v1.3.0
版本发布时间: 2023-02-20 19:57:16
modelscope/modelscope最新发布版本:v1.13.2(2024-03-22 17:57:37)
中文版本
该版本共新增上架51个模型,其中11个模型支持finetune能力。
模型功能特性说明
-
提供finetune的示例脚本,允许用户通过运行脚本命令行传参方式进行模型训练,详细可以参考github脚本
-
NLP领域新增了backbone + head的开发支持,允许用户任意组合已有的backbone(Encoder) 和任务head,方便在特定任务上切换不同模型进行建模,详细参考文档
-
贡献者文档完善模型贡献部分,详细参考接入流程概览
-
数据集接口支持本地文件直接加载 MsDataset.load('/to/path/abc.csv')
-
模型导出支持nlp_structbert_zero-shot、 nlp_csanmt_translation系列模型
更多SDK功能和变更可查看:https://github.com/modelscope/modelscope/releases/tag/v1.3.0
新模型列表及快捷访问
最佳实践教程
最后,我们还推出许多任务级别和模型级别的最佳实践教程文档,旨在帮助开发者更好地理解和应用模型。
欢迎关注我们的开源社区:https://github.com/modelscope/modelscope
English Version
Highlight
- Add vqa-degradation
- Add content check pipeline
- Add pipelines for en2zh-imt and zh2en-imt
- Add single and multiple human parsing models
- Add AdaInt model
- Add open vocabulary detection
- Support finetune for sentence-embedding
- Add bad image detection model and pipeline
- Support translation model exporting
- Add asr dataset for finetune
- Add ocr detection model and pipeline
- Add face quality assessment model
- Add video deinterlace model
- Add language model for audio task
- Add deeplpf for image color enhance and image debanding model
- Add ecbsr model for mobile image super-resolution
- Add msrresnetlite model for video super-resolution
- Support finetune and evaluation for image-fewshot-detection-defrcn
- Add yolopv2 model cv_yolopv2_image_driving_perception
- Add face liveness xc model
- Add paint-by-example model
- Add universal_matting pipeline
- Add multi-modal_gridvlp_classification_chinese-base-ecom-cate
- Add DINO detection with easycv
- Add speech speaker verification pipeline
- Add nerf-recon model
- Support finetune for real-time object detection with easycv
- Add single-camera depth estimation bts model
- Add MGIMN model
- Add fuse-in-decoder dialogue task
- Add vision_efficient_tuning models
- Add traffic-sign detection
- Add object_detection3d_depe model
- Add stable diffusion model for image inpainting
- Add head&phone detection models
- Add face_reconstruction model
- Add structured model probing pipeline for image classification
- Add video panorama segmentation with VideoKNet-SwinB
- Add image quality assessment mos(mean option score) model
- Add ddpm-segmentation pipeline
- Add plug mental model
- Add video-colorization pipeline
- Add image demoireing
- Add face recognition ir model
- Support batch inference for nlp_csanmt_translation_en2zh
- Add image_deblurring_dataset for REDS dataset
- Add new motion-generation model
- Add face recognition and face mask model
Breaking changes
- Adjust video_multi_target_tracking output
- Adjust video_human_matting output of video to support demo service
Feature
- Add default preprocessor for taskmodels
- Run ci cases base on code diff to reduct ci test time
- Support demo code to return path of result video for video human matting
- Add en2ru and ru2en pipeline ut v4
- Kws pipeline returns Chinese charactor by configuration
- Ast-scanning skip function level imports index
- Compatible with diffusers0.12.1
- Video depth estimation support cpu mode
- asr pipeline add output_dir parame
- Add RTS face recognition ood model
- Add image-defrcn-fewshot-detection
Improvements
- Remove requirements of mpi4py
- Remove pytorch-lightning version constrain
- Refine cv_image_defrcn trainer to avoid failed
- Support trainer prediction
- Allow pass prompt in kwargs & reduce GPU usage for image_inpainting
- Improve video frame interpolation pipeline
- Use package LoadImage for image io in image_quality_assessment_mos
- Remove text2sql_lgesql from nlp requirements
- Remove tensorboard hook as default
- Add model_revision parameter to ImageDetectionDamoyoloTrainer
- Update mgeo finetune test case for rerank
- Add args for asr_infer_pipeline, punc_pipeline, sv_pipeline & modify funasr version
- Add model type check and give easy-to-understand error prompts
- Replace
import torchaudio
to avoid unnecessary requirements in framework - Split training and evaluating code for nearfield kws trainer
- Update test image for image deblur
- Add output_dir for asr inferencewhen called
- Support cpu mode for video depth estimation
- Add zhconv to nlp requirements
- Modify the resumable cache path for oss utils
- Support the form of '/to/path/abc.csv' in MsDataset.load() function
- Add UT cases
- Limit pyarrow version
BugFix
- Fix bug in speaker verification infer
- Fix bug when add use_fast for text_ranking
- Fix two ckpt hooks save in the same dir
- Fix the bug that image_color_enhance_pipeline cannot run in CPU environment
- Fix bugs in audio fs, asr & sv demo services
- Fix gpt3 unexpected spaces
- Fix loading checkpoint errors for palm
- Fix data parallel bug for mgeo evaluation
- Fix the checkpoint is incompletely saved with tensor model parallel
- Fix _eval_iters_per_epoch None bug
- Fix damoyolo evaluater load checkpoint not matched
- Fix video matting demo format (mp4v to h264)
- Fix delete model revision
- Fix datasets version incompatible issue
- Fix the compatibility issue of datasets
- Fix postprocessor bugs with batch inference
- Fix asr backward compatibility during inference with tensorflow
- Fix for hand detect finetune
- Fix typos