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 16, 23:15 UTC
[CI/Build] Disable test_fp8.py (#5508)
Tyler Michael Smith committed
Jun 13, 2024 at 16:37 UTC
33e3b372429232cea44266d866906effaa705a10
1 file changed
+8
tests/models/test_fp8.py
+8
@@ -68,6 +68,14 @@ EXPECTED_STRS_MAP = {
68
}
69
70
71
+# This test compares against golden strings for exact match since
72
+# there is no baseline implementation to compare against
73
+# and is unstable w.r.t specifics of the fp8 implementation or
74
+# the hardware being run on.
75
+# Disabled to prevent it from breaking the build
76
+@pytest.mark.skip(
77
+ reason=
78
+ "Prevent unstable test based on golden strings from breaking the build.")
79
@pytest.mark.skipif(not is_quant_method_supported("fp8"),
80
reason="fp8 is not supported on this GPU type.")
81
@pytest.mark.parametrize("model_name", MODELS)