vLLM v0.1.5 — Release Notes & Engineering Impact

## Major Changes * Align beam search with `hf_model.generate`. * Stablelize AsyncLLMEngine with a background engine loop. * Add support for CodeLLaMA. * Add many model correctness tests. * Many other correctness fixes. ## What's Changed * Add support for CodeLlama by @Yard1 in https://github.com/vllm-project/vllm/pull/854 * [Fix] Fix a condition for ignored sequences by @zhuohan123 in https://github.com/vllm-project/vllm/pull/867 * use flash-attn via xformers by @tmm1 in https://github.com/vllm-project/vllm/pull/877 * Enable request body OpenAPI spec for OpenAI endpoints by @Peilun-Li in https://github.com/vllm-project/vllm/pull/865 * Accelerate LLaMA model loading by @JF-D in https://github.com/vllm-project/vllm/pull/234 * Improve _prune_hidden_states micro-benchmark by @tmm1 in https://github.com/vllm-project/vllm/pull/707 * fix: bug fix when penalties are negative by @pfldy2850 in https://github.com/vllm-project/vllm/pull/913 * [Docs] Minor fixes in supported models by @WoosukKwon in https://github.com/vllm-project/vllm/pull/920 * Fix README.md Link by @zhuohan123 in https://github.com/vllm-project/vllm/pull/927 * Add tests for models by @WoosukKwon in https://github.com/vllm-project/vllm/pull/922 * Avoid compiling kernels for double data type by @WoosukKwon in https://github.com/vllm-project/vllm/pull/933 * [BugFix] Fix NaN errors in paged attention kernel by @WoosukKwon in https://github.com/vllm-project/vllm/pull/936 * Refactor AsyncLLMEngine by @Yard1 in https://github.com/vllm-project/vllm/pull/880 * Only emit warning about internal tokenizer if it isn't being used by @nelson-liu in https://github.com/vllm-project/vllm/pull/939 * Align vLLM's beam search implementation with HF generate by @zhuohan123 in https://github.com/vllm-project/vllm/pull/857 * Initialize AsyncLLMEngine bg loop correctly by @Yard1 in https://github.com/vllm-project/vllm/pull/943 * FIx vLLM cannot launch by @HermitSun in https://github.com/vllm-project/vllm/pull/948 * Clean up kernel unit tests by @WoosukKwon in https://github.com/vllm-project/vllm/pull/938 * Use queue for finished requests by @Yard1 in https://github.com/vllm-project/vllm/pull/957 * [BugFix] Implement RoPE for GPT-J by @WoosukKwon in https://github.com/vllm-project/vllm/pull/941 * Set torch default dtype in a context manager by @Yard1 in https://github.com/vllm-project/vllm/pull/971 * Bump up transformers version in requirements.txt by @WoosukKwon in https://github.com/vllm-project/vllm/pull/976 * Make `AsyncLLMEngine` more robust & fix batched abort by @Yard1 in https://github.com/vllm-project/vllm/pull/969 * Enable safetensors loading for all models by @zhuohan123 in https://github.com/vllm-project/vllm/pull/974 * [FIX] Fix Alibi implementation in PagedAttention kernel by @zhuohan123 in https://github.com/vllm-project/vllm/pull/945 * Bump up the version to v0.1.5 by @WoosukKwon in https://github.com/vllm-project/vllm/pull/944 ## New Contributors * @tmm1 made their first contribution in https://github.com/vllm-project/vllm/pull/877 * @Peilun-Li made their first contribution in https://github.com/vllm-project/vllm/pull/865 * @JF-D made their first contribution in https://github.com/vllm-project/vllm/pull/234 * @pfldy2850 made their first contribution in https://github.com/vllm-project/vllm/pull/913 * @nelson-liu made their first contribution in https://github.com/vllm-project/vllm/pull/939 **Full Changelog**: https://github.com/vllm-project/vllm/compare/v0.1.4...v0.1.5