[Frontend][Bugfix] support prefill decode disaggregation on deepseek (#14824)

Signed-off-by: billishyahao <bill.he@amd.com>
Co-authored-by: Zhai Feiyue <80079571+ZhaiFeiyue@users.noreply.github.com>

vllm/model_executor/models/deepseek_v2.py

@@ -589,6 +589,7 @@ class DeepseekV2Model(nn.Module):
         model_config = vllm_config.model_config
         cache_config = vllm_config.cache_config
         quant_config = vllm_config.quant_config
+        self.config = config
 
         self.vocab_size = config.vocab_size