search

SwanHubX / SwanLab

⚡️SwanLab: your ML experiment notebook. 你的AI实验笔记本，跟踪与可视化你的机器学习全流程
https://swanlab.cn
Apache License 2.0
387 stars 35 forks source link

[FEATURE] record model info #167

Open Zeyi-Lin opened 8 months ago

Zeyi-Lin commented 8 months ago

根据 #166 中提到的打印模型权重信息的需求，计划很快上线相关功能

👾 What does the proposed API look like

import swanlab
import torchvision
from torchvision.models import ResNet50_Weights

# load pretrain resnet50
model = torchvision.models.resnet50(weights=ResNet50_Weights.IMAGENET1K_V2)

run = swanlab.init()
run.watch_model_graph(model)
Zeyi-Lin commented 7 months ago

UI like：

image
ZhikangNiu commented 7 months ago
  1. 感觉api叫做watch_model_graph 容易跟梯度图关联起来
  2. 好像更喜欢的是torchinfo那样的model的结构打印
  3. 如果对我来说，我可能更喜欢netron这种的可视化
  4. 可能需要检查下如果是多卡包裹模型的情况
Zeyi-Lin commented 7 months ago
  1. 感觉api叫做watch_model_graph 容易跟梯度图关联起来
  2. 好像更喜欢的是torchinfo那样的model的结构打印
  3. 如果对我来说，我可能更喜欢netron这种的可视化
  4. 可能需要检查下如果是多卡包裹模型的情况

❤️nice advice