From 63fcce4de1563309ea5195ba98d0a2e1ba4f5831 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Micha=C5=82=20Ganczarenko?= Date: Wed, 1 Jul 2026 17:39:12 +0300 Subject: [PATCH] [Bugfix] Fix GraniteMoeShared weight loading broken by #41184 (#47031) Signed-off-by: Co-authored-by: Kunshang Ji --- vllm/model_executor/models/granitemoeshared.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/vllm/model_executor/models/granitemoeshared.py b/vllm/model_executor/models/granitemoeshared.py index 7c8a92b88d..7abc682c58 100644 --- a/vllm/model_executor/models/granitemoeshared.py +++ b/vllm/model_executor/models/granitemoeshared.py @@ -214,11 +214,11 @@ class GraniteMoeSharedModel(nn.Module): for e in range(p.size(0)): w1_name = n.replace( ".block_sparse_moe.input_linear.weight", - f".block_sparse_moe.experts.routed_experts.{e}.w1.weight", + f".block_sparse_moe.experts.{e}.w1.weight", ) w3_name = n.replace( ".block_sparse_moe.input_linear.weight", - f".block_sparse_moe.experts.routed_experts.{e}.w3.weight", + f".block_sparse_moe.experts.{e}.w3.weight", ) w1_param, w3_param = p[e].chunk(2, dim=0) assert w1_name not in new_weights @@ -229,7 +229,7 @@ class GraniteMoeSharedModel(nn.Module): for e in range(p.size(0)): w2_name = n.replace( ".block_sparse_moe.output_linear.weight", - f".block_sparse_moe.experts.routed_experts.{e}.w2.weight", + f".block_sparse_moe.experts.{e}.w2.weight", ) w2_param = p[e] assert w2_name not in new_weights