v0.1.7

vllm-project/vllmv0.1.7Sep 11, 2023by github-actions[bot]

AI Summary

A minor release focused on bug fixes to ensure stability across various models, specifically addressing issues in ALiBi, Falcon-40B, and Code Llama.

Key Highlights

Fix bugs in ALiBi, Falcon-40B, and Code Llama

New Features

Fix transformers_module ModuleNotFoundError
Fix wrong dtype in PagedAttentionWithALiBi bias
Fix CUDA error when inferencing with Falcon-40B

Full Release Notes

A minor release to fix the bugs in ALiBi, Falcon-40B, and Code Llama.

## What's Changed
* fix "tansformers_module" ModuleNotFoundError when load model with `trust_remote_code=True` by @Jingru in https://github.com/vllm-project/vllm/pull/871
* Fix wrong dtype in PagedAttentionWithALiBi bias by @Yard1 in https://github.com/vllm-project/vllm/pull/996
* fix: CUDA error when inferencing with Falcon-40B base model by @kyujin-cho in https://github.com/vllm-project/vllm/pull/992
* [Docs] Update installation page by @WoosukKwon in https://github.com/vllm-project/vllm/pull/1005
* Update setup.py by @WoosukKwon in https://github.com/vllm-project/vllm/pull/1006
* Use FP32 in RoPE initialization by @WoosukKwon in https://github.com/vllm-project/vllm/pull/1004
* Bump up the version to v0.1.7 by @WoosukKwon in https://github.com/vllm-project/vllm/pull/1013

## New Contributors
* @Jingru made their first contribution in https://github.com/vllm-project/vllm/pull/871
* @kyujin-cho made their first contribution in https://github.com/vllm-project/vllm/pull/992

**Full Changelog**: https://github.com/vllm-project/vllm/compare/v0.1.6...v0.1.7