v0.8.2 release

What's Changed

• add user group by @Yunnglin in #251
• fix perf seed by @Yunnglin in #254
• add spawn env by @Yunnglin in #256
• Fix: sglang API response does not contain 'object' field. by @tghfly in #260
• fix parse response by @Yunnglin in #262
• fix predict by @Yunnglin in #264
• compat ragas 0.2.9 and remove chinese prompt cache by @Yunnglin in #265

New Contributors

• @tghfly made their first contribution in #260

Full Changelog: v0.8.1...v0.8.2

v0.8.1 release

What's Changed

• Unify opencompass and vlmeval output dirs by @Yunnglin in #242
• Perf add more metrics by @Yunnglin in #245
• Perf add trust remote parameter by @Yunnglin in #246
• Compat ms-swift<3.0 by @Yunnglin in #249
• Fix humaneval for native eval by @Yunnglin in #248

中文版本

• 统一 opencompass 和 vlmeval 输出目录，作者：@Yunnglin，相关链接：#242
• 模型压测：增加更多指标，作者：@Yunnglin，相关链接：#245
• 模型压测：添加trust remote参数，作者：@Yunnglin，相关链接：#246
• 兼容 ms-swift<3.0，作者：@Yunnglin，相关链接：#249
• 修复本地评估的 humaneval 问题，作者：@Yunnglin，相关链接：#248

Full Changelog: v0.8.0...v0.8.1

v0.8.0 release

Release Notes

1. Optimize Native eval and remove template_type #231
2. The evalscope perf command supports the --outputs-dir configuration. #232
3. Support ragas 0.2.7 #234

Bug Fixes

1. Fix longwriter docs #239
2. Fix lint for longwriter #240
3. Fix lint #237
4. Unify perf output #238

Documentation Updates

1. Fix longwriter docs #239
2. Optimize Native eval and remove template_type #231

中文说明

特性

1. 取消Native模式评测中template_type参数 #231
2. perf模块支持--output-dir #232
3. 支持适配最新的ragas 0.2.7版本 #234

缺陷修复

1. 修复longwriter代码示例，优化流程 #239
2. 修复lint，以及longwriter的lint #240 #237

文档更新

1. 更新longwriter文档 #239
2. 更新Native评测模式的相关文档 #231

v0.7.2 release

Release Note

1. Remove pyarrow version requirement #225
2. Optimize warning info #223

中文说明

1. 移除 pyarrow 版本要求 #225
2. 优化 warning 信息 #223

v0.7.1 release

Release Notes

1. Add PMMEval benchmark #222

中文说明

特性

1. 增加PMMEval评测集 #222

v0.7.0 release

Release Notes

1. Refactor the perf module, more robust and easier to use. #178
2. Add speed benchmarking in the perf module. #178
3. Add multi-modal benchmark flickr8k in the perf module for speed benchmark. #211

Bug Fixes

1. Add timeout for download punkt.zip #206
2. Fix parallel for speed benchmarking in the perf module. #215

Documentation Updates

1. Update VLM-Eval doc #209
2. Update perf module doc #178 #211

中文说明

特性

1. 重构perf模块，更鲁棒、更易用。 #178
2. 在perf模块中添加速度基准测试。 #178
3. 在perf模块中添加多模态基准 flickr8k 以进行速度基准测试。 #211

缺陷修复

1. 修复下载punkt.zip的超时问题。 #206
2. 修复perf模块中的速度基准测试并行问题。 #215

文档更新

1. 更新VLM-Eval文档。 #209
2. 更新perf模块文档。 #178 #211

v0.6.1 release

Release Notes

1. Add CMMLU benchmark #198
2. Add publish workflow #186
3. Adapt RAGAS v0.2.5 and update readme #205
4. Adapt MTEB v1.19 #196

Bug Fixes

1. Set datasets version: dataset>=3.0.0, <=3.0.1 #184
2. Set pyarrow version to <=17.0.0 to avoid installation issue on OSX. #187
3. Add timeout for download punkt.zip #206

Documentation Updates

1. Update OpenCompass list all datasets docs #199
2. Update RAGAS v0.2.5 docs #205

中文说明

特性

1. 支持CMMLU benchmark #198
2. 支持publish 流程 #186
3. 适配RAGAS v0.2.5并更新文档 #205
4. 适配 MTEB v1.19 #196

