Skip to content

Commit 4bf0a10

Browse files
committed
formatting
Summary Signed-off-by: HDCharles <charlesdavidhernandez@gmail.com>
1 parent 1a4da12 commit 4bf0a10

File tree

4 files changed

+20
-5
lines changed

4 files changed

+20
-5
lines changed
Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,7 @@
11
cadence: "nightly"
22
test_type: "regression"
33
model: Qwen/Qwen3-30B-A3B
4-
54
scheme: NVFP4
6-
75
dataset_id: HuggingFaceH4/ultrachat_200k
86
dataset_split: train_sft
97
num_calibration_samples: 20
Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,4 @@
11
cadence: "nightly"
22
test_type: "regression"
33
model: Qwen/Qwen3-30B-A3B
4-
54
scheme: FP8_DYNAMIC
Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,7 @@
11
cadence: "nightly"
22
test_type: "regression"
33
model: Qwen/Qwen3-30B-A3B
4-
54
dataset_id: HuggingFaceH4/ultrachat_200k
65
dataset_split: train_sft
76
num_calibration_samples: 20
8-
97
recipe: tests/e2e/vLLM/recipes/WNA16/recipe_w4a16_group_quant_first_10_layers.yaml
Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,20 @@
1+
quant_stage:
2+
quant_modifiers:
3+
GPTQModifier:
4+
ignore: [
5+
"lm_head",
6+
# Ignore layers (10+)
7+
"re:.*model\\.layers\\.([1-9][0-9])\\..*",
8+
]
9+
actorder: null
10+
config_groups:
11+
group_0:
12+
weights:
13+
num_bits: 4
14+
type: "int"
15+
symmetric: False
16+
strategy: "group"
17+
group_size: 128
18+
input_activations: null
19+
output_activations: null
20+
targets: ["Linear"]

0 commit comments

Comments
 (0)