...

examples/tmuxrunnerexamples/tasks/2/install_vllm.sh

@@ -4,11 +4,14 @@ set -euo pipefail
 source source ../../functions/base.sh
 
 
+curl -LsSf https://astral.sh/uv/install.sh | sh
+
+source $HOME/.local/bin/env
 
 # vllm serve openai/gpt-oss-20b
-# vllm serve openai/gpt-oss-20b --tensor-parallel-size 2
+# vllm serve openai/gpt-oss-20b --tensor-parallel-size 8
  
 # For 120B
-# vllm serve openai/gpt-oss-120b
+# vllm serve openai/gpt-oss-120b --tensor-parallel-size 8
 
 mark_done

examples/tmuxrunnerexamples/tasks/3/vllm.sh

@@ -13,8 +13,13 @@ exit 0
 touch "$0.done"
 exit 0
 
-uv pip install --pre torch==2.9.0.dev20250804+cu128 \
-  --index-url https://download.pytorch.org/whl/nightly/cu128
+cd /root
+
+uv venv
+source .venv/bin/activate
+
+uv pip install --pre torch==2.9.0.dev20250804+cu129 \
+  --index-url https://download.pytorch.org/whl/nightly/cu129
 
 uv pip install tiktoken ipython numpy psutil
 
@@ -22,7 +27,6 @@ uv pip install tiktoken ipython numpy psutil
 python -c "import torch; print(torch.__version__, torch.version.cuda)"
 # 2.9.0.dev20250804+cu128 12.8
 
-cd /root
 source .venv/bin/activate
 uv pip install --upgrade pip setuptools wheel ninja
 export MAX_JOBS=8
@@ -31,12 +35,13 @@ export NCCL_P2P_DISABLE=0
 export NCCL_DEBUG=INFO
 export CUDA_DEVICE_MAX_CONNECTIONS=1
 
-pip install flash-attn --no-build-isolation
+uv pip install vllm --torch-backend=auto
+uv pip install flash-attn --no-build-isolation
 
-uv pip install --pre vllm==0.10.1+gptoss \
-    --extra-index-url https://wheels.vllm.ai/gpt-oss/ \
-    --extra-index-url https://download.pytorch.org/whl/nightly/cu128 \
-    --index-strategy unsafe-best-match
+# uv pip install --pre vllm==0.10.1+gptoss \
+#     --extra-index-url https://wheels.vllm.ai/gpt-oss/ \
+#     --extra-index-url https://download.pytorch.org/whl/nightly/cu128 \
+#     --index-strategy unsafe-best-match
 
 
 mark_done