[Misc] Upgrade vllm hash to 12_14 (#5000)

### What this PR does / why we need it?

### Does this PR introduce _any_ user-facing change?
1. fix vllm-project/vllm#27938
2. fix vllm-project/vllm#27145
pooling models now supports chunked prefill and prefix caching,
3. fix vllm-project/vllm#30181
define the CPU fields in the field config where they really belong.
4. fix vllm-project/vllm#28168
define the CPU fields in the field config where they really belong.
5. fix vllm-project/vllm#30201
some moudle rename
6. fix vllm-project/vllm#29067
fusedmoe moudle refactor
7. fix vllm-project/vllm#29066
fusedmoe moudle refactor
8. fix vllm-project/vllm#29624
### How was this patch tested?

- vLLM version: v0.12.0
- vLLM main:
vllm-project/vllm@ad32e3e

---------

Signed-off-by: wangli <[email protected]>

Author: Potabk

.github/workflows/pr_test_full.yaml

@@ -74,7 +74,7 @@ jobs:
     name: e2e-full
     strategy:
       matrix:
-        vllm_version: [ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9, v0.12.0]
+        vllm_version: [97f2f160fda2805f9149b0e44da76b5d3b1f7c7e, v0.12.0]
     needs: [changes]
     if: ${{ needs.changes.outputs.e2e_tracker == 'true' }}
     uses: ./.github/workflows/_e2e_test.yaml

.github/workflows/pr_test_light.yaml

@@ -42,7 +42,7 @@ jobs:
   lint:
     uses: ./.github/workflows/_pre_commit.yml
     with:
-      vllm: ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9
+      vllm: 97f2f160fda2805f9149b0e44da76b5d3b1f7c7e
   changes:
     runs-on: linux-aarch64-a2-0
     outputs:
@@ -90,7 +90,7 @@ jobs:
         SOC_VERSION: ascend910b1
     strategy:
       matrix:
-        vllm_version: [ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9, v0.12.0]
+        vllm_version: [97f2f160fda2805f9149b0e44da76b5d3b1f7c7e, v0.12.0]
 
     steps:
       - name: Free up disk space
@@ -154,7 +154,7 @@ jobs:
     name: e2e-light
     strategy:
       matrix:
-        vllm_version: [ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9, v0.12.0]
+        vllm_version: [97f2f160fda2805f9149b0e44da76b5d3b1f7c7e, v0.12.0]
     # Note (yikun): If CI resource are limited we can split job into two chain jobs
     needs: [lint, changes]
     # only trigger e2e test after lint passed and the change is e2e related with pull request.

docs/source/community/versioning_policy.md

@@ -45,7 +45,7 @@ The table below is the release compatibility matrix for vLLM Ascend release.
 For main branch of vLLM Ascend, we usually make it compatible with the latest vLLM release and a newer commit hash of vLLM. Please note that this table is usually updated. Please check it regularly.
 | vLLM Ascend | vLLM         | Python           | Stable CANN | PyTorch/torch_npu  |
 |-------------|--------------|------------------|-------------|--------------------|
-|     main    | ad32e3e19ccf0526cb6744a5fed09a138a5fb2f9,  v0.12.0 tag | >= 3.10, < 3.12   | 8.3.RC2 | 2.8.0 / 2.8.0 |
+|     main    | 97f2f160fda2805f9149b0e44da76b5d3b1f7c7e,  v0.12.0 tag | >= 3.10, < 3.12   | 8.3.RC2 | 2.8.0 / 2.8.0 |
 
 ## Release cadence
 

tests/ut/compilation/test_acl_graph.py

@@ -803,7 +803,9 @@ def test_update_mla_dcp_pcp_params(self, _mock_graph_task_end):
             (q_nope, q_pe, k_nope, k_pe, block_table, seq_lens, num_heads,
              scale, num_kv_heads, out, lse))
 
-        update_mla_attn_dcp_pcp_params(self.update_stream, forward_context, 4)
+        with patch("torch_npu._C._npu_setStream", return_value=None):
+            update_mla_attn_dcp_pcp_params(self.update_stream, forward_context,
+                                           4)
 
         _mock_graph_task_end.assert_called_once()
 
@@ -842,6 +844,7 @@ def test_update_attn_dcp_pcp_params(self, _mock_graph_task_end):
              block_table, 128, actual_seq_lengths_kv, actual_seq_lengths_q,
              out, lse, 2, 0, 0))
 
-        update_attn_dcp_pcp_params(self.update_stream, forward_context, 4)
+        with patch("torch_npu._C._npu_setStream", return_value=None):
+            update_attn_dcp_pcp_params(self.update_stream, forward_context, 4)
 
         _mock_graph_task_end.assert_called_once()

tests/ut/spec_decode/test_eagle_proposer.py

@@ -95,6 +95,8 @@ def test_load_model_pp1(self, mock_pp_group, mock_get_model,
         mock_model = MagicMock()
         mock_model.model.embed_tokens = MagicMock()
         mock_model.lm_head = MagicMock()
+        mock_model.multimodal_cpu_fields = None
+        mock_model.merge_by_field_config = None
         mock_get_model.return_value = MagicMock()
         self.proposer.name = SpecDcodeType.EAGLE
 
@@ -117,6 +119,8 @@ def test_load_model_pp_gt1(self, mock_pp_group, mock_get_model,
 
         mock_model = MagicMock()
         original_embed = MagicMock()
+        mock_model.multimodal_cpu_fields = None
+        mock_model.merge_by_field_config = None
         mock_get_model.return_value = MagicMock(model=MagicMock(
             embed_tokens=original_embed))