A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 01:00 UTC
[TPU] Start using python 3.12 (#21000)
Signed-off-by: Xiongfei Wei <isaacwxf23@gmail.com>
XiongfeiWei committed
Jul 16, 2025 at 19:37 UTC
58760e12b18f8919e265f08c45a6364074280708
4 files changed
+9
-10
.buildkite/scripts/hardware_ci/run-tpu-v1-test.sh
+1
-1
@@ -70,7 +70,7 @@ export VLLM_XLA_CACHE_PATH=
70
echo "Using VLLM V1"
71
72
echo "--- Hardware Information ---"
73
-tpu-info
73
+# tpu-info
74
echo "--- Starting Tests ---"
75
set +e
76
overall_script_exit_code=0
docker/Dockerfile.tpu
+2
-2
@@ -1,5 +1,5 @@
1
-ARG NIGHTLY_DATE="20250124"
2
-ARG BASE_IMAGE="us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla:nightly_3.10_tpuvm_$NIGHTLY_DATE"
1
+ARG NIGHTLY_DATE="20250714"
2
+ARG BASE_IMAGE="us-central1-docker.pkg.dev/tpu-pytorch-releases/docker/xla:nightly_3.12_tpuvm_$NIGHTLY_DATE"
3
4
FROM $BASE_IMAGE
5
WORKDIR /workspace/vllm
docs/getting_started/installation/google_tpu.md
+2
-2
@@ -37,7 +37,7 @@ information, see [Storage options for Cloud TPU data](https://cloud.devsite.corp
37
38
- Google Cloud TPU VM
39
- TPU versions: v6e, v5e, v5p, v4
40
-- Python: 3.10 or newer
40
+- Python: 3.11 or newer
41
42
### Provision Cloud TPUs
43
@@ -117,7 +117,7 @@ source ~/.bashrc
117
Create and activate a Conda environment for vLLM:
118
119
```bash
120
-conda create -n vllm python=3.10 -y
120
+conda create -n vllm python=3.12 -y
121
conda activate vllm
122
```
123
requirements/tpu.txt
+4
-5
@@ -18,9 +18,8 @@ setuptools==78.1.0
18
--find-links https://storage.googleapis.com/libtpu-releases/index.html
19
--find-links https://storage.googleapis.com/jax-releases/jax_nightly_releases.html
20
--find-links https://storage.googleapis.com/jax-releases/jaxlib_nightly_releases.html
21
-torch==2.9.0.dev20250711
22
-torchvision==0.24.0.dev20250711
23
-torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.9.0.dev20250711-cp39-cp39-linux_x86_64.whl ; python_version == "3.9"
24
-torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.9.0.dev20250711-cp310-cp310-linux_x86_64.whl ; python_version == "3.10"
25
-torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.9.0.dev20250711-cp311-cp311-linux_x86_64.whl ; python_version == "3.11"
21
+torch==2.9.0.dev20250716
22
+torchvision==0.24.0.dev20250716
23
+torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.9.0.dev20250716-cp311-cp311-linux_x86_64.whl ; python_version == "3.11"
24
+torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.9.0.dev20250716-cp312-cp312-linux_x86_64.whl ; python_version == "3.12"
25