initial commit (#2223)

src/sparseml/transformers/sparsification/modification/base.py

@@ -24,7 +24,7 @@
 __all__ = ["check_transformers_version"]
 
 _TRANSFORMERS_MIN_VERSION = "4.39.0"
-_TRANSFORMERS_MAX_VERSION = "4.39.2"
+_TRANSFORMERS_MAX_VERSION = "4.39.3"
 
 
 def check_transformers_version(
@@ -56,7 +56,7 @@ def check_transformers_version(
         _LOGGER.warning(
             "Attempting to modify the transformers model to support "
             "the SparseML-specific functionalities. However, the detected "
-            f"transformers version ({current_version}) does not fall within the"
+            f"transformers version ({current_version}) does not fall within the "
             f"supported version range ({min_version} - {max_version}). "
             "This may lead to unexpected behavior. Please ensure that the "
             "correct transformers version is installed."

tests/sparseml/transformers/test_recipe_compatibility.py

@@ -0,0 +1,62 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import shutil
+
+import pytest
+
+import sparseml.core.session as session_manager
+from huggingface_hub import snapshot_download
+from sparseml.transformers import SparseAutoModelForCausalLM
+
+
+@pytest.fixture
+def model_path(tmp_path):
+    yield snapshot_download("stas/tiny-random-llama-2", local_dir=tmp_path)
+    shutil.rmtree(tmp_path)
+
+
+@pytest.fixture
+def recipe():
+    return """test_stage:
+  obcq_modifiers:
+    QuantizationModifier:
+      ignore:
+        - LlamaRotaryEmbedding
+        - LlamaRMSNorm
+        - {silu_activation}
+      scheme_overrides:
+        Embedding:
+          input_activations: null
+          weights:
+            num_bits: 8
+            symmetric: false"""
+
+
+def test_silu_alias_same_output(recipe, model_path):
+    model_ = SparseAutoModelForCausalLM.from_pretrained(
+        model_path, recipe=recipe.format(silu_activation="SiLU")
+    )
+    session_manager.create_session()
+    session_manager.active_session().reset()
+    model = SparseAutoModelForCausalLM.from_pretrained(
+        model_path, recipe=recipe.format(silu_activation="SiLUActivation")
+    )
+
+    dummy_input = model.dummy_inputs
+
+    out = model(**dummy_input)
+    out_ = model_(**dummy_input)
+
+    out.logits.allclose(out_.logits)