Merge branch 'master' into loadams/pyproject-toml

deepspeedai · Feb 26, 2025 · 73f8c71 · 73f8c71
2 parents fa4d66b + f0401ad
commit 73f8c71
Showing 1 changed file with 2 additions and 1 deletion.
diff --git a/deepspeed/module_inject/auto_tp.py b/deepspeed/module_inject/auto_tp.py
@@ -137,7 +137,8 @@ def is_load_module(module):
             "LPLayerNorm", "SharedEmbedding", "OPTLearnedPositionalEmbedding", "LlamaRMSNorm", "FalconLinear",
             "MistralRMSNorm", "T5LayerNorm", "MixtralRMSNorm", "Phi3RotaryEmbedding", "Phi3SuScaledRotaryEmbedding",
             "Phi3RMSNorm", "YuanRMSNorm", "YuanRotaryEmbedding", "Phi3LongRoPEScaledRotaryEmbedding", "Qwen2RMSNorm",
-            "DeepseekV2RMSNorm", "DeepseekV2YarnRotaryEmbedding", "MoEGate"
+            "DeepseekV2RMSNorm", "DeepseekV3RMSNorm", "DeepseekV2YarnRotaryEmbedding", "DeepseekV3YarnRotaryEmbedding",
+            "MoEGate"
         ]
         return module.__class__ in load_layers or module._get_name() in load_layer_names