oneflow

oneflow copied to clipboard

## This PR is done: - [x] Fix: https://github.com/Oneflow-Inc/OneTeam/issues/1571

lixiang007666

请问oneflow支持的分布式训练是否有k8s部署的方式

3

是否提供k8s下的operator支持oneflow的分布式训练任务的部署？

andyzheung

Further concern about oneflow.nn.FakeQuantization

1

But when the input is outside the range [quant_min,quant_max], shouldn't the gradient be 0.0 instead of 1.0? The following code snippet sets both quant_min and quant_max to 0 and defines...

xxxyyyzzz12345

## This PR is done: - [x] weight 一类的 op 不测试 sbp 为 partial_sum 的情况，并打开相关 op 的 Graph global test。 - [x] 重构 rnn_cell op 的测试脚本。

lixiang007666

## Summary A short description about the bug/issue ## Code to reproduce bug Please post a minimal example to repro the bug. GitHub Gist or repo is highly recommended. torch.stft链接：https://pytorch.org/docs/stable/generated/torch.stft.html?highlight=stft#torch.stft

hjshd

## This PR is done: - [x] 修复多卡 Global 环境下，affine op 的 check 错误。

lixiang007666

List styled ErrorFrame

5

将vector形式的StackedError重构成list形式的ErrorFrame。杜绝某些极端情况下显示错误和微量的内存泄漏。

lixinqi

op check error

5

liufengwei0103

liufengwei0103

Tmp compute x stream wait

1

本pr合并了 https://github.com/Oneflow-Inc/oneflow/pull/8570 和 https://github.com/Oneflow-Inc/oneflow/pull/8571 。不用合并，只是为了看联合优化效果。 本地的性能优化结果，对比oneflow和pytorch： 第一次运行 ``` lixinqi@oneflow-22:~/oneflow$ bash ci/test/test_speed_multi_client.sh OneFlow resnet50 time: 16.6ms (= 3329.9ms / 200, input_shape=[1, 3, 224, 224]) PyTorch resnet50 time: 29.0ms (= 5793.1ms /...

lixinqi