From c50397ca3a16771a3fa1d3e56205bf1a8533505c Mon Sep 17 00:00:00 2001
From: Paolo Albano <vincenzo.visco@igenius.ai>
Date: Tue, 4 Jun 2024 13:48:44 +0000
Subject: [PATCH] misc: update vllm dependecy to support attention size 160

---
 server/Makefile-vllm | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/server/Makefile-vllm b/server/Makefile-vllm
index 2f2b5ef6..57dcbcae 100644
--- a/server/Makefile-vllm
+++ b/server/Makefile-vllm
@@ -1,9 +1,9 @@
-commit_cuda := b5dfc61db88a81069e45b44f7cc99bd9e62a60fa
+commit_cuda := e0c577263c2c7f367198d4a51a9964136ea259db
 commit_rocm := c6ee53b1be97e3bbc791b95f22827501297f8921
 build-vllm-cuda:
 	if [ ! -d 'vllm' ]; then \
 		pip install -U ninja packaging --no-cache-dir && \
-		git clone https://github.com/Narsil/vllm.git vllm; \
+		git clone https://github.com/igeniusai/vllm.git vllm; \
 	fi
 	cd vllm  && git fetch && git checkout $(commit_cuda) && python setup.py build