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 14, 05:30 UTC
Re-enable Xet on TPU tests now that `hf_xet` has been updated (#22666)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Harry Mellor committed
Aug 12, 2025 at 03:54 UTC
839ab0034932e5e6863a8d837e5b04944fa0cac5
3 files changed
+2
-5
.buildkite/scripts/hardware_ci/run-tpu-v1-test-part2.sh
+1
-1
@@ -128,7 +128,7 @@ run_and_track_test() {
128
129
# --- Actual Test Execution ---
130
run_and_track_test 1 "test_struct_output_generate.py" \
131
- "HF_HUB_DISABLE_XET=1 python3 -m pytest -s -v /workspace/vllm/tests/v1/entrypoints/llm/test_struct_output_generate.py -k \"not test_structured_output_with_reasoning_matrices\""
131
+ "python3 -m pytest -s -v /workspace/vllm/tests/v1/entrypoints/llm/test_struct_output_generate.py -k \"not test_structured_output_with_reasoning_matrices\""
132
run_and_track_test 2 "test_moe_pallas.py" \
133
"python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_moe_pallas.py"
134
run_and_track_test 3 "test_lora.py" \
.buildkite/scripts/hardware_ci/run-tpu-v1-test.sh
+1
-1
@@ -134,7 +134,7 @@ run_and_track_test 1 "test_compilation.py" \
134
run_and_track_test 2 "test_basic.py" \
135
"python3 -m pytest -s -v /workspace/vllm/tests/v1/tpu/test_basic.py"
136
run_and_track_test 3 "test_accuracy.py::test_lm_eval_accuracy_v1_engine" \
137
- "HF_HUB_DISABLE_XET=1 python3 -m pytest -s -v /workspace/vllm/tests/entrypoints/llm/test_accuracy.py::test_lm_eval_accuracy_v1_engine"
137
+ "python3 -m pytest -s -v /workspace/vllm/tests/entrypoints/llm/test_accuracy.py::test_lm_eval_accuracy_v1_engine"
138
run_and_track_test 4 "test_quantization_accuracy.py" \
139
"python3 -m pytest -s -v /workspace/vllm/tests/tpu/test_quantization_accuracy.py"
140
run_and_track_test 5 "examples/offline_inference/tpu.py" \
tests/entrypoints/llm/test_accuracy.py
-3
@@ -96,9 +96,6 @@ def test_lm_eval_accuracy_v1_engine_fp8_kv_cache(
96
more_args = None
97
if current_platform.is_tpu():
98
# Limit compilation time for TPU V1
99
-
100
- # xet doesn't work well for Qwen/Qwen3-1.7B
101
- m.setenv("HF_HUB_DISABLE_XET", "1")
99
more_args = "max_model_len=2048,max_num_seqs=128,kv_cache_dtype=fp8"
100
101
# Add TP test (if provided)