Revert "Add ttnn.squeeze for lowering embedding when the rank of input is 1"

This reverts commit 39b81e7.

Author: kevinwuTT

tests/lowering/embedding/test_embedding.py

@@ -2,7 +2,6 @@
 import torch_ttnn
 import pytest
 import ttnn
-from tests.utils import assert_with_pcc
 
 
 class EmbeddingModule(torch.nn.Module):
@@ -24,7 +23,7 @@ def forward(self, input, weights):
 
 @pytest.mark.parametrize(
     "input_shapes",
-    [[((1, 2, 4, 5), (4, 3, 2, 9)), (10, 4)], [((0, 1, 2, 3)), (4, 2)]],
+    [[((1, 2, 4, 5), (4, 3, 2, 9)), (10, 4)]],
 )
 def test_embedding(device, input_shapes):
     m = EmbeddingModule()
@@ -42,7 +41,7 @@ def test_embedding(device, input_shapes):
     nodes = list(option._out_fx_graphs[0].nodes)
     assert [node.target for node in nodes].count(ttnn.embedding) == 1
     # Check inference result
-    assert_with_pcc(result_before, result_after)
+    assert torch.allclose(result_before, result_after)
 
 
 @pytest.mark.parametrize(

torch_ttnn/passes/lowering/to_tt_pass.py

@@ -558,13 +558,10 @@ def reshape_1d(code, args=args, kwargs=kwargs):
                 return g.call_function(ttnn.add, args=(beta_node, new_node))
 
             if node.target == torch.ops.aten.embedding.default:
-                tensor_meta = args[1].meta.get("val")
-                tiled = False if tensor_meta is None else tensor_meta.size()[-1] % ttnn.TILE_SIZE == 0
+                tiled = args[1].meta.get("val")
+                tiled = False if tiled is None else tiled.size()[-1] % ttnn.TILE_SIZE == 0
                 layout = TtnnTileLayout() if tiled else TtnnRowMajorLayout()
                 tensor = g.call_function(ttnn.embedding, (args[1], args[0]), {"layout": layout})
-                # TODO: Remove this squeeze when issue is fixed: https://github.com/tenstorrent/pytorch2.0_ttnn/issues/660
-                if len(tensor_meta.size()) == 1:
-                    tensor = g.call_function(ttnn.squeeze, (tensor, 0))
                 return tensor if tiled else g.call_function(ttnn.to_layout, (tensor, TtnnTileLayout()))
 
             if node.target == torch.ops.aten._log_softmax.default: