[main] flashcomm_v1 optim in Qwen Dense Models (#2802)

### What this PR does / why we need it?
Flashcomm_v1 optim in Qwen Dense Models.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
CI passed with new added/existing test.

- vLLM version: v0.10.1.1
- vLLM main:
5e537f45b4

Co-authored-by: 1024daniel <xxltju324@gmail.com>

vllm_ascend/worker/model_runner_v1.py

@@ -37,6 +37,7 @@ from vllm.attention.layer import Attention
 from vllm.compilation.counter import compilation_counter
 from vllm.compilation.monitor import set_cudagraph_capturing_enabled
 from vllm.config import CompilationLevel, CUDAGraphMode, VllmConfig
+from vllm.distributed import tensor_model_parallel_all_gather
 from vllm.distributed.kv_transfer import (get_kv_transfer_group,
                                           has_kv_transfer_group)
 from vllm.distributed.kv_transfer.kv_connector.v1 import KVConnectorBase_V1
@@ -1182,6 +1183,11 @@ class NPUModelRunner(LoRAModelRunnerMixin):
             intermediate_tensors=intermediate_tensors,
             inputs_embeds=inputs_embeds,
         )
+        if get_forward_context().flashcomm_v1_enabled:
+            hidden_states = tensor_model_parallel_all_gather(hidden_states, 0)
+            pad_size = get_forward_context().pad_size
+            if pad_size > 0:
+                hidden_states = hidden_states[:-pad_size, :]
         return hidden_states
 
     def _build_attn_state(self, num_reqs, num_scheduled_tokens,