[projects][sglang] Add

.ci/options.ini

@@ -3,12 +3,14 @@ automatic = https://github.com/vladmandic/automatic
 comfyui = https://github.com/comfyanonymous/ComfyUI
 llamacpp = https://github.com/ggerganov/llama.cpp
 nvidia_pstated = https://github.com/sasha0552/nvidia-pstated
+sglang = https://github.com/sgl-project/sglang
 
 [revisions]
 automatic = e7ec07f9783701629ca1411ad82aec87232501b9
 comfyui = v0.2.2
-llamacpp = b3828
+llamacpp = b3883
 nvidia_pstated = v1.0.5
+sglang = v0.3.2
 
 [files]
 nvidia_pstated = nvidia-pstated
@@ -17,4 +19,5 @@ nvidia_pstated = nvidia-pstated
 llamacpp = release
 comfyui = release
 nvidia_pstated = release+first
+sglang = release
 

airootfs/root/customize_airootfs/scripts/1000-sglang.sh.jinja2

@@ -0,0 +1,25 @@
+#!/bin/sh
+set -eu
+
+{% if platform == "cuda" %}
+# clone repository
+git clone "{{ repositories.sglang }}.git"
+
+# sglang dependencies
+pushd "sglang"
+  # create venv
+  python3 -m venv venv
+
+  # activate venv
+  source venv/bin/activate
+    # use repo with patched packages
+    export PIP_EXTRA_INDEX_URL="https://sasha0552.github.io/pascal-pkgs-ci/"
+
+    # install dependencies
+    pip3 install -e "python[all]"
+
+    # install flashinfer
+    pip3 install flashinfer-pascal
+  deactivate
+popd
+{% endif %}

airootfs/root/customize_airootfs/scripts/1000-vllm.sh.jinja2

@@ -7,4 +7,7 @@ export PIP_EXTRA_INDEX_URL="https://sasha0552.github.io/pascal-pkgs-ci/"
 
 # install patched vllm
 pipx install vllm-pascal
+
+# install flashinfer
+pipx inject vllm-pascal flashinfer-pascal
 {% endif %}