fix coverity scan issue (#3465)

Author: blzheng

intel_extension_for_pytorch/transformers/models/reference/models.py

@@ -829,7 +829,7 @@ def LlavaForConditionalGeneration_forward(
             )
 
             # Get the target length
-            target_length = input_ids.shape[1]
+            target_length = inputs_embeds.shape[1]
             past_length = first_layer_past_key_value.shape[-1]
 
             extended_attention_mask = torch.ones(

tests/cpu/hf_configs/deepseekv3/modeling_deepseek.py

@@ -531,8 +531,8 @@ def forward(self, hidden_states):
         topk_idx, topk_weight = self.gate(hidden_states)
         hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
         flat_topk_idx = topk_idx.view(-1)
-        if not self.training:
-            y = self.moe_infer(hidden_states, topk_idx, topk_weight).view(*orig_shape)
+        assert not self.training
+        y = self.moe_infer(hidden_states, topk_idx, topk_weight).view(*orig_shape)
         if self.config.n_shared_experts is not None:
             y = y + self.shared_experts(identity)
         return y