|
|
1150e325bb
|
Consolidate serving into kernel repo
- Dockerfile: COPY kernel source instead of git clone
- docker-compose: build context at repo root, all debug flags OFF
- vLLM patches: deepseek_v4.py, staging_kernel.py, deepseek_v4_attention.py
- serve_vllm.py script
- .dockerignore to keep image clean
|
2026-05-14 18:20:20 +00:00 |
|