We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent bd66b85 commit ba09652Copy full SHA for ba09652
vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_moe.py
@@ -142,7 +142,10 @@ def get_moe_method(
142
# group_size=None means channelwise
143
group_size = weight_quant.group_size or -1
144
# Prefer to use the MarlinMoE kernel when it is supported.
145
- if not check_moe_marlin_supports_layer(layer, group_size):
+ if (
146
+ not check_moe_marlin_supports_layer(layer, group_size)
147
+ or current_platform.is_rocm()
148
+ ):
149
if (
150
weight_quant.strategy == QuantizationStrategy.GROUP
151
and weight_quant.actorder
0 commit comments