[CI] Add nightly ci test for deepseek v3.1

.github/workflows/nightly_test_a3.yaml

@@ -86,6 +86,9 @@ jobs:
           - name: multi-node-kimi-k2-instruct-w8a8
             config_file_path: Kimi-K2-Instruct-W8A8.yaml
             size: 2
+          - name: multi-node-deepseek-v3.1
+            config_file_path: DeepSeek-V3.1-BF16.yaml
+            size: 2
     uses: ./.github/workflows/_e2e_nightly_multi_node.yaml
     with:
       soc_version: a3

tests/e2e/nightly/multi_node/config/DeepSeek-V3.1-BF16.yaml

@@ -0,0 +1,82 @@
+test_name: "test DeepSeek-V3.1-BF16 on A3"
+model: "unsloth/DeepSeek-V3.1-BF16"
+num_nodes: 2
+npu_per_node: 16
+env_common:
+  VLLM_USE_MODELSCOPE: true
+  HCCL_BUFFSIZE: 2048
+  SERVER_PORT: 8080
+  OMP_PROC_BIND: false
+  VLLM_ASCEND_ENABLE_FLASHCOMM1: 1
+  PYTORCH_NPU_ALLOC_CONF: expandable_segments:True
+  OMP_NUM_THREADS: 1
+  VLLM_ASCEND_ENABLE_MLAPO: 1
+  VLLM_ASCEND_BALANCE_SCHEDULING: 1
+  HCCL_INTRA_PCIE_ENABLE: 1
+  HCCL_INTRA_ROCE_ENABLE: 0
+
+deployment:
+  -
+    server_cmd: >
+      vllm serve unsloth/DeepSeek-V3.1-BF16
+      --host 0.0.0.0
+      --port $SERVER_PORT
+      --data-parallel-size 4
+      --tensor-parallel-size 8
+      --data-parallel-size-local 2
+      --data-parallel-address $LOCAL_IP
+      --data-parallel-rpc-port 13399
+      --no-enable-prefix-caching
+      --max-num-seqs 16
+      --max-model-len 8192
+      --max-num-batched-tokens 4096
+      --enable-expert-parallel
+      --trust-remote-code
+      --gpu-memory-utilization 0.95
+      --async-scheduling
+      --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
+      --compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY", "cudagraph_capture_sizes":[2, 4, 8, 16, 32]}'
+      --additional_config '{"enable_multistream_moe": true}'
+
+  -
+    server_cmd: >
+      vllm serve unsloth/DeepSeek-V3.1-BF16
+      --headless
+      --data-parallel-size 4
+      --tensor-parallel-size 8
+      --data-parallel-size-local 2
+      --data-parallel-start-rank 2
+      --data-parallel-address $MASTER_IP
+      --data-parallel-rpc-port 13399
+      --no-enable-prefix-caching
+      --max-num-seqs 16
+      --max-model-len 8192
+      --max-num-batched-tokens 4096
+      --enable-expert-parallel
+      --trust-remote-code
+      --gpu-memory-utilization 0.95
+      --async-scheduling
+      --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
+      --compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY", "cudagraph_capture_sizes":[2, 4, 8, 16, 32]}'
+      --additional_config '{"enable_multistream_moe": true}'
+benchmarks:
+  perf:
+    case_type: performance
+    dataset_path: vllm-ascend/GSM8K-in3500-bs2800
+    request_conf: vllm_api_stream_chat
+    dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
+    num_prompts: 512
+    max_out_len: 512
+    batch_size: 700
+    request_rate: 11.2
+    baseline: 1
+    threshold: 0.97
+  acc:
+    case_type: accuracy
+    dataset_path: vllm-ascend/gsm8k-lite
+    request_conf: vllm_api_general_chat
+    dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_chat_prompt
+    max_out_len: 4096
+    batch_size: 512
+    baseline: 95
+    threshold: 5