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, 00:30 UTC
[CI Failure] Fix test_kv_cache_model_load_and_run (#27717)
Signed-off-by: Huamin Li <3ericli@gmail.com>
Huamin Li committed
Oct 30, 2025 at 05:27 UTC
1994de99ea0bf8dd84257a19800f4f62526a7edf
1 file changed
+12
-1
tests/quantization/test_fp8.py
+12
-1
@@ -49,7 +49,18 @@ def test_model_load_and_run(
49
50
KV_CACHE_MODELS = [
51
# AutoFP8 format using separate .k_scale and .v_scale
52
- "nm-testing/Qwen2-1.5B-Instruct-FP8-K-V",
52
+ # The original checkpoint below was removed from the Hub. To unblock CI and
53
+ # until a small replacement with split K/V scales is found, skip this case.
54
+ # See PR #27717 for context.
55
+ pytest.param(
56
+ "nm-testing/Qwen2-1.5B-Instruct-FP8-K-V",
57
+ marks=pytest.mark.skip(
58
+ reason=(
59
+ "Checkpoint removed from HF; temporarily disabling this "
60
+ "AutoFP8 split K/V case (PR #27717)."
61
+ )
62
+ ),
63
+ ),
64
]
65
66