vision

vision copied to clipboard

基于project里面提供的train.py运行以下代码 ``` python train.py \ --data-path /home/kunyangzhou/project/dataset/coco \ --dataset coco \ --model retinanet_resnet50_fpn \ --batch-size 8 \ --pretrained \ --test-only ``` 然后出现报错 ``` Stack trace (most recent call last) in...

zkyseu

``` File ~/.cache/pants/named_caches/pex_root/venvs/s/526a591d/venv/lib/python3.8/site-packages/oneflow/framework/dtype.py:48 43 def convert_proto_dtype_to_oneflow_dtype(proto_dtype): 44 return oneflow._oneflow_internal.deprecated.GetDTypeByDataType(proto_dtype) 47 _ONEFLOW_DTYPE_TO_NUMPY_DTYPE = { ---> 48 oneflow.bool: np.bool, 49 oneflow.float: np.float32, 50 oneflow.float16: np.float16, 51 oneflow.float32: np.float32, 52 oneflow.float64: np.double, 53...

da-liii

+ https://libraries.io/pypi/torchvision + https://libraries.io/pypi/flowvision 比如TorchVision有六个依赖。

da-liii

当前release版本和Oneflow0.9.0没对齐

1

当前flowvision的release版本0.2.1和Oneflow0.9.0没对齐，比如oneflow0.9.0将`Tensor.zeros_`更改成了`Tensor.zero_`。 https://github.com/Oneflow-Inc/oneflow/pull/7593 flowvision这边的master分支在这个问题上已经更新了，但是release还没对齐。

SanftMonster

Add Beit model and pretrained weights

2

Add Beit model and pretrained weights

triple-mu

Add convit into models

1

Add convit into models

triple-mu

## ResNet-50 训练 参照当前 vision 下的 project 复现 resnet-50 训练和精度对齐。 ## 参考 - pytorch：[resent](https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py) - timm：[resnet](https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/resnet.py) ### 主要目标 - [ ] 2022.05.11 - 2022.5.12：熟悉 vision 下的分类模型训练代码，数据集配置并跑通。 - [ ] 2022.05.12...

triple-mu

Add convit model and pretrained weights

triple-mu

Add xcit model and pretrained weights

triple-mu

 ``` python import oneflow as flow import torch import numpy as np a = np.random.randn(3,3).astype(np.float32) b = 2 torch_a = torch.from_numpy(a) flow_a = flow.from_numpy(a) print(torch.div(torch_a,b,rounding_mode='floor')) print(flow.div(flow_a,b).floor()) print(flow.div(flow_a,b,rounding_mode='floor')) ```

triple-mu