dev

Yyzydevfeat: upgrade management and routing improvements

b0bb4f95创建于 11 天前历史提交

文件	最后提交记录	最后更新时间
experts	feat: upgrade management and routing improvements Major features: - Add elastic management for frequency tracking, load tracking, and routing broadcasts. - Add routing service APIs and expertkit-transport-rs with gRPC, shared-memory, RDMA, and load-balancing strategies. - Integrate worker-side weight management, peer weight transfer, byte-indexed weight serving. - Expand model/runtime support for DeepSeek v2/v3, Mixtral, Qwen3-MoE, vLLM integration, and Aliyun/local validation configs. Important fixes: - Fix fallback path stalls, worker disconnect fallback, torch backend device-index recognition, and DeepSeek-v2-lite runnable support. - Improve recovery by batching recovery upserts, deduplicating recovery triggers, keeping routing stable. - Align controller/worker liveness checks, routing subscriptions.	11 天前
models	feat: vLLM plugin support (#53) * feat: add config for ek-vllm plugin feat: integrate vLLM with ek framework and fix token output issues fix: add missing shared expert output summation in DeepSeek MoE forward pass * typo: unify env variables from EXPERTKIT to EK fix: correct vllm output * doc: Update vLLM plugin docs for latest configuration method * fix: Change the example model path from a local directory path to a Hugging Face model name.	1 年前
pbpy	feat: update proto for expertkit-vllm, need fix some bug	1 年前
utils	feat: vLLM plugin support (#53) * feat: add config for ek-vllm plugin feat: integrate vLLM with ek framework and fix token output issues fix: add missing shared expert output summation in DeepSeek MoE forward pass * typo: unify env variables from EXPERTKIT to EK fix: correct vllm output * doc: Update vLLM plugin docs for latest configuration method * fix: Change the example model path from a local directory path to a Hugging Face model name.	1 年前
__init__.py	feat[integration]: impl plugin for vllm, need for test	1 年前
grpc_client.py	feat: upgrade management and routing improvements Major features: - Add elastic management for frequency tracking, load tracking, and routing broadcasts. - Add routing service APIs and expertkit-transport-rs with gRPC, shared-memory, RDMA, and load-balancing strategies. - Integrate worker-side weight management, peer weight transfer, byte-indexed weight serving. - Expand model/runtime support for DeepSeek v2/v3, Mixtral, Qwen3-MoE, vLLM integration, and Aliyun/local validation configs. Important fixes: - Fix fallback path stalls, worker disconnect fallback, torch backend device-index recognition, and DeepSeek-v2-lite runnable support. - Improve recovery by batching recovery upserts, deduplicating recovery triggers, keeping routing stable. - Align controller/worker liveness checks, routing subscriptions.	11 天前
plugin.py	feat: upgrade management and routing improvements Major features: - Add elastic management for frequency tracking, load tracking, and routing broadcasts. - Add routing service APIs and expertkit-transport-rs with gRPC, shared-memory, RDMA, and load-balancing strategies. - Integrate worker-side weight management, peer weight transfer, byte-indexed weight serving. - Expand model/runtime support for DeepSeek v2/v3, Mixtral, Qwen3-MoE, vLLM integration, and Aliyun/local validation configs. Important fixes: - Fix fallback path stalls, worker disconnect fallback, torch backend device-index recognition, and DeepSeek-v2-lite runnable support. - Improve recovery by batching recovery upserts, deduplicating recovery triggers, keeping routing stable. - Align controller/worker liveness checks, routing subscriptions.	11 天前