want to contribute to our code, here is the guidance:

1. Choose an issue to contribute

Issues with label PR welcome, which means:
- A reproducible bug
- A function in plan

2. Install environment for development

We strongly suggest you to read our Document before developing
For setting environment, please check our Readme file

3. Build our project

You could run our demo to check whether the requirements are successfully installed:

4. Test

After the PR is submitted, we will format and test the code. Our tests are still far from perfect, so you are welcomed to add tests to our project!

1.7 KiB Raw Blame History