GitLab

correct the chat template of sglang_server 

MinerU2.0's chat template during training has additional "\n" compared with chatml's. This difference may slightly affect the performance under server mode. (because sglang_server use sglang's chat template).

add new enum values and improve MIN_BATCH_INFERENCE_SIZE documentation in pipeline_analyze.py

fix: include server_url in model key for backend model retrieval in vlm_analyze.py

fix: refine model path condition for transformers and sglang-engine backends in vlm_analyze.py

fix: correct syntax error in demo.py for VLM client backend

fix: correct syntax error in demo.py for VLM client backend

master->dev

update doc

update doc

Release 2.0.3

fix: update sglang version to 0.4.7 and adjust changelog for compatibility issues

master->dev

Release 2.0.2

Dev

master->dev

Release 2.0.1