Skip to content

Commit ba09652

Browse files
authored
[ROCM] Enable CompressedTensorsWNA16 (#27187)
Signed-off-by: JartX <[email protected]>
1 parent bd66b85 commit ba09652

File tree

1 file changed

+4
-1
lines changed

1 file changed

+4
-1
lines changed

vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -142,7 +142,10 @@ def get_moe_method(
142142
# group_size=None means channelwise
143143
group_size = weight_quant.group_size or -1
144144
# Prefer to use the MarlinMoE kernel when it is supported.
145-
if not check_moe_marlin_supports_layer(layer, group_size):
145+
if (
146+
not check_moe_marlin_supports_layer(layer, group_size)
147+
or current_platform.is_rocm()
148+
):
146149
if (
147150
weight_quant.strategy == QuantizationStrategy.GROUP
148151
and weight_quant.actorder

0 commit comments

Comments
 (0)