-
Notifications
You must be signed in to change notification settings - Fork 5.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[ROCM] bugfix for unittest #32392
[ROCM] bugfix for unittest #32392
Conversation
Thanks for your contribution! |
7347912
to
2b38b1f
Compare
96fd5be
to
9c66e1c
Compare
kNUM_CUDNN_FWD_ALGS, &find_count, &find_result, | ||
cudnn_workspace_ptr, workspace_size, false)); | ||
}; | ||
// if (!exhaustive_search && !deterministic) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
多一行注释删掉一下
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
已删除
@@ -32,11 +32,8 @@ if (WITH_GPU OR WITH_ROCM) | |||
file(APPEND ${pybind_file} "USE_CUDA_ONLY_OP(fused_batch_norm_act);\n") | |||
endif() | |||
# conv_fusion_op needs cudnn 7 above | |||
# HIP not support cudnnConvolutionBiasActivationForward | |||
if ((NOT WITH_ROCM) AND (NOT ${CUDNN_VERSION} VERSION_LESS 7100)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
这里需要保留NOT ${CUDNN_VERSION} VERSION_LESS 7100
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
已更新
cudnnTensorDescriptor_t cudnn_input_desc = input_desc.descriptor<T>( | ||
layout, framework::vectorize<int>(transformed_input.dims())); | ||
cudnnTensorDescriptor_t cudnn_output_desc = output_desc.descriptor<T>( | ||
layout, framework::vectorize<int>(transformed_output.dims())); | ||
cudnnFilterDescriptor_t cudnn_filter_desc = filter_desc.descriptor<T>( | ||
layout, framework::vectorize<int>(filter->dims())); | ||
// Now only support NCHW | ||
std::vector<int> bias_dim = { | ||
1, static_cast<int>(transformed_output.dims()[1]), 1, 1}; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
确认下这段代码为什么需要挪到上面去?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
已改回
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
PR types
Others
PR changes
Others
Describe
bugfix for unittest