Merge pull request #23 from InfiniTensor/feat/dashboard-streamlit

feat: add Streamlit dashboard for InfiniMetrics

Author: Chamberlain0w0

dashboard/__init__.py

@@ -0,0 +1,283 @@
+#!/usr/bin/env python3
+"""Main Streamlit application for InfiniMetrics dashboard."""
+
+import streamlit as st
+import pandas as pd
+from pathlib import Path
+import sys
+from datetime import datetime
+from infinimetrics.common.constants import AcceleratorType
+
+# Add project root to path
+project_root = Path(__file__).parent
+sys.path.append(str(project_root))
+
+from components.header import render_header
+from utils.data_loader import InfiniMetricsDataLoader, load_summary_file
+
+# Page configuration
+st.set_page_config(
+    page_title="InfiniMetrics Dashboard",
+    page_icon="🏭",
+    layout="wide",
+    initial_sidebar_state="expanded",
+)
+
+# Initialize session state
+if "data_loader" not in st.session_state:
+    st.session_state.data_loader = InfiniMetricsDataLoader()
+if "selected_accelerators" not in st.session_state:
+    st.session_state.selected_accelerators = []
+
+
+def main():
+    render_header()
+
+    # =========================
+    # Sidebar
+    # =========================
+
+    with st.sidebar:
+        st.markdown("## ⚙️ 设置")
+
+        results_dir = st.text_input(
+            "测试结果目录", value="./test_output", help="包含 JSON/CSV 测试结果的目录"
+        )
+
+        if results_dir != str(st.session_state.data_loader.results_dir):
+            st.session_state.data_loader = InfiniMetricsDataLoader(results_dir)
+
+        auto_refresh = st.toggle("自动刷新", value=False)
+        if auto_refresh:
+            st.rerun()
+
+        st.markdown("---")
+        st.markdown("## 🧠 筛选条件")
+
+        # Base accelerator types from constants.py
+        ACCELERATOR_OPTIONS = ["cpu"] + [a.value for a in AcceleratorType]
+
+        # UI display names (only labels live here)
+        ACCELERATOR_LABELS = {
+            "cpu": "CPU",
+            AcceleratorType.NVIDIA.value: "NVIDIA",
+            AcceleratorType.AMD.value: "AMD",
+            AcceleratorType.ASCEND.value: "昇腾 NPU",
+            AcceleratorType.CAMBRICON.value: "寒武纪 MLU",
+            AcceleratorType.GENERIC.value: "Generic",
+        }
+
+        selected_accs = st.multiselect(
+            "加速卡类型",
+            options=ACCELERATOR_OPTIONS,
+            default=ACCELERATOR_OPTIONS,
+            format_func=lambda x: ACCELERATOR_LABELS.get(x, x),
+        )
+        st.session_state.selected_accelerators = selected_accs
+
+        run_id_filter = st.text_input("Run ID 模糊搜索")
+        # test_type / testcase filtering will be applied dynamically after runs are loaded
+
+    render_dashboard(run_id_filter)
+
+
+def render_dashboard(run_id_filter: str):
+    st.markdown(
+        """
+        <h1 style="margin-bottom: 0.2em;">
+            📊 综合仪表板
+        </h1>
+        """,
+        unsafe_allow_html=True,
+    )
+
+    st.markdown(
+        """
+        <div style="
+            margin-top: 0.5em;
+            margin-bottom: 1.5em;
+            max-width: 1100px;
+            font-size: 1.05em;
+            line-height: 1.6;
+        ">
+            <strong>InfiniMetrics Dashboard</strong> 用于统一展示
+            <strong>通信（NCCL / 集合通信）</strong>、
+            <strong>推理（Direct / Service）</strong>、
+            <strong>算子（核心算子性能）</strong>
+            等 AI 加速卡性能测试结果。
+            <br/>
+            测试框架输出 <code>JSON</code>（环境 / 配置 / 标量指标） +
+            <code>CSV</code>（曲线 / 时序数据），
+            Dashboard 自动加载并支持多次运行的对比分析与可视化。
+        </div>
+        """,
+        unsafe_allow_html=True,
+    )
+
+    try:
+        runs = st.session_state.data_loader.list_test_runs()
+
+        # ========== Accelerator filtering  ==========
+        selected_accs = st.session_state.get("selected_accelerators", [])
+        if selected_accs:
+            runs = [
+                r
+                for r in runs
+                if set(r.get("accelerator_types", [])) & set(selected_accs)
+            ]
+
+        # ========== run_id filtering  ==========
+        if run_id_filter:
+            runs = [r for r in runs if run_id_filter in r.get("run_id", "")]
+
+        if not runs:
+            st.warning("No test results match the current filters.")
+            return
+
+        # ========== Sort by time (latest first) ==========
+        def _parse_time(t):
+            try:
+                return datetime.fromisoformat(t)
+            except Exception:
+                return datetime.min
+
+        runs = sorted(runs, key=lambda r: _parse_time(r.get("time", "")), reverse=True)
+
+        total = len(runs)
+        success = sum(1 for r in runs if r.get("success"))
+        fail = total - success
+
+        # ========== Categorize runs ==========
+        comm_runs = [r for r in runs if r.get("testcase", "").startswith("comm")]
+        infer_runs = [r for r in runs if r.get("testcase", "").startswith("infer")]
+
+        ops_runs, hw_runs = [], []
+        for r in runs:
+            p = str(r.get("path", "")).replace("\\", "/").lower()
+            tc = (r.get("testcase", "") or "").lower()
+            if "/operators/" in p or tc.startswith(("operator", "operators", "ops")):
+                ops_runs.append(r)
+            if "/hardware/" in p or tc.startswith("hardware"):
+                hw_runs.append(r)
+
+        # ========== KPI ==========
+        c1, c2, c3, c4, c5, c6 = st.columns(6)
+        c1.metric("总测试数", total)
+        c2.metric("成功率", f"{(success/total*100):.1f}%")
+        c3.metric("通信测试", len(comm_runs))
+        c4.metric("推理测试", len(infer_runs))
+        c5.metric("算子测试", len(ops_runs))
+        c6.metric("硬件检测", len(hw_runs))
+
+        st.caption(f"失败测试数：{fail}")
+        st.caption(f"当前筛选：加速卡={','.join(selected_accs) or '全部'}")
+
+        st.divider()
+
+        # ========== Latest results ==========
+        def _latest(lst):
+            return lst[0] if lst else None
+
+        latest_comm = _latest(comm_runs)
+        latest_infer = _latest(infer_runs)
+        latest_ops = _latest(ops_runs)
+
+        colA, colB, colC = st.columns(3)
+
+        with colA:
+            st.markdown("#### 🔗 通信（最新）")
+            if not latest_comm:
+                st.info("暂无通信结果")
+            else:
+                st.write(f"- testcase: `{latest_comm.get('testcase','')}`")
+                st.write(f"- time: {latest_comm.get('time','')}")
+                st.write(f"- status: {'✅' if latest_comm.get('success') else '❌'}")
+
+        with colB:
+            st.markdown("#### 🚀 推理（最新）")
+            if not latest_infer:
+                st.info("暂无推理结果")
+            else:
+                st.write(f"- testcase: `{latest_infer.get('testcase','')}`")
+                st.write(f"- time: {latest_infer.get('time','')}")
+                st.write(f"- status: {'✅' if latest_infer.get('success') else '❌'}")
+
+        with colC:
+            st.markdown("#### ⚡ 算子（最新）")
+            if not latest_ops:
+                st.info("暂无算子结果")
+            else:
+                st.write(f"- testcase: `{latest_ops.get('testcase','')}`")
+                st.write(f"- time: {latest_ops.get('time','')}")
+                st.write(f"- status: {'✅' if latest_ops.get('success') else '❌'}")
+
+        st.divider()
+
+        # ========== Recent runs table ==========
+        st.markdown("### 🕒 最近测试运行")
+        df = pd.DataFrame(
+            [
+                {
+                    "类型": (r.get("testcase", "").split(".")[0] or "UNKNOWN").upper(),
+                    "加速卡": ", ".join(r.get("accelerator_types", [])),
+                    "时间": r.get("time", ""),
+                    "状态": "✅" if r.get("success") else "❌",
+                    "run_id": r.get("run_id", "")[:32],
+                }
+                for r in runs[:15]
+            ]
+        )
+        st.dataframe(df, use_container_width=True, hide_index=True)
+
+        # ========== Dispatcher summary ==========
+        summaries = load_summary_file()
+
+        if not summaries:
+            st.info("No dispatcher_summary file found")
+            return
+
+        st.markdown("### 🧾 Dispatcher 汇总记录")
+
+        rows = []
+        for s in summaries:
+            rows.append(
+                {
+                    "时间": s.get("timestamp"),
+                    "总测试数": s.get("total_tests"),
+                    "成功": s.get("successful_tests"),
+                    "失败": s.get("failed_tests"),
+                    "成功率": (
+                        f"{s['successful_tests'] / s['total_tests'] * 100:.1f}%"
+                        if s.get("total_tests")
+                        else "-"
+                    ),
+                    "文件": s.get("file"),
+                }
+            )
+
+        df = pd.DataFrame(rows).sort_values("时间", ascending=False)
+
+        st.dataframe(
+            df,
+            use_container_width=True,
+            hide_index=True,
+        )
+
+        # ========== Quick navigation ==========
+        st.markdown("---")
+        st.markdown("### 🚀 快速导航")
+
+        col1, col2, col3 = st.columns(3)
+        if col1.button("🔗 通信测试分析", use_container_width=True):
+            st.switch_page("pages/communication.py")
+        if col2.button("⚡ 算子测试分析", use_container_width=True):
+            st.switch_page("pages/operator.py")
+        if col3.button("🤖 推理测试分析", use_container_width=True):
+            st.switch_page("pages/inference.py")
+
+    except Exception as e:
+        st.error(f"Dashboard 加载失败: {e}")
+
+
+if __name__ == "__main__":
+    main()

dashboard/app.py

@@ -0,0 +1,28 @@
+#!/usr/bin/env python3
+"""Common utilities for dashboard pages."""
+
+import streamlit as st
+import sys
+from pathlib import Path
+
+
+def init_page(page_title: str, page_icon: str):
+    """
+    Common page initialization:
+    - -Set Streamlit page configuration
+    - initialize DataLoader
+    - set project path
+    """
+    # Add project root to Python path
+    project_root = Path(__file__).parent.parent
+    if str(project_root) not in sys.path:
+        sys.path.append(str(project_root))
+
+    # Page configuration
+    st.set_page_config(page_title=page_title, page_icon=page_icon, layout="wide")
+
+    # Initialize DataLoader
+    if "data_loader" not in st.session_state:
+        from utils.data_loader import InfiniMetricsDataLoader
+
+        st.session_state.data_loader = InfiniMetricsDataLoader()

dashboard/common.py

@@ -0,0 +1,37 @@
+#!/usr/bin/env python3
+"""Header component for InfiniMetrics dashboard."""
+
+import streamlit as st
+
+
+def render_header():
+    """Render the dashboard header."""
+    st.markdown(
+        """
+        <style>
+        .main-header {
+            font-size: 2.5rem;
+            font-weight: bold;
+            color: #1E3A8A;
+            text-align: center;
+            margin-bottom: 1rem;
+        }
+        .sub-header {
+            font-size: 1.2rem;
+            color: #6B7280;
+            text-align: center;
+            margin-bottom: 2rem;
+        }
+        </style>
+    """,
+        unsafe_allow_html=True,
+    )
+
+    st.markdown(
+        '<div class="main-header">🏭 InfiniMetrics 测试结果展示平台</div>',
+        unsafe_allow_html=True,
+    )
+    st.markdown(
+        '<div class="sub-header">AI加速卡通信、算力、推理性能一站式分析与可视化</div>',
+        unsafe_allow_html=True,
+    )