PR #8810: Fix import of get_gpt_layer_ammo_spec

Signed-off-by: Jaemin Choi <[email protected]>
NVIDIA · Apr 9, 2024 · 938005f · 938005f
1 parent 95124e7
commit 938005f
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/nemo/collections/nlp/models/language_modeling/megatron_gpt_model.py b/nemo/collections/nlp/models/language_modeling/megatron_gpt_model.py
@@ -92,7 +92,7 @@
     from megatron.core import InferenceParams, parallel_state, tensor_parallel
     from megatron.core.datasets.blended_megatron_dataset_builder import BlendedMegatronDatasetBuilder
     from megatron.core.datasets.gpt_dataset import GPTDataset, GPTDatasetConfig, MockGPTDataset
-    from megatron.core.deploy.gpt.model_specs import get_gpt_layer_ammo_spec
+    from megatron.core.inference.gpt.model_specs import get_gpt_layer_ammo_spec
     from megatron.core.models.gpt import GPTModel as MCoreGPTModel
     from megatron.core.models.gpt.gpt_layer_specs import (
         get_gpt_layer_local_spec,