Skip to content

Commit a2e676b

Browse files
committed
fix qwen3_moe switch_glu object
1 parent 069aa50 commit a2e676b

File tree

2 files changed

+13
-1
lines changed

2 files changed

+13
-1
lines changed

src/dnet/api/catalog.py

Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -204,5 +204,17 @@
204204
"quantization": "4bit",
205205
"alias": "qwen3-moe-30b",
206206
},
207+
{
208+
"id": "mlx-community/Qwen3-Coder-30B-A3B-Instruct-4bit",
209+
"arch": "qwen3_moe",
210+
"quantization": "4bit",
211+
"alias": "qwen3-moe-30b",
212+
},
213+
{
214+
"id": "mlx-community/Qwen3-Coder-30B-A3B-Instruct-8bit",
215+
"arch": "qwen3_moe",
216+
"quantization": "8bit",
217+
"alias": "qwen3-moe-30b",
218+
},
207219
]
208220
}

src/dnet/core/models/qwen3_moe.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -14,7 +14,7 @@ def __init__(self, args: ModelArgs):
1414
self.num_experts = args.num_experts
1515
self.gate = nn.Linear(args.hidden_size, self.num_experts, bias=False)
1616
self.switch_mlp = SwitchGLU(
17-
args.hidden_size, args.intermediate_size, self.num_experts
17+
args.hidden_size, args.moe_intermediate_size, self.num_experts
1818
)
1919

2020
def __call__(self, x: mx.array):

0 commit comments

Comments
 (0)