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 13, 17:00 UTC
Disable nm-testing models with issues in CI (#28206)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Nov 6, 2025 at 22:19 UTC
f32229293e3c779004d9f90503b4375a3716879d
3 files changed
+4
-16
.buildkite/lm-eval-harness/configs/Qwen2-1.5B-Instruct-W8A16-compressed-tensors.yaml
deleted
-12
@@ -1,12 +0,0 @@
1
-# For vllm script, with -t option (tensor parallel size).
2
-# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m nm-testing/Qwen2-1.5B-Instruct-W8A16-Channelwise -b "auto" -l 1000 -f 5 -t 1
3
-model_name: "nm-testing/Qwen2-1.5B-Instruct-W8A16-Channelwise"
4
-tasks:
5
-- name: "gsm8k"
6
- metrics:
7
- - name: "exact_match,strict-match"
8
- value: 0.595
9
- - name: "exact_match,flexible-extract"
10
- value: 0.582
11
-limit: 1000
12
-num_fewshot: 5
benchmarks/kernels/benchmark_grouped_gemm_cutlass.py
+2
-2
@@ -16,8 +16,8 @@ from vllm.model_executor.layers.fused_moe.fused_moe import (
16
from vllm.utils.argparse_utils import FlexibleArgumentParser
17
18
DEFAULT_MODELS = [
19
- "nm-testing/Mixtral-8x7B-Instruct-v0.1",
20
- "nm-testing/deepseekv2-lite",
19
+ "mistralai/Mixtral-8x7B-Instruct-v0.1",
20
+ "deepseek-ai/DeepSeek-V2-Lite",
21
"ibm-granite/granite-3.0-1b-a400m",
22
"ibm-granite/granite-3.0-3b-a800m",
23
]
benchmarks/kernels/benchmark_shapes.py
+2
-2
@@ -78,11 +78,11 @@ WEIGHT_SHAPES = {
78
}
79
80
WEIGHT_SHAPES_MOE = {
81
- "nm-testing/Mixtral-8x7B-Instruct-v0.1": [
81
+ "mistralai/Mixtral-8x7B-Instruct-v0.1": [
82
[8, 2, 4096, 28672],
83
[8, 2, 14336, 4096],
84
],
85
- "nm-testing/deepseekv2-lite": [
85
+ "deepseek-ai/DeepSeek-V2-Lite": [
86
[64, 6, 2048, 1408],
87
],
88
"ibm-granite/granite-3.0-1b-a400m": [