Set Sparse compression to save_compressed

vllm-project · Oct 6, 2024 · da4e0dc · da4e0dc
1 parent 7c2ab3a
commit da4e0dc
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/llmcompressor/transformers/sparsification/compressed_tensors_utils.py b/src/llmcompressor/transformers/sparsification/compressed_tensors_utils.py
@@ -95,7 +95,7 @@ def save_pretrained_wrapper(
                     "skip_compression_stats=True"
                 )
                 sparsity_config = SparsityConfigMetadata.from_pretrained(
-                    model, state_dict=state_dict, compress=False
+                    model, state_dict=state_dict, compress=save_compressed
                 )
 
             quantization_format = infer_quantization_format(