Merge pull request #4359 from opendatalab/dev

feat: add Hygon entry to acceleration cards list
2026-03-27 11:08:32 +07:00 · 2026-01-13 19:51:36 +08:00
parent 30c5d10e05 03698c656e
commit db40932e6d
4 changed files with 181 additions and 9 deletions
--- a/docker/china/dcu.Dockerfile
+++ b/docker/china/dcu.Dockerfile
@@ -0,0 +1,34 @@
+# Base image containing the vLLM inference environment, requiring amd64(x86-64) CPU + Hygon DCU.
+FROM harbor.sourcefind.cn:5443/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk25.04.2-1226-das1.7-py3.10-20251226
+
+
+# Install libgl for opencv support & Noto fonts for Chinese characters
+RUN apt-get update && \
+    apt-get install -y \
+        fonts-noto-core \
+        fonts-noto-cjk \
+        fontconfig && \
+    fc-cache -fv && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Install mineru latest
+RUN python3 -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
+    python3 -m pip install mineru[api,gradio] \
+                            "matplotlib>=3.10,<4" \
+                            "ultralytics>=8.3.48,<9" \
+                            "doclayout_yolo==0.0.4" \
+                            "ftfy>=6.3.1,<7" \
+                            "shapely>=2.0.7,<3" \
+                            "pyclipper>=1.3.0,<2" \
+                            "omegaconf>=2.3.0,<3" \
+                            numpy==1.25.0 \
+                            opencv-python==4.11.0.86 \
+                            -i https://mirrors.aliyun.com/pypi/simple && \
+    python3 -m pip cache purge
+
+# Download models and update the configuration file
+RUN /bin/bash -c "mineru-models-download -s modelscope -m all"
+
+# Set the entry point to activate the virtual environment and run the command line tool
+ENTRYPOINT ["/bin/bash", "-c", "export MINERU_MODEL_SOURCE=local && exec \"$@\"", "--"]
--- a/docs/zh/usage/acceleration_cards/Hygon.md
+++ b/docs/zh/usage/acceleration_cards/Hygon.md
@@ -0,0 +1,115 @@
+## 1. 测试平台
+以下为本指南测试使用的平台信息，供参考：
+```
+os: Ubuntu 22.04.3 LTS  
+cpu: Hygon Hygon C86-4G(x86-64)
+dcu: BW200
+driver: 6.3.13-V1.12.0a
+docker: 20.10.24
+```
+
+## 2. 环境准备
+
+### 2.1 使用 Dockerfile 构建镜像
+
+```bash
+wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/dcu.Dockerfile
+docker build --network=host -t mineru:dcu-vllm-latest -f dcu.Dockerfile .
+```
+
+
+## 3. 启动 Docker 容器
+
+```bash
+docker run -u root --name mineru_docker \
+    --network=host \
+    --ipc=host \
+    --shm-size=16G \
+    --device=/dev/kfd \
+    --device=/dev/mkfd \
+    --device=/dev/dri \
+    -v /opt/hyhal:/opt/hyhal \
+    --group-add video \
+    --cap-add=SYS_PTRACE \
+    --security-opt seccomp=unconfined \
+    -e MINERU_MODEL_SOURCE=local \
+    -it mineru:dcu-vllm-latest \
+    /bin/bash
+```
+
+执行该命令后，您将进入到Docker容器的交互式终端，您可以直接在容器内运行MinerU相关命令来使用MinerU的功能。
+您也可以直接通过替换`/bin/bash`为服务启动命令来启动MinerU服务，详细说明请参考[通过命令启动服务](https://opendatalab.github.io/MinerU/zh/usage/quick_usage/#apiwebuihttp-clientserver)。
+
+
+## 4. 注意事项
+
+不同环境下，MinerU对Hygon加速卡的支持情况如下表所示：
+
+<table border="1">
+  <thead>
+    <tr>
+      <th rowspan="2" colspan="2">使用场景</th>
+      <th colspan="2">容器环境</th>
+    </tr>
+    <tr>
+      <th>vllm</th>
+    </tr>
+  </thead>
+  <tbody>
+    <tr>
+      <td rowspan="3">命令行工具(mineru)</td>
+      <td>pipeline</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td>&lt;vlm/hybrid&gt;-auto-engine</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td>&lt;vlm/hybrid&gt;-http-client</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td rowspan="3">fastapi服务(mineru-api)</td>
+      <td>pipeline</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td>&lt;vlm/hybrid&gt;-auto-engine</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td>&lt;vlm/hybrid&gt;-http-client</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td rowspan="3">gradio界面(mineru-gradio)</td>
+      <td>pipeline</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td>&lt;vlm/hybrid&gt;-auto-engine</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td>&lt;vlm/hybrid&gt;-http-client</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td colspan="2">openai-server服务（mineru-openai-server）</td>
+      <td>🟢</td>
+    </tr>
+    <tr>
+      <td colspan="2">数据并行 (--data-parallel-size)</td>
+      <td>🟢</td>
+    </tr>
+  </tbody>
+</table>
+
+注：  
+🟢: 支持，运行较稳定，精度与Nvidia GPU基本一致  
+🟡: 支持但较不稳定，在某些场景下可能出现异常，或精度存在一定差异  
+🔴: 不支持，无法运行，或精度存在较大差异
+
+>[!TIP]
+>DCU加速卡指定可用加速卡的方式与AMD GPU类似，请参考[GPU isolation techniques](https://rocm.docs.amd.com/en/docs-6.2.4/conceptual/gpu-isolation.html)
--- a/docs/zh/usage/index.md
+++ b/docs/zh/usage/index.md
@@ -12,6 +12,7 @@
    * [昇腾 Ascend](acceleration_cards/Ascend.md) 🚀
    * [平头哥 T-Head](acceleration_cards/THead.md) 🚀
    * [沐曦 METAX](acceleration_cards/METAX.md) 🚀
+    * [海光 Hygon](acceleration_cards/Hygon.md) 🚀
    * [AMD](acceleration_cards/AMD.md)  [#3662](https://github.com/opendatalab/MinerU/discussions/3662) ❤️
    * [太初元碁 Tecorigin](acceleration_cards/Tecorigin.md) [#3767](https://github.com/opendatalab/MinerU/pull/3767) ❤️
    * [寒武纪 Cambricon](acceleration_cards/Cambricon.md) [#4004](https://github.com/opendatalab/MinerU/discussions/4004) ❤️
--- a/mineru/utils/table_merge.py
+++ b/mineru/utils/table_merge.py
@@ -9,13 +9,18 @@ from mineru.utils.char_utils import full_to_half
 from mineru.utils.enum_class import BlockType, SplitFlag


-CONTINUATION_MARKERS = [
+CONTINUATION_END_MARKERS = [
    "(续)",
    "(续表)",
    "(续上表)",
    "(continued)",
    "(cont.)",
    "(cont’d)",
+    "(…continued)",
+]
+
+CONTINUATION_INLINE_MARKERS = [
+    "(continued)",
 ]


@@ -163,20 +168,32 @@ def detect_table_headers(soup1, soup2, max_header_rows=5):
 def can_merge_tables(current_table_block, previous_table_block):
    """判断两个表格是否可以合并"""
    # 检查表格是否有caption和footnote
+    # 计算previous_table_block中的footnote数量
+    footnote_count = sum(1 for block in previous_table_block["blocks"] if block["type"] == BlockType.TABLE_FOOTNOTE)
    # 如果有TABLE_CAPTION类型的块,检查是否至少有一个以"(续)"结尾
    caption_blocks = [block for block in current_table_block["blocks"] if block["type"] == BlockType.TABLE_CAPTION]
    if caption_blocks:
-        # 如果所有caption都不以"(续)"、"(续表)"、"(continued)"或"(cont.)"结尾,则不合并
+        # 检查是否至少有一个caption包含续表标识
+        has_continuation_marker = False
+        for block in caption_blocks:
+            caption_text = full_to_half(merge_para_with_text(block).strip()).lower()
+            if (
+                    any(caption_text.endswith(marker.lower()) for marker in CONTINUATION_END_MARKERS)
+                    or any(marker.lower() in caption_text for marker in CONTINUATION_INLINE_MARKERS)
+            ):
+                has_continuation_marker = True
+                break

-        if not any(
-                any(full_to_half(merge_para_with_text(block).strip()).lower().endswith(marker.lower())
-                    for marker in CONTINUATION_MARKERS)
-                for block in caption_blocks
-        ):
+        # 如果所有caption都不包含续表标识，则不允许合并
+        if not has_continuation_marker:
            return False, None, None, None, None

-    if any(block["type"] == BlockType.TABLE_FOOTNOTE for block in previous_table_block["blocks"]):
-        return False, None, None, None, None
+        # 如果current_table_block的caption存在续标识,放宽footnote的限制允许previous_table_block有最多一条footnote
+        if footnote_count > 1:
+            return False, None, None, None, None
+    else:
+        if footnote_count > 0:
+            return False, None, None, None, None

    # 获取两个表格的HTML内容
    current_html = ""
@@ -363,6 +380,11 @@ def perform_table_merge(soup1, soup2, previous_table_block, wait_merge_table_foo
                row.extract()
                tbody1.append(row)

+    # 清空previous_table_block的footnote
+    previous_table_block["blocks"] = [
+        block for block in previous_table_block["blocks"]
+        if block["type"] != BlockType.TABLE_FOOTNOTE
+    ]
    # 添加待合并表格的footnote到前一个表格中
    for table_footnote in wait_merge_table_footnotes:
        temp_table_footnote = table_footnote.copy()