v2.11.1

• chore(deps): update runpod to latest version

v2.11.0

• add ENABLE_EXPERT_PARALLEL engine arg for MoE models

New Contributors

• @velaraptor-runpod

v2.10.0

• feat: cse 853 vllm template params
• fix(config): update allowed cuda versions in hub and tests config
• fix: remove space from gpuIds
• feat: bump transformers to allow Qwen3-VL

New Contributors

• @eugene-runpod made their first contribution in #230
• @wwydmanski made their first contribution in #225

v2.9.6

• fix: allow also CUDA 12.8 & 12.9

v2.9.5

• chore: update vllm to 0.11.0
• fix: max concurrency = 30 instead of 300

v2.9.4

• removed HF_TOKEN again

v2.9.3

• fix: added back the HF_TOKEN

v2.9.2

• docs: add reasoning parser
• fix: remove "access token" as this is handled by the platform

v2.9.1

• feat: better hub support & concise README for the main repo (#215)

v2.9.0

• feat: prepare worker-vllm for the hub
• fix: allow "None" as value & parse the value of RAW_OPENAI_OUTPUT correctly