Skip to content

Conversation

@shresthamalik
Copy link
Contributor

@shresthamalik shresthamalik commented Dec 17, 2019

  • Enabled --var build to use parallel executor integrating weights-on-device and data pipelining
  • moved ngraph_var files outside the var build


// For graphs that were run through AOT
// Graph rewrite is not done, and there is no entry in catalog
// If there is not entry in catalog all outputs need to be copied
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

typo not -- > no

Copy link
Contributor

@sayantan-nervana sayantan-nervana left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One minor comment about reusing ng::join

m_use_parallel_executor = backend->executable_can_create_tensors();
#else
m_use_parallel_executor = false;
#endif
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

note to self: important section here.

void* dst_ptr = (void*)DMAHelper::base(tf_output_tensors[output_index]);
ng_outputs[output_index]->read(
dst_ptr, ng_outputs[output_index]->get_element_count() *
ng_outputs[output_index]->get_element_type().size());
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

note to self: instead of calling read on everything, call read only on output_indexes_to_be_copied

ss << val << " ";
}
cout << ss.str() << endl;
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

there is ng::join. Maybe we don't need this.

cout << title << " " << ng::join(input_vector) << "\n";

Copy link
Contributor

@vishakha-nervana vishakha-nervana left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, Do we have any c++ test for this use case?

Copy link
Contributor

@sayantan-nervana sayantan-nervana left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@shresthamalik shresthamalik merged commit b267ba1 into master Dec 20, 2019
@shresthamalik shresthamalik deleted the shrestha/var_in_compute branch December 20, 2019 18:29
@shresthamalik shresthamalik restored the shrestha/var_in_compute branch December 20, 2019 21:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants