Merge pull request #3196 from myhloli/dev

fix: update Mineru2ImageProcessor constructor to accept additional arguments to adaptation sglang 0.4.9.post5

README.md

@@ -43,6 +43,10 @@
 </div>
 
 # Changelog
+- 2025/07/28 version 2.1.8 Released
+  - `sglang` 0.4.9.post5 version adaptation
+- 2025/07/27 version 2.1.7 Released
+  - `transformers` 4.54.0 version adaptation
 - 2025/07/26 2.1.6 Released
   - Fixed table parsing issues in handwritten documents when using `vlm` backend
   - Fixed visualization box position drift issue when document is rotated #3175

README_zh-CN.md

@@ -43,16 +43,20 @@
 </div>
 
 # 更新记录
-- 2025/07/26 2.1.6发布
+- 2025/07/28 2.1.8 发布
+  - `sglang` 0.4.9.post5 版本适配
+- 2025/07/27 2.1.7 发布
+  - `transformers` 4.54.0 版本适配
+- 2025/07/26 2.1.6 发布
   - 修复`vlm`后端解析部分手写文档时的表格异常问题
   - 修复文档旋转时可视化框位置漂移问题 #3175
-- 2025/07/24 2.1.5发布
+- 2025/07/24 2.1.5 发布
   - `sglang` 0.4.9 版本适配，同步升级dockerfile基础镜像为sglang 0.4.9.post3
-- 2025/07/23 2.1.4发布
+- 2025/07/23 2.1.4 发布
   - bug修复
     - 修复`pipeline`后端中`MFR`步骤在某些情况下显存消耗过大的问题 #2771
     - 修复某些情况下`image`/`table`与`caption`/`footnote`匹配不准确的问题 #3129
-- 2025/07/16 2.1.1发布
+- 2025/07/16 2.1.1 发布
   - bug修复 
     - 修复`pipeline`在某些情况可能发生的文本块内容丢失问题 #3005
     - 修复`sglang-client`需要安装`torch`等不必要的包的问题 #2968
@@ -60,7 +64,7 @@
   - 易用性更新
     - 更新`compose.yaml`，便于用户直接启动`sglang-server`、`mineru-api`、`mineru-gradio`服务
     - 启用全新的[在线文档站点](https://opendatalab.github.io/MinerU/zh/)，简化readme，提供更好的文档体验
-- 2025/07/05 2.1.0发布
+- 2025/07/05 2.1.0 发布
   - 这是 MinerU 2 的第一个大版本更新，包含了大量新功能和改进，包含众多性能优化、体验优化和bug修复，具体更新内容如下： 
   - 性能优化： 
     - 大幅提升某些特定分辨率（长边2000像素左右）文档的预处理速度

docker/china/Dockerfile

@@ -1,7 +1,7 @@
 # Use the official sglang image
-FROM lmsysorg/sglang:v0.4.9.post3-cu126
+FROM lmsysorg/sglang:v0.4.9.post5-cu126
 # For blackwell GPU, use the following line instead:
-# FROM lmsysorg/sglang:v0.4.9.post3-cu128-b200
+# FROM lmsysorg/sglang:v0.4.9.post5-cu128-b200
 
 # Install libgl for opencv support & Noto fonts for Chinese characters
 RUN apt-get update && \

docker/global/Dockerfile

@@ -1,7 +1,7 @@
 # Use the official sglang image
-FROM lmsysorg/sglang:v0.4.9.post3-cu126
+FROM lmsysorg/sglang:v0.4.9.post5-cu126
 # For blackwell GPU, use the following line instead:
-# FROM lmsysorg/sglang:v0.4.9.post3-cu128-b200
+# FROM lmsysorg/sglang:v0.4.9.post5-cu128-b200
 
