iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch ModelScope模型tensorboard应该怎么配置呢?
2000元阿里云代金券免费领取,2核4G云服务器仅664元/3年,新老用户都有优惠,立即抢购>>>
您参考这个文档在cfg_modify_fn中加一下TensorboardHook,https://modelscope.cn/docs/%E8%AE%AD%E7%BB%83%E7%9A%84%E8%AF%A6%E7%BB%86%E5%8F%82%E6%95%B0 此回答整理自钉群“魔搭ModelScope开发者联盟群 ①”
2000元阿里云代金券免费领取,2核4G云服务器仅664元/3年,新老用户都有优惠,立即抢购>>>
使用PyTorch的torch.utils.tensorboard模块,可以在训练循环中记录指标。
from torch.utils.tensorboard import SummaryWriter
# 创建一个SummaryWriter实例
writer = SummaryWriter('runs/my_experiment')
# 假设你有一个训练循环
for epoch in range(num_epochs):
for batch in dataloader:
# ... 训练代码 ...
# 记录损失和指标
loss = ...
accuracy = ...
# 将损失和准确率添加到SummaryWriter
writer.add_scalar('loss', loss, global_step=epoch)
writer.add_scalar('accuracy', accuracy, global_step=epoch)
# 每轮结束后,你可能还想记录其他信息,如模型的权重和梯度