open-mmlab/mmpretrain issues and pull requests

#1968 - Dimension Mismatch in Binary Classification with CrossEntropyLoss and use_sigmoid=True

Issue - State: open - Opened by egoravv 28 days ago - 1 comment

#1967 - [Bug] RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead

Issue - State: open - Opened by kirottu1 about 1 month ago

#1966 - [Bug] Incompatibility between mmcv, mmdet, mmtrack.

Issue - State: open - Opened by deepalchemist about 1 month ago

#1965 - Inference time gradual slow down

Issue - State: open - Opened by vincentw1997 about 1 month ago

#1963 - [Feature] mmpretrain to timm model

Issue - State: open - Opened by shinpaul14 about 2 months ago

#1962 - [Bug] fail to build with python 3.13

Issue - State: open - Opened by hubutui about 2 months ago

#1959 - Ignore

Issue - State: closed - Opened by Tegala 3 months ago

#1957 - [Bug] ConvNeXtV2 预训练权重加载键值对不匹配

Issue - State: closed - Opened by caixiaoshun 3 months ago

#1956 - fix register torchvision.transforms error

Pull Request - State: open - Opened by FelixFu520 4 months ago - 1 comment

#1954 - [Bug] Your pre -training model accuracy is different from logs

Issue - State: open - Opened by freedomspiritJohn 4 months ago

#1953 - [Bug] cspnet的实际用的是darknet53，但是配置命名却是50

Issue - State: open - Opened by rexlee8776 4 months ago

#1952 - revert changes

Pull Request - State: closed - Opened by wemoveon2 4 months ago - 1 comment

#1951 - [SVM-796] upgrade torch

Pull Request - State: closed - Opened by wemoveon2 5 months ago - 1 comment

#1950 - If I have trained the model and want to use pytorch to implement data pre-processing, how should I disassemble the code?

Issue - State: open - Opened by goldwater668 5 months ago

#1947 - fix: add missing device argument in image_demo

Pull Request - State: open - Opened by asas1asas200 5 months ago

#1946 - [Feature] 没有原版Vit的预训练权重和训练参数！ Lack of Vanilla ViT Support!

Issue - State: open - Opened by 2catycm 5 months ago

#1945 - fix typo mush to much

Pull Request - State: open - Opened by asas1asas200 5 months ago - 2 comments

#1944 - how to convert rgb to grayscale in config setting?

Issue - State: open - Opened by shuyuan-wang 6 months ago

#1943 - ConvNeXt V2 Whether to use fcma pre-training？

Issue - State: open - Opened by goldwater668 6 months ago

#1942 - [Bug] Issue with Multi-GPU Training on Windows 10 in MMPretrain

Issue - State: closed - Opened by Riki123666 6 months ago

#1941 - Input image size

Issue - State: open - Opened by marthajoddrell 6 months ago - 3 comments

#1940 - [Feature] If I use convnet to train a model, and then use the model for feature extraction and image retrieval, how should I apply this framework?

Issue - State: open - Opened by watertianyi 6 months ago - 1 comment

#1939 - [Bug] Getting exit code 137 in validation step while training

Issue - State: open - Opened by idonahum1 6 months ago

#1938 - [Bug] my bad

Issue - State: open - Opened by victormay 6 months ago - 1 comment

#1937 - [Feature] 请问如何对这些模型进行微调，说明文档里并未找到

Issue - State: open - Opened by abcc235 6 months ago - 1 comment

#1934 - [Bug] konflict between ReduceOnPlateauLR and val_evaluator

Issue - State: open - Opened by hamlet2019 6 months ago

#1931 - assert isinstance(inputs, list)

Issue - State: open - Opened by JinMana 6 months ago

#1930 - [Fix] Typo in param size of ResNet34

Pull Request - State: open - Opened by wangruohui 6 months ago

#1927 - [Bug] cpu not fully used, data_time load slow

Issue - State: open - Opened by YAwei666 7 months ago - 2 comments

#1926 - [Bug] cannot find a doc page

Issue - State: open - Opened by sirq2tap 7 months ago

#1925 - How can I train the model using the APPLE M3 "mps" device?

Issue - State: open - Opened by zh-zhang1984 7 months ago

#1924 - SimMIM support MFF?[Feature]

Issue - State: open - Opened by sjtu-cz 7 months ago

#1923 - [Feature] support mlflow do advance experiment management

Issue - State: open - Opened by zhanglazhuan 7 months ago - 1 comment

#1922 - [Feature] About Release

Issue - State: open - Opened by HarborYuan 7 months ago - 1 comment

#1920 - [Docs] A100算力加持！书生大模型实战营第3期全面升级，趣味闯关模式等你开启

Issue - State: open - Opened by boshallen 7 months ago

#1919 - [Bug] Error when trying to replicate the tutorial for pre-training with custom dataset.

Issue - State: open - Opened by siddhi-wiai 8 months ago - 1 comment

#1918 - Master

Pull Request - State: closed - Opened by crazyboy9103 8 months ago - 1 comment

#1917 - [Bug] Layer normalisation not implemented for transformers

Issue - State: open - Opened by teethoe 8 months ago - 1 comment

#1915 - [Feature] 请求添加MAE可视化重建的代码。

Issue - State: open - Opened by EndeavorWang 8 months ago - 1 comment

#1913 - Correct EfficientNetV2 URL in metafile.yml

Pull Request - State: closed - Opened by bryanbocao 8 months ago

#1911 - How to inference a multi-label image? is it possible to provide an example?

Issue - State: open - Opened by liulan93 8 months ago

#1909 - [Feature] Support StarNet

Pull Request - State: open - Opened by bear-coder-9527 8 months ago

#1908 - [Feature] Support StarNet

Pull Request - State: closed - Opened by bear-coder-9527 8 months ago - 1 comment

#1907 - New Backbone StarNet for mmlab

Pull Request - State: open - Opened by Lazy-coder-9527 8 months ago - 1 comment

#1906 - New Backbone StarNet for mmlab

Pull Request - State: closed - Opened by Lazy-coder-9527 8 months ago - 1 comment

#1905 - New Backbone StarNet

Pull Request - State: open - Opened by Lazy-coder-9527 8 months ago - 1 comment

#1904 - [Bug] BEiT performance

Issue - State: open - Opened by djaniak 8 months ago

#1903 - Can I use the mmpretrain/tools/visualization/vis_cam.py in the mmdetection package for CAM ?

Issue - State: open - Opened by xyb1314 9 months ago

#1901 - [Feature] why remove the gpu-id?

Issue - State: open - Opened by Maxwell2016LeChouchou 9 months ago

#1900 - [Bug] Current installing process filed with JIT missing error

Issue - State: open - Opened by HelloWorldLTY 9 months ago

#1899 - EVA 02 vit_h pre training weight loading problem

Issue - State: open - Opened by lyc200522 9 months ago

#1898 - RuntimeError: The configuration file type in the inheritance chain must match the current configuration file type, either "lazy_import" or non-"lazy_import".

Issue - State: open - Opened by santoshamd 9 months ago - 2 comments

#1897 - [Feature] How can I use torch.compile with mode? For example, torch.compile(model, mode="reduce-overhead")

Issue - State: open - Opened by lyp-liuyipeng 9 months ago

#1896 - Accuracy is low on the testing set (split from the same distributation of the training set) but is high when conducting K-Fold validation on the training set.

Issue - State: open - Opened by freshn 9 months ago - 1 comment

#1894 - Fix deploy error

Pull Request - State: open - Opened by Y-T-G 10 months ago - 1 comment

#1893 - [Feature] 如何加载预训练模型进行finetune？

Issue - State: open - Opened by chen1234520 10 months ago - 4 comments

#1892 - [Bug] Cannot replicate reported CIFAR-100 results for multiple models

Issue - State: open - Opened by saifkhichi96 10 months ago - 1 comment

#1891 - fix minor bug with `get_model` function

Pull Request - State: open - Opened by haruishi43 10 months ago - 1 comment

#1890 - DINOv2 training

Issue - State: closed - Opened by LichunZhang 10 months ago - 1 comment

#1889 - [Bug] Spark预训练权重加载错误！

Issue - State: open - Opened by WWork1 10 months ago - 4 comments

#1888 - [Bug] KeyError: 'gt_label'

Issue - State: open - Opened by yanshuangying888 10 months ago - 5 comments

#1885 - [Bug] mim resources not found: [...]/.venv/lib/python3.10/site-packages/mmpretrain/.mim, you can try to install the latest mmpretrain.

Issue - State: open - Opened by maygeisler 11 months ago

#1883 - [Feature] please add i-jepa

Issue - State: open - Opened by dirtycomputer 11 months ago

#1881 - [Feature] Can I view val loss while the code is running

Issue - State: open - Opened by slantingsun 12 months ago

#1880 - [Bug] Cannot register custom collate_fn function

Issue - State: open - Opened by ShangWeiKuo 12 months ago

#1879 - RTDT-626_Add_test_epoch

Pull Request - State: closed - Opened by rostyslavhereha 12 months ago - 1 comment

#1878 - [Bug] TypeError: init() got an unexpected keyword argument 'init_cfg'

Issue - State: closed - Opened by Carina02 12 months ago - 1 comment

#1875 - Update simmim_neck.py

Pull Request - State: open - Opened by ashutoshsingh0223 12 months ago - 2 comments

#1874 - [Bug] There is a problem with the setting of optim_wrapper and param_scheduler, which always leads to the best results in the top 100 epochs.

Issue - State: open - Opened by YUNIyx 12 months ago - 1 comment

#1872 - [Feature] Training Support for Image Captioning

Issue - State: open - Opened by ShenZheng2000 about 1 year ago

#1871 - [Feature] Support for pretaining convnext-v2

Issue - State: closed - Opened by chaoer about 1 year ago - 5 comments

#1869 - [Feature] 进行自监督学习

Issue - State: open - Opened by ChenjieXu about 1 year ago

#1868 - compatible with New Config in `mmpretrain.apis.get_model`

Pull Request - State: open - Opened by youqingxiaozhua about 1 year ago - 1 comment

#1867 - [Bug] SimMIM pretrained weights DO NOT MATCH the descriptions

Issue - State: open - Opened by SheffieldCao about 1 year ago

#1866 - Support input of non three-channel image

Pull Request - State: closed - Opened by MGAMZ about 1 year ago - 2 comments

#1865 - setting learning rate

Issue - State: open - Opened by AllieLan about 1 year ago

#1862 - SwinV2 for downstream

Issue - State: open - Opened by sipie800 about 1 year ago - 1 comment

#1861 - accuracy/top1 does not meet expectations, and the predicted results are in the same category

Issue - State: open - Opened by 6sz about 1 year ago - 1 comment

#1860 - Bump version to v1.2.0

Pull Request - State: closed - Opened by mzr1996 about 1 year ago - 1 comment

#1859 - [Bug] 关于模型扩展的问题

Issue - State: open - Opened by qsd-github about 1 year ago - 5 comments

#1858 - [Bug] Error when training model - TypeError: BaseDataset.init() got an unexpected keyword argument 'split'

Issue - State: open - Opened by TNodeCode about 1 year ago - 6 comments

#1857 - Update multi_task.py

Pull Request - State: open - Opened by eantono-pge about 1 year ago - 1 comment

#1856 - [ModuleNotFoundError: No module named 'mmcv.cnn.utils.weight_init']

Issue - State: open - Opened by jinglin80 about 1 year ago

#1855 - [Feature] dose open-mmlab has project related semi-supervision?

Issue - State: open - Opened by HaoLiuHust about 1 year ago

#1854 - [Bug] can not find AveragePrecision, MultiLabelMetric,MultiLabelMetric in mmengine

Issue - State: open - Opened by zengziru about 1 year ago - 1 comment

#1853 - [Feature] Support LLaVA 1.5

Pull Request - State: closed - Opened by mzr1996 about 1 year ago

#1852 - [Bug] error when training vit model with torch.compile

Issue - State: open - Opened by HaoLiuHust about 1 year ago

#1850 - about the loss of Mocov3, no decreasing?

Issue - State: open - Opened by RobinHan24 about 1 year ago

#1848 - [Bug] Error occurred while obtaining the model parameters

Issue - State: open - Opened by fengchezhendexiatian about 1 year ago

#1847 - [Bug] Why is deit-tiny's acc@1 different from the official result?

Issue - State: open - Opened by seungwoo-s about 1 year ago

#1846 - minigpt4权重来源确认

Issue - State: open - Opened by FDInSky about 1 year ago

#1845 - mobileone-s1 prediction scores are low

Issue - State: open - Opened by Phil-Lin about 1 year ago - 1 comment

#1844 - [Feature] How to train a classifier by Linear Probing?

Issue - State: open - Opened by jwwangchn about 1 year ago - 2 comments

#1843 - [Feature] 是否支持通过torch hub中的结构和权重训练？

Issue - State: open - Opened by HaoLiuHust about 1 year ago

#1842 - [Bug] Duplicate NORM Application in VisionTransformer when setting out_type=avg_featmap

Issue - State: open - Opened by lovelyczli about 1 year ago

#1841 - [Bug] I tested vit-base-p16_64xb64_in1k-384px.py on the Oxford 102 Category Flower dataset, but I got a top1-accuracy greater than 1.

Issue - State: open - Opened by six-wood over 1 year ago - 1 comment

#1840 - [Bug] Error C++17 or later compatible compiler is required to

Issue - State: open - Opened by zhaohad over 1 year ago - 1 comment

#1839 - [Bug] ImageClassificationInferencer may be initialized incorrectly in non-'mmpretrain' scope.

Issue - State: open - Opened by ZhangYouyi over 1 year ago - 1 comment

#1838 - [Bug] Error when loading weights for models

Issue - State: open - Opened by OnlyForWW over 1 year ago - 2 comments

#1837 - [Bug] 2node distribued train got timeout error [E ProcessGroupNCCL.cpp:587] [Rank 14] Watchdog caught collective operation timeout: WorkNCCL(OpType=ALLREDUCE, Timeout(ms)=1800000) ran for 1800834 milliseconds before timing out.

Issue - State: closed - Opened by ray-lee-94 over 1 year ago - 1 comment

GitHub / open-mmlab/mmpretrain issues and pull requests