Conv2dConfig reallocate_halo_output default to true

Idea behind this flag is to enable reallocation of
halo output buffer in place of input buffer if case
deallocate_activation flag is also set.
In this case halo output is moved up and memory
fragmentation is reduced.

In case deallocate_activation is not set reallocate_halo_output
won't have an effect as ttnn::move will be a no-op
since input buffer is not deallocated.

Current problem is that if user sets deallocate_activation
it doesn't help them with memory issue unless they
set reallocate_halo_output as well.

This change is addressing this by setting reallocate_halo_output
to true by default.

This helps by increassing pass rate in ttnn torch
traces.

tests/sweep_framework/sweeps/conv2d/short/conv2d_short_sweep.py

@@ -1620,7 +1620,6 @@ def test_conv2d_localrun(device, input_spec):
     [1, 960, 960, 24, 24, 5, 5, 1, 1, 2, 2, 960, False, 1],  # 5
     [1, 816, 816, 19, 19, 5, 5, 1, 1, 2, 2, 816, False, 1],  # 19
     [1, 816, 816, 23, 23, 5, 5, 2, 2, 0, 0, 816, False, 1],  # 20
-    [1, 1056, 1056, 96, 96, 3, 3, 2, 2, 1, 1, 4, False, 1],  # 127
     [1, 528, 528, 192, 192, 3, 3, 2, 2, 1, 1, 2, False, 1],  # 220
     [1, 2904, 2904, 48, 48, 3, 3, 2, 2, 1, 1, 11, False, 1],  # 294
     [1, 3024, 1232, 14, 14, 1, 1, 2, 2, 0, 0, 1, False, 1],  # 1421
@@ -1630,8 +1629,6 @@ def test_conv2d_localrun(device, input_spec):
     [1, 768, 3, 224, 224, 32, 32, 32, 32, 0, 0, 1, False, 1],  # 1460
     [1, 768, 3, 224, 224, 32, 32, 32, 32, 0, 0, 1, True, 1],  # 1461
     [1, 768, 3, 384, 512, 32, 32, 32, 32, 0, 0, 1, True, 1],  # 1464
-    [1, 64, 3, 720, 1280, 7, 7, 2, 2, 3, 3, 1, False, 1],  # 1471
-    [1, 64, 3, 800, 1088, 7, 7, 2, 2, 3, 3, 1, False, 1],  # 1472
     [1, 1, 64, 480, 640, 3, 3, 1, 1, 1, 1, 1, True, 1],  # 1495
     [1, 64, 64, 480, 640, 3, 3, 1, 1, 1, 1, 1, True, 1],  # 1496
     [1, 640, 1920, 32, 32, 3, 3, 1, 1, 1, 1, 1, True, 1],  # 1522

ttnn/cpp/ttnn/operations/conv/conv2d/conv2d_utils.hpp

@@ -24,8 +24,9 @@ struct Conv2dConfig {
     DataType weights_dtype = DataType::BFLOAT16;
     string activation = "";
     uint32_t input_channels_alignment = 32;
-    bool deallocate_activation = false;
-    bool reallocate_halo_output = false;
+    bool deallocate_activation = false; // If user tensor will be deallocated if it's on device.
+    bool reallocate_halo_output = true; // If true after halo device op is done, the output tensor will be reallocated.
+                                        // in case deallocate_activation is set to true.
     uint32_t act_block_h_override = 0; // This argument is ignored when shard_layout == WIDTH_SHARDED.
     uint32_t act_block_w_div = 1; // Amount by which the maximum possible act_block_width is divided. Max act_block_w = in_channels / (total_num_cores * TILE_WIDTH);
                                   // Ignored when shard_layout == HEIGHT_SHARDED or BLOCK_SHARDED