{"release":{"tag":{"name":"24.12","path":"/deep-spark/deepsparkinference/tags/24.12","tree_path":"/deep-spark/deepsparkinference/tree/24.12","message":"24.12 Release","commit":{"id":"04705c916aed947a0feee1eaf430d0815607467b","short_id":"04705c9","title":"update cspdarknet53 name","title_markdown":"update cspdarknet53 name","description":"","description_markdown":"","message":"update cspdarknet53 name\n","message_markdown":"update cspdarknet53 name","detail_path":"/deep-spark/deepsparkinference/commit/04705c916aed947a0feee1eaf430d0815607467b","commits_path":"/deep-spark/deepsparkinference/commits/04705c916aed947a0feee1eaf430d0815607467b","tree_path":"/deep-spark/deepsparkinference/tree/04705c916aed947a0feee1eaf430d0815607467b","author":{"name":"honglyua","email":"hongliang.yuan@iluvatar.com","username":"honglyua","user_path":"/honglyua","enterprise_user_path":null,"image_path":"no_portrait.png#honglyua-honglyua","is_gitee_user":true,"is_enterprise_user":null,"widget_url":""},"committer":{"name":"honglyua","email":"hongliang.yuan@iluvatar.com","username":"honglyua","user_path":"/honglyua","enterprise_user_path":null,"image_path":"no_portrait.png#honglyua-honglyua","is_gitee_user":true,"is_enterprise_user":null,"widget_url":""},"authored_date":"2024-12-20T10:58:13+08:00","committed_date":"2024-12-20T10:58:13+08:00","signature":null,"build_state":null},"archive_path":"/deep-spark/deepsparkinference/repository/archive/24.12","signature":null},"operating":{"edit":false,"download":true,"destroy":false,"enterprise_forbid_zip":false},"release":{"title":"DeepSparkInference 24.12 Release","path":"/deep-spark/deepsparkinference/releases/tag/24.12","tag_path":"/deep-spark/deepsparkinference/tree/24.12","project_id":33771133,"created_at":"2024-12-23T10:46:53+08:00","is_prerelease":false,"description":"## 24.12 Release Notes\r\n\r\n### 模型与算法\r\n● 新增了24个推理小模型示例，其中支持IGIE推理引擎的15个，支持IxRT推理引擎的9个。\r\n● 新增了9个大语言模型的推理示例，其中支持vLLM的8个，支持IxFormer的1个。\r\n\r\n\r\n\u003Ctable\u003E\r\n  \u003Ctr colspan=4\u003E\r\n  \u003Cth colspan=3\u003EIGIE\u003C/th\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EConvNeXt-Base\u003C/td\u003E\r\n    \u003Ctd\u003EDenseNet201\u003C/td\u003E\r\n    \u003Ctd\u003EEfficientNet-B3\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EEfficientNetV2-S\u003C/td\u003E\r\n    \u003Ctd\u003EMNASNet0_5\u003C/td\u003E\r\n    \u003Ctd\u003EMViTv2_base\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003ERegnet_y_1_6gf\u003C/td\u003E\r\n    \u003Ctd\u003EResNetV1_D50\u003C/td\u003E\r\n    \u003Ctd\u003EResNeXt101_64x4d\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EShuffleNetV2_x1_5\u003C/td\u003E\r\n    \u003Ctd\u003EKie_layoutXLM\u003C/td\u003E\r\n    \u003Ctd\u003ERec_SVTR\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EYOLOv9\u003C/td\u003E\r\n    \u003Ctd\u003EYOLOv10\u003C/td\u003E\r\n    \u003Ctd\u003EPAA\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n    \u003Cth colspan=4\u003EIxRT\u003C/th\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003ECenterNet\u003C/td\u003E\r\n    \u003Ctd\u003EOpenPose\u003C/td\u003E\r\n    \u003Ctd\u003ERTMPose\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003ECSPDarkNet53\u003C/td\u003E\r\n    \u003Ctd\u003EDensNet161\u003C/td\u003E\r\n    \u003Ctd\u003EDensNet169\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EEfficientNetB2\u003C/td\u003E\r\n    \u003Ctd\u003EResNeXt50_32x4d\u003C/td\u003E\r\n    \u003Ctd\u003EConvNeXt-Small\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003C/tr\u003E\r\n    \u003Cth colspan=4\u003E大模型推理\u003C/th\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003ECLIP (IxFormer)\u003C/td\u003E\r\n    \u003Ctd\u003EChatGLM3-6B-32K (vLLM）\u003C/td\u003E\r\n    \u003Ctd\u003ELlama2-7B (vLLM)\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EMiniCPM-V-2 (vLLM）\u003C/td\u003E\r\n    \u003Ctd\u003EQwen-7B (vLLM)\u003C/td\u003E\r\n    \u003Ctd\u003EQwen1.5-32B-Chat (vLLM）\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n  \u003Ctr\u003E\r\n    \u003Ctd\u003EQwen2-72B-Instruct (vLLM）\u003C/td\u003E\r\n    \u003Ctd\u003EQwen2-7B-Instruct (vLLM）\u003C/td\u003E\r\n    \u003Ctd\u003EStableLM2-1.6B (vLLM）\u003C/td\u003E\r\n  \u003C/tr\u003E\r\n\u003C/table\u003E\r\n\r\n### 问题修复\r\n● 新增了IGIE推理模型自动化测试的运行脚本。\r\n● 修复了YOLOv8 IxRT模型运行推理脚本报错的问题。\r\n● 更新了YOLOv9和YOLOv10的IGIE模型的配置文件。\r\n● 完善了IxRT模型BERT，Mask RCNN，MobileNetV2和YOLOX的end2end推理时间打印。\r\n\r\n### 版本关联\r\nDeepSparkInference 24.12对应天数软件栈4.1.2版本。\r\n\r\n### 感谢以下社区贡献者\r\n\r\nYoungPeng，majorli6，xinchi.tian，xiaomei.wang，honglyua，qiang.zhang。","author":{"name":"honglyua","username":"honglyua","path":"/honglyua","avatar_url":"no_portrait.png#honglyua-honglyua"},"attach_files":[],"zip_download_url":"/deep-spark/deepsparkinference/releases/tag/24.12.zip","tar_download_url":"/deep-spark/deepsparkinference/releases/tag/24.12.tar.gz"}}}