SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

tests/roformer: fix couple roformer tests on gpus (#38570)

Browse Source 
Fix "RuntimeError: Expected all tensors to be on the same device,
but found at least two devices, cuda:0 and cpu" error running the
following roformer tests on GPUs (CUDA or XPU):

```
tests/models/roformer/test_modeling_roformer.py::RoFormerSinusoidalPositionalEmbeddingTest::test_basic
tests/models/roformer/test_modeling_roformer.py::RoFormerSelfAttentionRotaryPositionEmbeddingTest::test_apply_rotary_position_embeddings
```

Signed-off-by: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
This commit is contained in:
Dmitry Rogozhkin
2025-06-04 09:45:56 -07:00
committed by GitHub
parent b9c17c5dc0
commit 8046aff520
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/models/roformer/test_modeling_roformer.py
View File

@@ -513,8 +513,9 @@ class RoFormerSinusoidalPositionalEmbeddingTest(unittest.TestCase):
def test_basic(self): def test_basic(self):
input_ids = torch.tensor([[4, 10]], dtype=torch.long, device=torch_device) input_ids = torch.tensor([[4, 10]], dtype=torch.long, device=torch_device)
emb1 = RoFormerSinusoidalPositionalEmbedding(num_positions=6, embedding_dim=6).to(torch_device) emb1 = RoFormerSinusoidalPositionalEmbedding(num_positions=6, embedding_dim=6)
emb1._init_weight() emb1._init_weight()
emb1 = emb1.to(torch_device)
emb = emb1(input_ids.shape) emb = emb1(input_ids.shape)
desired_weights = torch.tensor( desired_weights = torch.tensor(
[[0.0000, 0.0000, 0.0000, 1.0000, 1.0000, 1.0000], [0.8415, 0.0464, 0.0022, 0.5403, 0.9989, 1.0000]] [[0.0000, 0.0000, 0.0000, 1.0000, 1.0000, 1.0000], [0.8415, 0.0464, 0.0022, 0.5403, 0.9989, 1.0000]]
@@ -554,8 +555,9 @@ class RoFormerSelfAttentionRotaryPositionEmbeddingTest(unittest.TestCase):
key_layer = ( key_layer = (
-torch.arange(2 * 12 * 16 * 64, dtype=torch.float, device=torch_device).reshape(2, 12, 16, 64) / 100 -torch.arange(2 * 12 * 16 * 64, dtype=torch.float, device=torch_device).reshape(2, 12, 16, 64) / 100
).to(torch_device) ).to(torch_device)
embed_positions = RoFormerSinusoidalPositionalEmbedding(num_positions=32, embedding_dim=64).to(torch_device) embed_positions = RoFormerSinusoidalPositionalEmbedding(num_positions=32, embedding_dim=64)
embed_positions._init_weight() embed_positions._init_weight()
embed_positions = embed_positions.to(torch_device)
sinusoidal_pos = embed_positions([2, 16, 768])[None, None, :, :] sinusoidal_pos = embed_positions([2, 16, 768])[None, None, :, :]
query_layer, key_layer = RoFormerSelfAttention.apply_rotary_position_embeddings( query_layer, key_layer = RoFormerSelfAttention.apply_rotary_position_embeddings(