#4957: remove construct_2d_input_tensor, instead use pytorch to pad i…

…nput tensor

tests/tt_eager/python_api_testing/unit_testing/test_configs_for_untilize_with_halo_and_conv.py

@@ -119,14 +119,14 @@ def test_generate_all_configs_and_references(
     input_padded_height = input_h + 2 * pad_h
     # Generate following configs by tracing conv -
     logger.info("Trace conv and generate following configs - pad_metadata and data_top_left_indices.")
-    (
-        pad_metadata,
-        data_top_left_indices,
-        input_padded_tensor,
-    ) = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
-        conv_params, input_nchw_shape, input_pyt_tensor.reshape(-1).tolist()
+    pad_metadata, data_top_left_indices = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
+        conv_params, input_nchw_shape
     )
 
+    logger.info("Generate input tensor")
+    input_padded_pyt_tensor = torch.nn.functional.pad(input_pyt_tensor, (pad_w, pad_w, pad_h, pad_h), value=0)
+    input_padded_pyt_tensor = input_padded_pyt_tensor.permute(0, 2, 3, 1)
+    input_padded_tensor = input_padded_pyt_tensor.reshape(-1).tolist()
     # run trace conv reference to validate pad_metadata and data_top_left_indices
     logger.info("Validate pad_metadata and data_top_left_indices.")
 

tests/tt_eager/python_api_testing/unit_testing/test_untilize_with_halo_v2.py

@@ -324,16 +324,16 @@ def test_generate_all_configs_and_references(
     input_padded_height = input_h + 2 * pad_h
     # Generate following configs by tracing conv -
     logger.info("Trace conv and generate follwing configs - pad_metadata and data_top_left_indices.")
-    (
-        pad_metadata,
-        data_top_left_indices,
-        input_padded_tensor,
-    ) = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
-        conv_params, input_nchw_shape, input_pyt_tensor.reshape(-1).tolist()
+    pad_metadata, data_top_left_indices = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
+        conv_params, input_nchw_shape
     )
     # # print("Data top left indices - ", data_top_left_indices)
     # # print("Pad meta data -", pad_metadata)
 
+    logger.info("Generate input tensor")
+    input_padded_pyt_tensor = torch.nn.functional.pad(input_pyt_tensor, (pad_w, pad_w, pad_h, pad_h), value=0)
+    input_padded_pyt_tensor = input_padded_pyt_tensor.permute(0, 2, 3, 1)
+    input_padded_tensor = input_padded_pyt_tensor.reshape(-1).tolist()
     # run trace conv reference to validate pad_metadata and data_top_left_indices
 
     # Generate more configs -
@@ -349,7 +349,6 @@ def test_generate_all_configs_and_references(
         num_cores_nhw,
         filter_h,
         filter_w,
-        # input_c,
     )
     # print("req_conv_input_shard_start_end-", req_conv_input_shard_start_end)
     # print("tensor_metadata-", tensor_metadata)

tt_eager/tt_dnn/op_library/sliding_window_op_infra/tt_py_composite_conv.py

@@ -593,17 +593,10 @@ def set_op_configs(
 
             input_padded_width = input_w + 2 * pad_w
 
-            dummy = torch.rand(batch_size * input_h * input_w, dtype=torch.bfloat16)
-            dummy = torch.reshape(dummy, input_nchw_shape)
-
             # TODO: We should remove C from input_nchw_shape since none of the specs depend on it
             # TODO: Pass sliding_window_op_params instead of conv_param?
-            (
-                pad_metadata,
-                data_top_left_indices,
-                _,
-            ) = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
-                conv_params, input_nchw_shape, dummy.reshape(-1).tolist()
+            pad_metadata, data_top_left_indices = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
+                conv_params, input_nchw_shape
             )
 
             req_conv_input_shard_start_end, tensor_metadata = decompose_conv_into_shards_and_generate_tensor_metadata(

tt_eager/tt_dnn/op_library/sliding_window_op_infra/tt_py_max_pool.py

@@ -227,17 +227,8 @@ def set_op_configs(self, sliding_window_op_params_hash, reader_patterns_cache):
 
             input_padded_width = input_w + 2 * pad_w
 
-            dummy = torch.rand(batch_size * input_h * input_w, dtype=torch.bfloat16)
-            dummy = torch.reshape(dummy, input_nchw_shape)
-
-            (
-                pad_metadata,
-                data_top_left_indices,
-                _,
-            ) = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
-                (1, 1, window_h, window_w, stride_h, stride_w, pad_h, pad_w, 1, 1),
-                input_nchw_shape,
-                dummy.reshape(-1).tolist(),
+            pad_metadata, data_top_left_indices = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
+                (1, 1, window_h, window_w, stride_h, stride_w, pad_h, pad_w, 1, 1), input_nchw_shape
             )
 
             req_conv_input_shard_start_end, tensor_metadata = decompose_conv_into_shards_and_generate_tensor_metadata(

