From 891b2bfe715619affac64c6fa339944d47ebcf1a Mon Sep 17 00:00:00 2001
From: Icey <1790571317@qq.com>
Date: Mon, 25 Aug 2025 09:39:30 +0800
Subject: [PATCH] Accuracy report formatting (#2279)

### What this PR does / why we need it?
Accuracy report formatting

### Does this PR introduce _any_ user-facing change?
no

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


- vLLM version: v0.10.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/53415653ff24be03e7c90f5b42ef9cb3f72aad71

---------

Signed-off-by: Icey <1790571317@qq.com>
---
 tests/e2e/models/report_template.md          | 17 +++++++----------
 tests/e2e/models/test_lm_eval_correctness.py | 11 ++++++++---
 2 files changed, 15 insertions(+), 13 deletions(-)

diff --git a/tests/e2e/models/report_template.md b/tests/e2e/models/report_template.md
index ddaa9c7..8402545 100644
--- a/tests/e2e/models/report_template.md
+++ b/tests/e2e/models/report_template.md
@@ -1,24 +1,21 @@
 # {{ model_name }}
 
-**vLLM Version**: vLLM: {{ vllm_version }} ([{{ vllm_commit[:7] }}](https://github.com/vllm-project/vllm/commit/{{ vllm_commit }})),
-**vLLM Ascend Version**: {{ vllm_ascend_version }} ([{{ vllm_ascend_commit[:7] }}](https://github.com/vllm-project/vllm-ascend/commit/{{ vllm_ascend_commit }}))  
-**Software Environment**: CANN: {{ cann_version }}, PyTorch: {{ torch_version }}, torch-npu: {{ torch_npu_version }}  
-**Hardware Environment**: Atlas A2 Series  
-**Datasets**: {{ datasets }}  
-**Parallel Mode**: TP  
-**Execution Mode**: ACLGraph  
+- **vLLM Version**: vLLM: {{ vllm_version }} ([{{ vllm_commit[:7] }}](https://github.com/vllm-project/vllm/commit/{{ vllm_commit }})), **vLLM Ascend Version**: {{ vllm_ascend_version }} ([{{ vllm_ascend_commit[:7] }}](https://github.com/vllm-project/vllm-ascend/commit/{{ vllm_ascend_commit }}))  
+- **Software Environment**: **CANN**: {{ cann_version }}, **PyTorch**: {{ torch_version }}, **torch-npu**: {{ torch_npu_version }}  
+- **Hardware Environment**: Atlas A2 Series  
+- **Parallel mode**: {{ parallel_mode }}
+- **Execution mode**: ACLGraph
 
 **Command**:  
 
 ```bash
 export MODEL_ARGS={{ model_args }}
 lm_eval --model {{ model_type }} --model_args $MODEL_ARGS --tasks {{ datasets }} \
---apply_chat_template {{ apply_chat_template }} --fewshot_as_multiturn {{ fewshot_as_multiturn }} {% if num_fewshot is defined and num_fewshot != "N/A" %} --num_fewshot {{ num_fewshot }} {% endif %} \
---limit {{ limit }} --batch_size {{ batch_size}}
+{% if apply_chat_template %} --apply_chat_template {{ apply_chat_template }} {% endif %} {% if fewshot_as_multiturn %} --fewshot_as_multiturn {{ fewshot_as_multiturn }} {% endif %} {% if num_fewshot is defined and num_fewshot != "N/A" %} --num_fewshot {{ num_fewshot }} {% endif %} {% if limit is defined and limit != "N/A" %} --limit {{ limit }} {% endif %} --batch_size {{ batch_size}}
 ```
 
 | Task                  | Metric      | Value     | Stderr |
 |-----------------------|-------------|----------:|-------:|
 {% for row in rows -%}
-| {{ row.task.rjust(23) }} | {{ row.metric.rjust(15) }} |{{ row.value }} | ± {{ "%.4f" | format(row.stderr | float) }} |
+| {{ row.task }} | {{ row.metric }} | {{ row.value }} | ± {{ "%.4f" | format(row.stderr | float) }} |
 {% endfor %}
diff --git a/tests/e2e/models/test_lm_eval_correctness.py b/tests/e2e/models/test_lm_eval_correctness.py
index 567d3de..18768e1 100644
--- a/tests/e2e/models/test_lm_eval_correctness.py
+++ b/tests/e2e/models/test_lm_eval_correctness.py
@@ -65,6 +65,10 @@ def generate_report(tp_size, eval_config, report_data, report_dir, env_config):
     template = env.get_template("report_template.md")
     model_args = build_model_args(eval_config, tp_size)
 
+    parallel_mode = f"TP{model_args.get('tensor_parallel_size', 1)}"
+    if model_args.get('enable_expert_parallel', False):
+        parallel_mode += " + EP"
+
     report_content = template.render(
         vllm_version=env_config.vllm_version,
         vllm_commit=env_config.vllm_commit,
@@ -79,10 +83,11 @@ def generate_report(tp_size, eval_config, report_data, report_dir, env_config):
         datasets=",".join([task["name"] for task in eval_config["tasks"]]),
         apply_chat_template=eval_config.get("apply_chat_template", True),
         fewshot_as_multiturn=eval_config.get("fewshot_as_multiturn", True),
-        limit=eval_config.get("limit", None),
+        limit=eval_config.get("limit", "N/A"),
         batch_size="auto",
         num_fewshot=eval_config.get("num_fewshot", "N/A"),
-        rows=report_data["rows"])
+        rows=report_data["rows"],
+        parallel_mode=parallel_mode)
 
     report_output = os.path.join(
         report_dir, f"{os.path.basename(eval_config['model_name'])}.md")
@@ -123,7 +128,7 @@ def test_lm_eval_correctness_param(config_filename, tp_size, report_dir,
         for metric in task["metrics"]:
             metric_name = metric["name"]
             ground_truth = metric["value"]
-            measured_value = task_result[metric_name]
+            measured_value = round(task_result[metric_name], 4)
             task_success = bool(
                 np.isclose(ground_truth, measured_value, rtol=RTOL))
             success = success and task_success