lamm-mit
/

Cephalo-Phi-3-MoE-vision-128k-3x4b-beta

Model card Files Files and versions Community

mjbuehler commited on Jun 1

Commit

cd35978

•

1 Parent(s): 1e0f2e3

Upload moe_phi3_v.py

Files changed (1) hide show

moe_phi3_v.py +1 -1

moe_phi3_v.py CHANGED Viewed

@@ -347,7 +347,7 @@ class Phi3VForCausalLMMoE(Phi3VForCausalLM):
                     return self.gate(x)
             gating_layer = SimpleGatingLayer(self.config.hidden_size, num_experts, layer_dtype=self.layer_dtype).to(self.custom_device)
-            print(f"Now train layer={layer_idx}...")
             criterion = nn.CrossEntropyLoss()
             optimizer = Adam(gating_layer.parameters(), lr=lr)

                     return self.gate(x)
             gating_layer = SimpleGatingLayer(self.config.hidden_size, num_experts, layer_dtype=self.layer_dtype).to(self.custom_device)
             criterion = nn.CrossEntropyLoss()
             optimizer = Adam(gating_layer.parameters(), lr=lr)