New Contributors
Welcomed three new contributors who made impactful contributions:
- MartinRepo: Slurm deployment documentation. (135)
- dalongbao: GitHub Actions for model testing. (145)
- dblate: Keep-alive configuration feature. (166)
New Features
- **ROCm Support**: Introduced ROCm support, expanding compatibility with AMD GPUs. (118)
- **Hardware Information Collector**: Added a hardware information collector for worker nodes to enhance deployment monitoring. (100)
- **Keep-Alive Configuration**: Added a new `keep_alive` configuration option for better connection management. (166)
Documentation Improvements
- **Developer Guide**: Added a guide for developers to facilitate adapting ServerlessLLM for new hardware support. (130)
- **Architecture Overview**: Provided an in-depth introduction to the architecture of ServerlessLLM. (134)
- **Docker Setup**: Enhanced Docker documentation, including GPU device specification and a guide for Docker Compose setup. (137, 153)
- **Slurm Deployment Guide**: Documented steps for deploying ServerlessLLM on Slurm clusters. (135)
Key Bug Fixes
- **Ray Component Stability**: Resolved segmentation faults by adding control node resources to Ray components. (146)
- **Transformer Backend**: Fixed storage path issues. (148)
- **SllmLocalStore**: Fixed a loading loop issue and enhanced compatibility with the automatic hardware configuration detection feature. (165)
- **API Improvements**: Resolved issues in the embedding and generation APIs for smoother functionality. (152)
Continuous Integration Enhancements
- **Docker Workflow**: Integrated Docker build and push jobs into the publish workflow, automating image deployment. (138)
- **Model Testing**: Added GitHub Actions testing for model storage and deployment support. (145)
What's Changed
* Update `sllm-store` README.md by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/129
* feat: ROCm support by SiyangShao in https://github.com/ServerlessLLM/ServerlessLLM/pull/118
* docs: add developer guide by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/130
* ci(workflow): add Docker build and push job to publish workflow by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/138
* docs: add an architecture introduction by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/134
* chore: update wechat qr code by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/139
* docs: Add Docker Compose Setup for ServerlessLLM by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/137
* docs: update README.md by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/140
* fix: CI for build rocm package by SiyangShao in https://github.com/ServerlessLLM/ServerlessLLM/pull/136
* docs: add documentation for deployment on slurm by MartinRepo in https://github.com/ServerlessLLM/ServerlessLLM/pull/135
* fix: correct git clone path in documentation by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/142
* fix: add control_node resource to Ray components to prevent segfault by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/146
* Fy/new entrypoint by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/147
* [fix] remove storage path from model path in transformer backend by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/148
* docs: update Docker to add guide for specifying GPU devices by JustinTong0323 in https://github.com/ServerlessLLM/ServerlessLLM/pull/153
* docs: correct directory name in Docker quickstart documentation by JustinTong0323 in https://github.com/ServerlessLLM/ServerlessLLM/pull/154
* fix: Resolve issues in the embedding API and generation API by SecretSettler in https://github.com/ServerlessLLM/ServerlessLLM/pull/152
* feat(serve): deploy hardware information collector for worker nodes by JustinTong0323 in https://github.com/ServerlessLLM/ServerlessLLM/pull/100
* Specify supported hardware types and provide a link to AMD GPU by lrq619 in https://github.com/ServerlessLLM/ServerlessLLM/pull/158
* Adding testing for model storage and support in Github Actions by dalongbao in https://github.com/ServerlessLLM/ServerlessLLM/pull/145
* fix: update requirements for worker to support the hardware information collector by JustinTong0323 in https://github.com/ServerlessLLM/ServerlessLLM/pull/162
* fix: Fix dtype error for vllm downloader by SecretSettler in https://github.com/ServerlessLLM/ServerlessLLM/pull/163
* Update publish.yml by lrq619 in https://github.com/ServerlessLLM/ServerlessLLM/pull/167
* feat: add keep_alive config by dblate in https://github.com/ServerlessLLM/ServerlessLLM/pull/166
* [FIX] Fix loading loop in `SllmLocalStore` by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/165
* docs: update qr code image for wechat by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/173
* release: boost version to 0.5.2 by future-xy in https://github.com/ServerlessLLM/ServerlessLLM/pull/175
New Contributors
* MartinRepo made their first contribution in https://github.com/ServerlessLLM/ServerlessLLM/pull/135
* dalongbao made their first contribution in https://github.com/ServerlessLLM/ServerlessLLM/pull/145
* dblate made their first contribution in https://github.com/ServerlessLLM/ServerlessLLM/pull/166
**Full Changelog**: https://github.com/ServerlessLLM/ServerlessLLM/compare/v0.5.1...v0.5.2