缺陷修复

1. 设置datasets 版本，修复兼容性问题: dataset>=3.0.0, <=3.0.1 #184
2. 设置 pyarrow版本：<=17.0.0 修复在OSX操作系统下的安装问题 #187
3. 增加下载punkt.zip时的超时时间 #206

文档更新

1. 更新OpenCompass作为backend时所支持的数据集列表文档 #199
2. 更新RAGAS v0.2.5 文档 #205

Release v0.6.0

Release Notes

1. Support multi-modal RAG evaluation #149
   • Add CLIP_Benchmark
   • Add end-to-end multi-modal RAG evaluation in Ragas
2. To be compatible with Ragas v0.2.3 #165 #171
3. Support truncating input for CLIP models #163 #164
4. Support saving knowledge graphs when generating datasets in Ragas #175

Bug Fixes

1. Fix issue of abnormal metrics during CMTEB evaluation #157
2. Fix issue of GenerationConfig being None #173
3. Update datasets version constraints #184
4. Add publish workflow #186

Documentation Updates

1. Update VLMEvalKit documentation #166
2. Update multi-modal RAG blog #172

中文说明

特性

1. 添加多模态RAG评测支持 #149
   • 支持CLIP_Benchmark
   • 支持Ragas端到端多模态RAG评测
2. 兼容Ragas v0.2.3 #165 #171
3. 支持CLIP模型截断输入 #163 #164
4. 支持Ragas生成数据集时保存知识图谱 #175

缺陷修复

1. 修复CMTEB评估时指标异常的问题 #157
2. 修复GenerationConfig为None的异常 #173
3. 更新datasets版本限制 #184
4. 增加publish workflow #186

文档更新

1. 更新VLMEvalKit文档 #166
2. 更新多模态RAG博客 #172

Release v0.5.5

Release Notes

1. Added Dataset Support:

   • Enhanced multimodal evaluation capabilities, now supporting MMBench-Video, Video-MME, and MVBench video evaluations #146
   • Added cmb dataset #117

2. Support for LongBench-write quality evaluation of long text generation #136

3. Automatic downloading of punkt_tab.zip from nltk #140

4. Support for RAG evaluation #127:

   • Support for embeddings/reranker evaluation: Integration of MTEB (Massive Text Embedding Benchmark) and CMTEB (Chinese Massive Text Embedding Benchmark), supporting tasks such as retrieval and reranking
   • Support for end-to-end RAG evaluation: Integration of the ragas framework, supporting automatic generation of evaluation datasets and evaluation based on judge models

5. Documentation Updates:

   • Added "Blog" section #126, #135
   • Added support for dataset page #121
   • Updated function usage instructions #125, #134, #138, #137, #127

6. Updated dependencies: nltk>=3.9 and rouge-score>=0.1.0 #145, #143

中文说明

1. 新增数据集支持：

   • 完善多模态评测功能，支持MMBench-Video，Video-MME，MVBench视频评测 #146
   • 新增cmb数据集 #117

2. 支持LongBench-write 长文本生成的质量评测 #136

3. 支持从nltk自动下载 punkt_tab.zip #140

4. 支持RAG评测：#127

   • 支持embeddings/reranker 评测：集成MTEB（Massive Text Embedding Benchmark）和 CMTEB（Chinese Massive Text Embedding Benchmark），支持检索、重排等任务评估
   • 支持RAG端到端评测：集成ragas框架，支持自动生成评测数据集和基于裁判员模型的评测

5. 文档更新

   • 增加 “博客” 板块 #126, #135
   • 增加支持的数据集页面 #121
   • 更新功能使用说明 #125, #134, #138, #137, #127

6. 更新依赖nltk>=3.9和rouge-score>=0.1.0 #145, #143

Release v0.5.2

Highlight features

• Support Multi-modal models evaluation (VLM Eval)
• Transform the synchronous API to asynchronous for OpenAI's API format, speed up the evaluation process up to 10x .
• Support installation with format: pip install evalscope[opencompass] or pip install evalscope[vlmeval]

Breaking Changes

None

What's Changed

1. Support Multi-modal models evaluation (VLM Eval)
2. Transform the synchronous API to asynchronous for OpenAI's API format, speed up the evaluation process up to 10x .
3. Support installation with format: pip install evalscope[opencompass] or pip install evalscope[vlmeval]
4. Update README
5. Add UT cases for VLM eval
6. Update examples for OpenCompass and VLMEval eval backends
7. Update version restrictions for ms-opencompass and ms-vlmeval dependencies.