[Feature] Enable inference support for Deepseekr1-w8a8-MTP (#1994)

Support the inference of the Deepseekr1-w8a8-mtp model with
statically-quantized shared_head in MTP layers.

- vLLM version: v0.9.2
- vLLM main:
6eca337ce0

Signed-off-by: curryliu <120010041@link.cuhk.edu.cn>

vllm_ascend/models/deepseek_v2.py

@@ -868,7 +868,9 @@ class CustomDeepseekV2ForCausalLM(DeepseekV2ForCausalLM):
         if get_pp_group().is_last_rank:
             self.lm_head = ParallelLMHead(config.vocab_size,
                                           config.hidden_size,
-                                          quant_config=quant_config)
+                                          quant_config=quant_config,
+                                          prefix=maybe_prefix(
+                                              prefix, "lm_head"))
         else:
             self.lm_head = PPMissingLayer()
         self.logits_processor = LogitsProcessor(config.vocab_size)