Skip to content
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

Dev tmp odd channels #23

Closed
wants to merge 9 commits into from
Closed

Dev tmp odd channels #23

wants to merge 9 commits into from

Conversation

UpasanaSridhar
Copy link
Contributor

No description provided.

printf("input col stride %d\n", input_col_stride);

#endif
FLOAT_ABSTRACT_OP_END(op_type, op_class, step, a_cur, b_cur, c_cur, W_elements, _C_ob);
Copy link
Collaborator

Choose a reason for hiding this comment

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

I think, this is where arm/intrinsics.h is failing to compile: _C_ob is not constexpr as macro assumes (or the compiler is not advanced enough to ignore constexpr...why doesn't this fail on Zen2 platform?).

throw std::invalid_argument(
"AveragePool2D<float> ERROR: in_channels unsupported.");
}
// else
Copy link
Collaborator

Choose a reason for hiding this comment

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

missing code for odd channels?

Copy link
Contributor Author

@UpasanaSridhar UpasanaSridhar Jan 25, 2024

Choose a reason for hiding this comment

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

Haven't Tested odd channels for this layer type yet. Added the else back

@@ -270,7 +276,7 @@ void inline kernel(
if (op_type == OP_MAX_POOL || op_type == OP_MUL)
{
/// @note using platform C_ob
Copy link
Collaborator

Choose a reason for hiding this comment

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

With the change from FLOAT_C_ob to _C_ob, this comment is no longer correct. Remove?

@@ -303,7 +309,7 @@ void inline kernel(
for (uint32_t ii = 0; ii < _F_cb / _UNROLL; ii++)
{
/// @note using platform C_ob
Copy link
Collaborator

Choose a reason for hiding this comment

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

With the change from FLOAT_C_ob to _C_ob, this comment is no longer correct. Remove?

* O: [G/G_b, K/K_b, O_h, O_w, G_b, K_b]
*
* For the case where the number of channels is not a multiple of the blocking size,
* I:
Copy link
Collaborator

Choose a reason for hiding this comment

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

Missing new documentation

{
// printf("Generic handling for input channels that are not a multiple\n");
// printf("call parameters: ");
if (stride == 1)
Copy link
Collaborator

Choose a reason for hiding this comment

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

ERROR(?) This new if was not indented properly and the closing brace is missing. I think it belongs at line 447-448. This also indicates that there is no unit test exercising this function.

throw std::invalid_argument(
"ReLUActivation<float> ERROR: in_channels unsupported.");
}

Copy link
Collaborator

@mcmillan03 mcmillan03 Feb 16, 2024

Choose a reason for hiding this comment

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

TODO: add odd channel support for LeakyReLUActivation, Dropout, Softmax, Upsample, Accum, Bias, PartialBias?

@mcmillan03 mcmillan03 closed this Jan 7, 2025
@mcmillan03 mcmillan03 deleted the dev_tmp_odd_channels branch January 7, 2025 16:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants