vllm 0.3.2

v0.3.2

on Python PyPI

Major Changes

This version adds support for the OLMo and Gemma Model, as well as seed parameter.

What's Changed

• Defensively copy sampling_params by @njhill in #2881
• multi-LoRA as extra models in OpenAI server by @jvmncs in #2775
• Add code-revision config argument for Hugging Face Hub by @mbm-ai in #2892
• [Minor] Small fix to make distributed init logic in worker looks cleaner by @zhuohan123 in #2905
• [Test] Add basic correctness test by @zhuohan123 in #2908
• Support OLMo models. by @Isotr0py in #2832
• Add warning to prevent changes to benchmark api server by @simon-mo in #2858
• Fix vllm:prompt_tokens_total metric calculation by @ronensc in #2869
• [ROCm] include gfx908 as supported by @jamestwhedbee in #2792
• [FIX] Fix beam search test by @zhuohan123 in #2930
• Make vLLM logging formatting optional by @Yard1 in #2877
• Add metrics to RequestOutput by @Yard1 in #2876
• Add Gemma model by @xiangxu-google in #2964
• Upgrade transformers to v4.38.0 by @WoosukKwon in #2965
• [FIX] Add Gemma model to the doc by @zhuohan123 in #2966
• [ROCm] Upgrade transformers to v4.38.0 by @WoosukKwon in #2967
• Support per-request seed by @njhill in #2514
• Bump up version to v0.3.2 by @zhuohan123 in #2968

New Contributors

• @jvmncs made their first contribution in #2775
• @mbm-ai made their first contribution in #2892
• @Isotr0py made their first contribution in #2832
• @jamestwhedbee made their first contribution in #2792

Full Changelog: v0.3.1...v0.3.2