tt_eager/tt_dnn/op_library/sliding_window_op_infra/tt_py_untilize_with_halo.py

@@ -110,15 +110,8 @@ def set_op_configs(
             # output_channels, input_channels, filter_h, filter_w, stride_h, stride_w, pad_h, pad_w, dilation, groups
             sliding_window_op_all_params = [1, 1, window_h, window_w, stride_h, stride_w, pad_h, pad_w, 1, 1]
             input_nchw_shape = [input_n, 1, input_h, input_w]
-
-            dummy = torch.rand(input_n * input_h * input_w, dtype=torch.bfloat16)
-            dummy = torch.reshape(dummy, input_nchw_shape)
-            (
-                pad_metadata,
-                data_top_left_indices,
-                _,
-            ) = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
-                sliding_window_op_all_params, input_nchw_shape, dummy.reshape(-1).tolist()
+            pad_metadata, data_top_left_indices = trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
+                sliding_window_op_all_params, input_nchw_shape
             )
             sliding_window_output_shard_nhw_size = get_sliding_window_op_output_shard_nhw_size(
                 num_cores_nhw,

tt_eager/tt_dnn/op_library/sliding_window_op_infra/untilize_with_halo_config_generation_and_validation.py

@@ -4,16 +4,13 @@
 
 import torch
 import numpy as np
-from loguru import logger
 
 from tests.tt_eager.python_api_testing.sweep_tests.comparison_funcs import comp_equal, comp_allclose_and_pcc
 
 
 def trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
-    conv_params, input_nchw_shape, input_tensor, pad_val: torch.int16 = 0x0
+    conv_params, input_nchw_shape  # , input_tensor, pad_val: torch.int16 = 0x0
 ):
-    if pad_val == 0xF7FF:
-        pad_val = -1.03e34  ## TODO: how to do this in python properly???
     assert len(conv_params) == 10
     output_channels, input_channels, filter_h, filter_w, stride_h, stride_w, pad_h, pad_w, dilation, groups = [
         conv_params[i] for i in range(10)
@@ -22,9 +19,6 @@ def trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
     assert len(input_nchw_shape) == 4
     input_n, input_c, input_h, input_w = [input_nchw_shape[i] for i in range(4)]
 
-    input_tensor_nchw = np.reshape(input_tensor, input_nchw_shape)
-    input_tensor_nhwc = np.transpose(input_tensor_nchw, (0, 2, 3, 1))
-    input_tensor_nhwc = np.reshape(input_tensor_nhwc, (np.prod(input_nchw_shape)))
     # image 1 data
     # 1  2  3  4  5  6  7  8
     # 9  10 11 12 13 14 15 16
@@ -54,41 +48,33 @@ def trace_conv_to_generate_data_top_left_indices_and_pad_metadata(
     # We encode above shown padded tensor into pad_metadata (list of boolean - true if padding location)
     # pad_meta_data: [true, true, ..., false, ...]
     index = 0
-    input_idx = 0
-    input_pad_idx = 0
     padded_input_h = input_h + (2 * pad_h)
     padded_input_w = input_w + (2 * pad_w)
     pad_metadata = np.full(input_n * padded_input_h * padded_input_w, False, dtype=bool)
-    input_padded_tensor = np.full(
-        input_n * input_c * padded_input_h * padded_input_w, pad_val, dtype=type(input_tensor_nhwc[0])
-    )
     for n in range(input_n):
         for h in range(padded_input_h):
             for w in range(padded_input_w):
                 if h < pad_h or h >= (input_h + pad_h) or w < pad_w or w >= (input_w + pad_w):
                     pad_metadata[index] = True
-                    input_pad_idx += input_c
-                else:
-                    for c in range(input_c):
-                        input_padded_tensor[input_pad_idx + c] = input_tensor_nhwc[input_idx]
-                        input_idx += 1
-                    input_pad_idx += input_c
                 index += 1
 
     # TODO: add support for dilation > 1
     output_h = ((int)((padded_input_h - filter_h) / stride_h)) + 1
     output_w = ((int)((padded_input_w - filter_w) / stride_w)) + 1
     # generate a list of input indices corresponding to the top left position of sliding window
     # the index refers to the location in the padded tensor
-    data_top_left_indices = []
+    # data_top_left_indices = []
+    index = 0
+    data_top_left_indices = np.full(input_n * output_h * output_w, 0, dtype=int)
     for n in range(input_n):
         for oh in range(output_h):
             for ow in range(output_w):
                 ih = oh * stride_h
                 iw = ow * stride_w
                 channel_idx = (n * padded_input_h * padded_input_w) + (ih * padded_input_w) + iw
-                data_top_left_indices.append(channel_idx)
-    return pad_metadata.tolist(), data_top_left_indices, input_padded_tensor.tolist()
+                data_top_left_indices[index] = channel_idx
+                index += 1
+    return pad_metadata.tolist(), data_top_left_indices.tolist()
 
 
 def validate_input_padded_tensor_and_data_top_left_indices_and_pad_metadata(