#0: Fix failing test case for width sharded non-32 multiple output width #16224
Conversation
sankarmanoj-tt
requested review from
mywoodstock,
shwetankTT and
pavlejosipovic
as code owners
sankarmanoj-tt
force-pushed
the
smanoj/fix_test_case_7
branch
from
8db5c45 to
41070e7
Compare
ttnn/cpp/ttnn/operations/conv/conv2d/prepare_conv2d_weights.cpp
Outdated
Show resolved
Hide resolved
|
It would be nice to remove this test case from |
sankarmanoj-tt
force-pushed
the
smanoj/fix_test_case_7
branch
from
41070e7 to
1007506
Compare
|
@pavlejosipovic I've added a UT. It exposed some more bugs, which I have also fixed. I removed the test from the sweep. |
sankarmanoj-tt
force-pushed
the
smanoj/fix_test_case_7
branch
from
83628b8 to
3f80dad
Compare
| @@ -530,20 +530,18 @@ def test_conv_features_multi_device( | |||
| @pytest.mark.parametrize( | |||
| "batch_size, output_channels, input_channels, input_height, input_width, filter_height, filter_width, pad_h, pad_w, act_block_w_div", | |||
| ( | |||
| (2, 128, 128, 9, 9, 3, 3, 0, 0, 1), | |||
There was a problem hiding this comment.
any reason why where few test cases removed?
There was a problem hiding this comment.
They were unnecessary.
There was a problem hiding this comment.
test coverage for width sharding is miniscule as is, but if you think they are not needed ok
| @@ -191,6 +192,7 @@ Result conv2d( | |||
|
|
|||
| if (bypass_halo) { | |||
| if (input_tensor_post_tm.layout() == Layout::TILE) { | |||
| input_tensor_post_tm = ttnn::reshape(input_tensor_post_tm, input_tensor_post_tm.get_padded_shape()); | |||
There was a problem hiding this comment.
why do we need this reshape here?
There was a problem hiding this comment.
to_layout was not using the padded shape, and was instead using the logical shape. This was causing an error.
There was a problem hiding this comment.
at least we should log an issue with unit tests and reference it here.
Would be better if we can just fix to layout.
Halo with untialize is doing this properly?
Problem description
tests/sweep_framework/sweeps/conv2d/short/conv2d_short_sweep.py::test_conv2d_localrun_fail_only[device_params0-input_spec7]test case fails.What's changed
Fixed width sharded weights preparation, where out_channels % 32 != 0
Checklist