 # Install libgl for opencv support & Noto fonts for Chinese characters
 RUN apt-get update && \

docs/en/quick_start/docker_deployment.md

@@ -10,8 +10,8 @@ docker build -t mineru-sglang:latest -f Dockerfile .
 ```
 
 > [!TIP]
-> The [Dockerfile](https://github.com/opendatalab/MinerU/blob/master/docker/global/Dockerfile) uses `lmsysorg/sglang:v0.4.9.post3-cu126` as the base image by default, supporting Turing/Ampere/Ada Lovelace/Hopper platforms.
-> If you are using the newer `Blackwell` platform, please modify the base image to `lmsysorg/sglang:v0.4.9.post3-cu128-b200` before executing the build operation.
+> The [Dockerfile](https://github.com/opendatalab/MinerU/blob/master/docker/global/Dockerfile) uses `lmsysorg/sglang:v0.4.9.post5-cu126` as the base image by default, supporting Turing/Ampere/Ada Lovelace/Hopper platforms.
+> If you are using the newer `Blackwell` platform, please modify the base image to `lmsysorg/sglang:v0.4.9.post5-cu128-b200` before executing the build operation.
 
 ## Docker Description
 

docs/zh/quick_start/docker_deployment.md

@@ -10,8 +10,8 @@ docker build -t mineru-sglang:latest -f Dockerfile .
 ```
 
 > [!TIP]
-> [Dockerfile](https://github.com/opendatalab/MinerU/blob/master/docker/china/Dockerfile)默认使用`lmsysorg/sglang:v0.4.9.post3-cu126`作为基础镜像，支持Turing/Ampere/Ada Lovelace/Hopper平台，
-> 如您使用较新的`Blackwell`平台，请将基础镜像修改为`lmsysorg/sglang:v0.4.9.post3-cu128-b200` 再执行build操作。
+> [Dockerfile](https://github.com/opendatalab/MinerU/blob/master/docker/china/Dockerfile)默认使用`lmsysorg/sglang:v0.4.9.post5-cu126`作为基础镜像，支持Turing/Ampere/Ada Lovelace/Hopper平台，
+> 如您使用较新的`Blackwell`平台，请将基础镜像修改为`lmsysorg/sglang:v0.4.9.post5-cu128-b200` 再执行build操作。
 
 ## Docker说明
 

mineru/model/mfr/unimernet/unimernet_hf/unimer_mbart/modeling_unimer_mbart.py

@@ -1416,7 +1416,11 @@ class UnimerMBartDecoder(UnimerMBartPreTrainedModel):
             raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")
 
         # past_key_values_length
-        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
+        # past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
+        past_key_values_length = 0
+        if past_key_values is not None:
+            if isinstance(past_key_values, (list, tuple)) and past_key_values:
+                past_key_values_length = past_key_values[0][0].shape[2]
 
         if inputs_embeds is None:
             inputs_embeds = self.embed_tokens(input_ids)
@@ -1501,7 +1505,12 @@ class UnimerMBartDecoder(UnimerMBartPreTrainedModel):
                 if dropout_probability < self.layerdrop:
                     continue
 
-            past_key_value = past_key_values[idx] if past_key_values is not None else None
+            # past_key_value = past_key_values[idx] if past_key_values is not None else None
+            past_key_value = past_key_values[idx] if (
+                    past_key_values is not None and
+                    isinstance(past_key_values, (list, tuple)) and
+                    idx < len(past_key_values)
+            ) else None
 
             if self.gradient_checkpointing and self.training:
                 layer_outputs = self._gradient_checkpointing_func(

mineru/model/vlm_sglang_model/image_processor.py

@@ -56,8 +56,8 @@ def process_anyres_image(image, processor, grid_pinpoints):
 
 
 class Mineru2ImageProcessor(BaseProcessor):
-    def __init__(self, hf_config, server_args, _processor):
-        super().__init__(hf_config, server_args, _processor)
+    def __init__(self, hf_config, server_args, _processor, *args, **kwargs):
+        super().__init__(hf_config, server_args, _processor, *args, **kwargs)
 
     @staticmethod
     def _process_single_image_task(

mineru/version.py

@@ -1 +1 @@
-__version__ = "2.1.5"
+__version__ = "2.1.7"