ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

[NPU] Fix minicpm-2B error (#12527)

Browse source
This commit is contained in:
binbin Deng 2024-12-11 16:49:32 +08:00 committed by GitHub
parent fd9cf767ed
commit 509bdb4661
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 3 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
python/llm/src/ipex_llm/transformers/npu_pipeline_model/convert_pipeline.py
View file

@ -435,8 +435,10 @@ def convert_llm_for_deploy(model: torch.nn.Module,
layernorm_const = os.environ.get("IPEX_LLM_NPU_LAYERNORM_CONST", "1") == "1" layernorm_const = os.environ.get("IPEX_LLM_NPU_LAYERNORM_CONST", "1") == "1"
lm_head_low_bit = getattr(model.config, "bigdl_transformers_low_bit", "sym_int4_rtn") lm_head_low_bit = getattr(model.config, "bigdl_transformers_low_bit", "sym_int4_rtn")
if not isinstance(model.lm_head, SlicedLMHead): if hasattr(model, "lm_head") and not isinstance(model.lm_head, SlicedLMHead):
lm_head_low_bit = model.lm_head.qtype lm_head_low_bit = model.lm_head.qtype
elif hasattr(model, "lm_head_0") and not isinstance(model.lm_head_0, SlicedLMHead):
lm_head_low_bit = model.lm_head_0.qtype
else: else:
lm_head_low_bit = model.lm_head.lm_heads[0].qtype lm_head_low_bit = model.lm_head.lm_heads[0].qtype