Remove unused sparseml.export utilities (#950)

* remove sparseml utilities Signed-off-by: Kyle Sayers <kylesayrs@gmail.com> * use in model_load Signed-off-by: Kyle Sayers <kylesayrs@gmail.com> * remove use of RECIPE FILE NAME Signed-off-by: Kyle Sayers <kylesayrs@gmail.com> * rename to RECIPE_FILE_NAME, avoid circular import Signed-off-by: Kyle Sayers <kylesayrs@gmail.com> * remove qa ignore Signed-off-by: Kyle Sayers <kylesayrs@gmail.com> --------- Signed-off-by: Kyle Sayers <kylesayrs@gmail.com> Co-authored-by: Dipika Sikka <dipikasikka1@gmail.com>
vllm-project · Dec 5, 2024 · 1830382 · 1830382
1 parent 1494fd6
commit 1830382
Show file tree

Hide file tree

Showing 4 changed files with 13 additions and 447 deletions.
diff --git a/src/llmcompressor/pytorch/model_load/helpers.py b/src/llmcompressor/pytorch/model_load/helpers.py
@@ -24,8 +24,6 @@
     "save_completed_stages",
 ]
 
-RECIPE_FILE_NAME = "recipe.yaml"
-
 
 def log_model_load(
     model: Module, model_name_or_path: str, model_type: str, delayed_load: bool
@@ -106,6 +104,8 @@ def save_model_and_recipe(
     :param save_safetensors: whether to save as safetensors or pickle (bin)
     :param save_compressed: whether to compress sparse weights on disk
     """
+    # avoid circular import
+    from llmcompressor.transformers.utils.helpers import RECIPE_FILE_NAME
 
     model.save_pretrained(
         save_path, save_compressed=save_compressed, safe_serialization=save_safetensors

diff --git a/src/llmcompressor/transformers/finetune/session_mixin.py b/src/llmcompressor/transformers/finetune/session_mixin.py
@@ -24,8 +24,9 @@
 from llmcompressor.modifiers.distillation.utils.pytorch.model_wrapper import (
     KDModelWrapper,
 )
-from llmcompressor.pytorch.model_load.helpers import RECIPE_FILE_NAME, get_session_model
+from llmcompressor.pytorch.model_load.helpers import get_session_model
 from llmcompressor.pytorch.utils import ModuleSparsificationInfo
+from llmcompressor.transformers import RECIPE_FILE_NAME
 from llmcompressor.transformers.finetune.callbacks import (
     DisableHalfPrecisionCallback,
     TrainingLoopCallbacks,

diff --git a/src/llmcompressor/transformers/sparsification/compressed_tensors_utils.py b/src/llmcompressor/transformers/sparsification/compressed_tensors_utils.py
@@ -24,6 +24,7 @@
 from llmcompressor.transformers.compression.sparsity_config import (
     SparsityConfigMetadata,
 )
+from llmcompressor.transformers.utils import RECIPE_FILE_NAME
 from llmcompressor.utils.fsdp.helpers import (
     find_and_move_state_dicts_to_cpu,
     unwrap_and_export_model,
@@ -189,7 +190,7 @@ def skip(*args, **kwargs):
                 )
                 compressor.update_config(save_directory)
 
-            recipe_path = os.path.join(save_directory, "recipe.yaml")
+            recipe_path = os.path.join(save_directory, RECIPE_FILE_NAME)
             session = active_session()
 
             if (recipe_yaml_str := session.get_serialized_recipe()) is not None: