2021-06-17 记事

2021-06-17 记事

Focal Loss的定义

理论定义:Focal Loss可以看作是一个损失函数，它使容易分类的样本权重降低，而对难分类的样本权重增加。

pandas转json

result = df.to_json(orient=”records”)
parsed = json.loads(result)
json.dumps(parsed, indent=4)

[
{
“col 1”: “a”,
“col 2”: “b”
},
{
“col 1”: “c”,
“col 2”: “d”
}
]

keras也不错

● 团体

● #Pytorch Lightning

好多都不错的 https://github.com/PyTorchLightning​

● https://github.com/PyTorchLightning/lightning-flash

PyTorchLightning/lightning-flash：用于快速原型设计、基线、微调和解决深度学习问题的任务集合。

● https://github.com/PyTorchLightning/lightning-tutorials

● https://github.com/PyTorchLightning/lightning-tutorials

● https://github.com/PyTorchLightning/deep-learning-project-template

● https://github.com/PyTorchLightning/lightning-bolts

● Fastai

https://github.com/fastai

● https://github.com/fastai/fastbook

● github主题

● pytorch-lightning

https://github.com/topics/pytorch-lightning​

● 笔记本

https://github.com/topics/jupyter-notebook

● 学习资源

● PyTorch Tutorial

PyTorch Tutorial for Deep Learning Researchers https://github.com/yunjey/pytorch-tutorial​

● Notebooks using the Hugging Face libraries hugs

https://github.com/huggingface/notebooks​

● 可以借鉴

● NN Template

https://github.com/lucmos/nn-template Generic template to bootstrap your PyTorch project. Click on  and avoid writing boilerplate code for:

● lightning-hydra-template

https://github.com/PyTorchLightning/lightning-flash

●

Deep Learning project template best practices with Pytorch Lightning, Hydra, Tensorboard. https://github.com/lkhphuc/lightning-hydra-template​

●

My repo for training neural nets using pytorch-lightning and hydra https://github.com/Erlemar/pytorch_tempest​

● keras官方示例

https://keras.io/examples/​

● 模型优化压缩

● 蒸馏资源

https://github.com/karanchahal/distiller

● net2net

https://github.com/CompVis/net2net 具有条件可逆神经网络的网络到网络转换​

● Knowledge Distillation Toolkit

This toolkit allows you to compress a machine learning model using knowledge distillation. To use this toolkit, you need to provide a teacher model, a student model, data loaders for training and validation, and an inference pipeline. This toolkit is based on PyTorch and PyTorch Lightning , so teacher and student models need to be PyTorch neural network modules, and data loaders need to be PyTorch data loaders. https://github.com/georgian-io/Knowledge-Distillation-Toolkit​