Skip to content

Commit 2baf479

Browse files
authored
Merge branch 'main' into savitha/og2-readme-metrics-update
2 parents 965d885 + 85d0c16 commit 2baf479

2 files changed

Lines changed: 10 additions & 4 deletions

File tree

bionemo-recipes/recipes/esm2_accelerate_te/tests/test_accelerate_amplify.py

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -20,6 +20,8 @@
2020
recipe tests.
2121
"""
2222

23+
import pytest
24+
2325
# Local helper function import, resolved in conftest.py
2426
from launch import launch_accelerate, requires_multi_gpu
2527

@@ -39,6 +41,10 @@ def test_te_with_fp8_config(tmp_path):
3941
assert train_loss < 3.0, f"Final train_loss {train_loss} should be less than 3.0"
4042

4143

44+
@pytest.mark.xfail(
45+
strict=True,
46+
reason="BIO-466: AMPLIFY HF model does not implement get_input_embeddings, required by accelerate FSDP2 (transformers>=5.6).",
47+
)
4248
def test_te_with_fsdp2_config(tmp_path):
4349
train_loss = launch_accelerate("fsdp2_te.yaml", tmp_path, 1, "L0_sanity_amplify")
4450
assert train_loss < 3.0, f"Final train_loss {train_loss} should be less than 3.0"
@@ -56,6 +62,10 @@ def test_te_with_fp8_config_two_gpus(tmp_path):
5662
assert train_loss < 3.0, f"Final train_loss {train_loss} should be less than 3.0"
5763

5864

65+
@pytest.mark.xfail(
66+
strict=True,
67+
reason="BIO-466: AMPLIFY HF model does not implement get_input_embeddings, required by accelerate FSDP2 (transformers>=5.6).",
68+
)
5969
@requires_multi_gpu
6070
def test_te_with_fsdp2_config_two_gpus(tmp_path):
6171
train_loss = launch_accelerate("fsdp2_te.yaml", tmp_path, 2, "L0_sanity_amplify")

bionemo-recipes/recipes/esm2_accelerate_te/tests/test_accelerate_esm2.py

Lines changed: 0 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -13,8 +13,6 @@
1313
# See the License for the specific language governing permissions and
1414
# limitations under the License.
1515

16-
import pytest
17-
1816
# Local helper function import, resolved in conftest.py
1917
from launch import launch_accelerate, requires_multi_gpu
2018

@@ -49,7 +47,6 @@ def test_hf_with_default_config(tmp_path):
4947
assert train_loss < 3.0, f"Final train_loss {train_loss} should be less than 3.0"
5048

5149

52-
@pytest.mark.xfail(reason="BIONEMO-3331: FSDP2 and HF model failing with 25.11+ torch container.")
5350
def test_hf_with_fsdp2_config(tmp_path):
5451
train_loss = launch_accelerate("fsdp2_hf.yaml", tmp_path, 1, "L0_sanity", "model_tag=facebook/esm2_t6_8M_UR50D")
5552
assert train_loss < 3.0, f"Final train_loss {train_loss} should be less than 3.0"
@@ -85,7 +82,6 @@ def test_hf_with_fsdp1_config_two_gpus(tmp_path):
8582
assert train_loss < 3.0, f"Final train_loss {train_loss} should be less than 3.0"
8683

8784

88-
@pytest.mark.xfail(reason="BIONEMO-3331: FSDP2 and HF model failing with 25.11+ torch container.")
8985
@requires_multi_gpu
9086
def test_hf_with_fsdp2_config_two_gpus(tmp_path):
9187
train_loss = launch_accelerate("fsdp2_hf.yaml", tmp_path, 2, "L0_sanity", "model_tag=facebook/esm2_t6_8M_UR50D")

0 commit comments

Comments
 (0)