From 6be00e8a394e34d9ce021d0bd460902bfab57fe8 Mon Sep 17 00:00:00 2001 From: luxingyu2023 Date: Sun, 27 Apr 2025 21:09:28 +0800 Subject: [PATCH] fix bug when patching gptmodel in mindspore adaptor --- mindspeed_llm/mindspore/mindspore_adaptor.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mindspeed_llm/mindspore/mindspore_adaptor.py b/mindspeed_llm/mindspore/mindspore_adaptor.py index 6586c2c3a..3d039c790 100644 --- a/mindspeed_llm/mindspore/mindspore_adaptor.py +++ b/mindspeed_llm/mindspore/mindspore_adaptor.py @@ -14,7 +14,7 @@ class MindSporeAdaptation(MegatronAdaptationABC): from mindspeed.mindspore.core.data_parallel.distributed_data_parallel import distributed_data_parallel_init_with_cp from mindspeed.mindspore.core.transformer.moe.experts import groupedmlp_init_wrapper, groupedmlp_forward - MegatronAdaptation.register('megatron.core.models.gpt.gpt_model.GPTModel', GPTModel) + MegatronAdaptation.register('megatron.core.models.gpt.gpt_model.GPTModel', GPTModel, force_patch=True) MegatronAdaptation.register('megatron.core.distributed.distributed_data_parallel.DistributedDataParallel.__init__', distributed_data_parallel_init_with_cp, force_patch=True) MegatronAdaptation.register('megatron.core.transformer.moe.moe_layer.MoELayer.__init__', -- Gitee