Revert "Add simple CPU offloading support" (#2252)

We'll re-add the commit to correctly ack Kaichao's authorship

python/sglang/srt/models/olmoe.py

@@ -48,7 +48,6 @@ from sglang.srt.layers.vocab_parallel_embedding import (
     VocabParallelEmbedding,
 )
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch
-from sglang.srt.utils import make_layers
 
 
 class OlmoeMoE(nn.Module):
@@ -262,13 +261,11 @@ class OlmoeModel(nn.Module):
             config.vocab_size,
             config.hidden_size,
         )
-        self.layers = make_layers(
-            config.num_hidden_layers,
-            lambda idx, prefix: OlmoeDecoderLayer(
-                config=config,
-                quant_config=quant_config,
-                layer_id=idx,
-            ),
+        self.layers = nn.ModuleList(
+            [
+                OlmoeDecoderLayer(config, layer_id, quant_config=quant_config)
+                for layer_id in range(config.num_hidden_layers)
+            ]
         )
         self.norm = RMSNorm(config.hidden_size, eps=1e-5)