Unverified Commit 5d16d0fa authored by Chen Zhang's avatar Chen Zhang Committed by GitHub
Browse files

[DCP] check return_lse for all layers in dcp (#27929)


Signed-off-by: default avatarChen Zhang <zhangch99@outlook.com>
parent 0606bea2
......@@ -4680,10 +4680,7 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
kv_transfer_group.set_host_xfer_buffer_ops(copy_kv_blocks)
if self.dcp_world_size > 1:
layer_names = self.attn_groups[0][0].layer_names
layers = get_layers_from_vllm_config(
self.vllm_config, AttentionLayerBase, layer_names
)
layers = get_layers_from_vllm_config(self.vllm_config, AttentionLayerBase)
for layer in layers.values():
assert layer.impl.need_to_return_lse_for_decode, (
"DCP requires attention impls to return"
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment