patch transformers to allow CP + FA2

tests/monkeypatch/test_trainer_context_parallel_patch.py

@@ -0,0 +1,67 @@
+"""Tests for the Trainer context parallel patch."""
+
+import pytest
+from transformers import Trainer
+
+from axolotl.monkeypatch.transformers.trainer_context_parallel import (
+    GUARD_PATTERN,
+    PATCHED_GUARD,
+    patch_prepare_context_parallel_inputs,
+)
+
+
+@pytest.fixture
+def restore_trainer_prepare_method():
+    """Ensure Trainer._prepare_context_parallel_inputs is restored after a test."""
+    original_method = getattr(
+        Trainer,
+        "_original_prepare_context_parallel_inputs",
+        Trainer._prepare_context_parallel_inputs,
+    )
+    patched_attr_present = hasattr(
+        Trainer, "_axolotl_prepare_context_parallel_inputs_patched"
+    )
+
+    yield
+
+    Trainer._prepare_context_parallel_inputs = original_method
+    if patched_attr_present:
+        delattr(Trainer, "_axolotl_prepare_context_parallel_inputs_patched")
+    if hasattr(Trainer, "_original_prepare_context_parallel_inputs"):
+        delattr(Trainer, "_original_prepare_context_parallel_inputs")
+    if hasattr(Trainer, "_axolotl_prepare_context_parallel_inputs_source"):
+        delattr(Trainer, "_axolotl_prepare_context_parallel_inputs_source")
+
+
+def test_patch_replaces_guard_for_flash_attention(restore_trainer_prepare_method):
+    """Patch should swap the guard to allow sdpa or flash attention."""
+    # Ensure we start from the unpatched method
+    if hasattr(Trainer, "_original_prepare_context_parallel_inputs"):
+        Trainer._prepare_context_parallel_inputs = (
+            Trainer._original_prepare_context_parallel_inputs
+        )
+        delattr(Trainer, "_original_prepare_context_parallel_inputs")
+    if hasattr(Trainer, "_axolotl_prepare_context_parallel_inputs_patched"):
+        delattr(Trainer, "_axolotl_prepare_context_parallel_inputs_patched")
+
+    patch_prepare_context_parallel_inputs()
+
+    patched_method = Trainer._prepare_context_parallel_inputs
+    assert patched_method is not None
+    assert getattr(Trainer, "_axolotl_prepare_context_parallel_inputs_patched", False)
+
+    source = Trainer._axolotl_prepare_context_parallel_inputs_source
+    # Original guard should be gone, patched guard should be present
+    assert GUARD_PATTERN not in source
+    assert PATCHED_GUARD in source
+
+
+def test_patch_is_idempotent(restore_trainer_prepare_method):
+    """Calling the patch twice should leave the same patched function in place."""
+    patch_prepare_context_parallel_inputs()
+    first_patched = Trainer._prepare_context_parallel_inputs
+
+    patch_prepare_context_parallel_inputs()
+    second_patched = Trainer._prepare_context_parallel_inputs
+
+    assert first_patched is second_patched